1 Commits

Author SHA1 Message Date
Per Stark 8fe4ac9fec release: 1.0.0 2026-01-11 18:37:07 +01:00
278 changed files with 11976 additions and 24843 deletions
+49
View File
@@ -0,0 +1,49 @@
- name: Prepare lib dir
run: mkdir -p lib
# Linux
- name: Fetch ONNX Runtime (Linux)
if: runner.os == 'Linux'
env:
ORT_VER: 1.22.0
run: |
set -euo pipefail
ARCH="$(uname -m)"
case "$ARCH" in
x86_64) URL="https://github.com/microsoft/onnxruntime/releases/download/v${ORT_VER}/onnxruntime-linux-x64-${ORT_VER}.tgz" ;;
aarch64) URL="https://github.com/microsoft/onnxruntime/releases/download/v${ORT_VER}/onnxruntime-linux-aarch64-${ORT_VER}.tgz" ;;
*) echo "Unsupported arch $ARCH"; exit 1 ;;
esac
curl -fsSL -o ort.tgz "$URL"
tar -xzf ort.tgz
cp -v onnxruntime-*/lib/libonnxruntime.so* lib/
# macOS
- name: Fetch ONNX Runtime (macOS)
if: runner.os == 'macOS'
env:
ORT_VER: 1.22.0
run: |
set -euo pipefail
curl -fsSL -o ort.tgz "https://github.com/microsoft/onnxruntime/releases/download/v${ORT_VER}/onnxruntime-osx-universal2-${ORT_VER}.tgz"
tar -xzf ort.tgz
# copy the main dylib; rename to stable name if needed
cp -v onnxruntime-*/lib/libonnxruntime*.dylib lib/
# optional: ensure a stable name
if [ ! -f lib/libonnxruntime.dylib ]; then
cp -v lib/libonnxruntime*.dylib lib/libonnxruntime.dylib
fi
# Windows
- name: Fetch ONNX Runtime (Windows)
if: runner.os == 'Windows'
shell: pwsh
env:
ORT_VER: 1.22.0
run: |
$url = "https://github.com/microsoft/onnxruntime/releases/download/v$env:ORT_VER/onnxruntime-win-x64-$env:ORT_VER.zip"
Invoke-WebRequest $url -OutFile ort.zip
Expand-Archive ort.zip -DestinationPath ort
$dll = Get-ChildItem -Recurse -Path ort -Filter onnxruntime.dll | Select-Object -First 1
Copy-Item $dll.FullName lib\onnxruntime.dll
+7 -14
View File
@@ -24,18 +24,9 @@ jobs:
with:
submodules: recursive
- name: Install Nix
uses: cachix/install-nix-action@v27
with:
extra_nix_config: |
experimental-features = nix-command flakes
- name: Verify ort-version matches nixpkgs onnxruntime
run: nix flake check --system x86_64-linux -L
- name: Install dist
shell: bash
run: "curl --proto '=https' --tlsv1.2 -LsSf https://github.com/axodotdev/cargo-dist/releases/download/v0.30.3/cargo-dist-installer.sh | sh"
run: "curl --proto '=https' --tlsv1.2 -LsSf https://github.com/axodotdev/cargo-dist/releases/download/v0.30.0/cargo-dist-installer.sh | sh"
- name: Cache dist
uses: actions/upload-artifact@v4
@@ -76,10 +67,6 @@ jobs:
with:
submodules: recursive
- name: Load ONNX Runtime version
shell: bash
run: echo "ORT_VER=$(tr -d '[:space:]' < ort-version)" >> "$GITHUB_ENV"
- name: Install Rust non-interactively if not already installed
if: ${{ matrix.container }}
run: |
@@ -120,6 +107,8 @@ jobs:
- name: Fetch ONNX Runtime (Linux)
if: runner.os == 'Linux'
env:
ORT_VER: 1.22.0
run: |
set -euo pipefail
ARCH="$(uname -m)"
@@ -136,6 +125,8 @@ jobs:
- name: Fetch ONNX Runtime (macOS)
if: runner.os == 'macOS'
env:
ORT_VER: 1.22.0
run: |
set -euo pipefail
curl -fsSL -o ort.tgz "https://github.com/microsoft/onnxruntime/releases/download/v${ORT_VER}/onnxruntime-osx-universal2-${ORT_VER}.tgz"
@@ -146,6 +137,8 @@ jobs:
- name: Fetch ONNX Runtime (Windows)
if: runner.os == 'Windows'
shell: pwsh
env:
ORT_VER: 1.22.0
run: |
$url = "https://github.com/microsoft/onnxruntime/releases/download/v$env:ORT_VER/onnxruntime-win-x64-$env:ORT_VER.zip"
Invoke-WebRequest $url -OutFile ort.zip
-4
View File
@@ -25,7 +25,3 @@ devenv.local.nix
# html-router/assets/style.css
html-router/node_modules
.fastembed_cache/
# insta: pending (unreviewed) snapshots; accepted *.snap files are committed
*.snap.new
.insta.bak
-24
View File
@@ -1,28 +1,4 @@
# Changelog
## Unreleased
## 1.0.3 (2026-06-12)
- Search: filter results by type — knowledge entities, ingested content, or both
- Admin: choose the local FastEmbed model from the admin UI; changes save immediately and apply after restart (re-embeds when the vector dimension changes)
- Performance: pooled FastEmbed workers and batched embedding generation for faster ingestion and search
- Performance: lower search and chat latency from backend allocation and retrieval optimizations
- Fix: modal dialogs (scratchpad editor, admin prompts, entity creation) open and close more reliably
- Fix: improved knowledge-entity relationship suggestions when creating entities manually
- Fix: API key revocation now correctly clears the stored key
## 1.0.2 (2026-02-15)
- Fix: edge case where navigation back to a chat page could trigger a new response generation
- Fix: chat references now validate and render more reliably
- Fix: improved admin access checks for restricted routes
- Performance: faster chat sidebar loads from cached conversation archive data
- API: harmonized ingest endpoint naming and added configurable ingest safety limits
- Security: hardened query handling and ingestion logging to reduce injection and data exposure risk
## 1.0.1 (2026-02-11)
- Shipped an S3 storage backend so content can be stored in object storage instead of local disk, with configuration support for S3 deployments.
- Introduced user theme preferences with the new Obsidian Prism look and improved dark mode styling.
- Fixed edge cases, including content deletion behavior and compatibility for older user records.
## 1.0.0 (2026-01-02)
- **Locally generated embeddings are now default**. If you want to continue using API embeddings, set EMBEDDING_BACKEND to openai. This will download a ONNX model and recreate all embeddings. But in most instances it's very worth it. Removing the network bound call to create embeddings. Creating embeddings on my N100 device is extremely fast. Typically a search response is provided in less than 50ms.
- Added a benchmarks create for evaluating the retrieval process
Generated
+1593 -1679
View File
File diff suppressed because it is too large Load Diff
+5 -10
View File
@@ -40,7 +40,7 @@ serde_json = "1.0.128"
serde = { version = "1", features = ["derive"] }
sha2 = "0.10.8"
surrealdb-migrations = "2.2.2"
surrealdb = { version = "2" }
surrealdb = { version = "2", features = ["kv-mem"] }
tempfile = "3.12.0"
text-splitter = { version = "0.18.1", features = ["markdown", "tokenizers"] }
tokenizers = { version = "0.20.4", features = ["http"] }
@@ -56,20 +56,15 @@ url = { version = "2.5.2", features = ["serde"] }
uuid = { version = "1.10.0", features = ["v4", "serde"] }
tokio-retry = "0.3.0"
base64 = "0.22.1"
object_store = { version = "0.11.2", features = ["aws"] }
object_store = { version = "0.11.2" }
bytes = "1.7.1"
state-machines = "0.9"
pdf-extract = "0.9"
lopdf = "0.32"
state-machines = "0.2.0"
fastembed = { version = "5.2.0", default-features = false, features = ["hf-hub-native-tls", "ort-load-dynamic"] }
[profile.dist]
inherits = "release"
lto = "thin"
[workspace.lints.rust]
unexpected_cfgs = { level = "warn", check-cfg = ["cfg(feature, values(\"inspect\"))"] }
[workspace.lints.clippy]
# Performance-focused lints
perf = { level = "warn", priority = -1 }
@@ -111,11 +106,11 @@ missing_errors_doc = "allow"
missing_panics_doc = "warn"
module_name_repetitions = "warn"
wildcard_dependencies = "warn"
missing_docs_in_private_items = "allow"
missing_docs_in_private_items = "warn"
# Allow noisy lints that don't add value for this project
needless_raw_string_hashes = "allow"
multiple_bound_locations = "allow"
cargo_common_metadata = "allow"
multiple-crate-versions = "allow"
module_name_repetition = "allow"
+4 -6
View File
@@ -1,5 +1,5 @@
# === Builder ===
FROM rust:1.91.1-bookworm AS builder
FROM rust:1.86-bookworm AS builder
WORKDIR /usr/src/minne
RUN apt-get update && apt-get install -y --no-install-recommends \
pkg-config clang cmake git && rm -rf /var/lib/apt/lists/*
@@ -30,11 +30,9 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
libgomp1 libstdc++6 curl \
&& rm -rf /var/lib/apt/lists/*
# ONNX Runtime (CPU). Version is read from ort-version (override with --build-arg ORT_VERSION=...).
COPY ort-version /tmp/ort-version
ARG ORT_VERSION
RUN ORT_VERSION="${ORT_VERSION:-$(tr -d '[:space:]' < /tmp/ort-version)}" && \
mkdir -p /opt/onnxruntime && \
# ONNX Runtime (CPU). Change if you bump ort.
ARG ORT_VERSION=1.22.0
RUN mkdir -p /opt/onnxruntime && \
curl -fsSL -o /tmp/ort.tgz \
"https://github.com/microsoft/onnxruntime/releases/download/v${ORT_VERSION}/onnxruntime-linux-x64-${ORT_VERSION}.tgz" && \
tar -xzf /tmp/ort.tgz -C /opt/onnxruntime --strip-components=1 && rm /tmp/ort.tgz
+31 -230
View File
@@ -1,265 +1,66 @@
# Minne - A Graph-Powered Personal Knowledge Base
# Minne
**Minne (Swedish for "memory")** is a personal knowledge management system and save-for-later application for capturing, organizing, and accessing your information. Inspired by the Zettelkasten method, it uses a graph database to automatically create connections between your notes without manual linking overhead.
**A graph-powered personal knowledge base that makes storing easy.**
Capture content effortlessly, let AI discover connections, and explore your knowledge visually. Self-hosted and privacy-focused.
[![Release Status](https://github.com/perstarkse/minne/actions/workflows/release.yml/badge.svg)](https://github.com/perstarkse/minne/actions/workflows/release.yml)
[![License: AGPL v3](https://img.shields.io/badge/License-AGPL_v3-blue.svg)](https://www.gnu.org/licenses/agpl-3.0)
[![Latest Release](https://img.shields.io/github/v/release/perstarkse/minne?sort=semver)](https://github.com/perstarkse/minne/releases/latest)
![Screenshot](screenshot-graph.webp)
![Screenshot](./screenshot-graph.webp)
## Demo deployment
## Try It
To test _Minne_ out, enter [this](https://minne.stark.pub) and sign in to a read-only demo deployment to view and test functionality out.
## Noteworthy Features
- **Search & Chat Interface** - Find content or knowledge instantly with full-text search, or use the chat mode and conversational AI to find and reason about content
- **Manual and AI-assisted connections** - Build entities and relationships manually with full control, let AI create entities and relationships automatically, or blend both approaches with AI suggestions for manual approval
- **Hybrid Retrieval System** - Search combining vector similarity & full-text search
- **Scratchpad Feature** - Quickly capture thoughts and convert them to permanent content when ready
- **Visual Graph Explorer** - Interactive D3-based navigation of your knowledge entities and connections
- **Multi-Format Support** - Ingest text, URLs, PDFs, audio files, and images into your knowledge base
- **Performance Focus** - Built with Rust and server-side rendering for speed and efficiency
- **Self-Hosted & Privacy-Focused** - Full control over your data, and compatible with any OpenAI-compatible API that supports structured outputs
## The "Why" Behind Minne
For a while I've been fascinated by personal knowledge management systems. I wanted something that made it incredibly easy to capture content - snippets of text, URLs, and other media - while automatically discovering connections between ideas. But I also wanted to maintain control over my knowledge structure.
Traditional tools like Logseq and Obsidian are excellent, but the manual linking process often became a hindrance. Meanwhile, fully automated systems sometimes miss important context or create relationships I wouldn't have chosen myself.
So I built Minne to offer the best of both worlds: effortless content capture with AI-assisted relationship discovery, but with the flexibility to manually curate, edit, or override any connections. You can let AI handle the heavy lifting of extracting entities and finding relationships, take full control yourself, or use a hybrid approach where AI suggests connections that you can approve or modify.
While developing Minne, I discovered [KaraKeep](https://github.com/karakeep-app/karakeep) (formerly Hoarder), which is an excellent application in a similar space you probably want to check it out! However, if you're interested in a PKM that offers both intelligent automation and manual curation, with the ability to chat with your knowledge base, then Minne might be worth testing.
## Table of Contents
- [Quick Start](#quick-start)
- [Features in Detail](#features-in-detail)
- [Configuration](#configuration)
- [Tech Stack](#tech-stack)
- [Application Architecture](#application-architecture)
- [AI Configuration](#ai-configuration--model-selection)
- [Roadmap](#roadmap)
- [Development](#development)
- [Contributing](#contributing)
- [License](#license)
**[Live Demo](https://minne-demo.stark.pub)** — Read-only demo deployment
## Quick Start
The fastest way to get Minne running is with Docker Compose:
```bash
# Clone the repository
git clone https://github.com/perstarkse/minne.git
cd minne
# Start Minne and its database
# Set your OpenAI API key in docker-compose.yml, then:
docker compose up -d
# Access at http://localhost:3000
# Open http://localhost:3000
```
**Required Setup:**
- Replace `your_openai_api_key_here` in `docker-compose.yml` with your actual API key
- Configure `OPENAI_BASE_URL` if using a custom AI provider (like Ollama)
For detailed installation options, see [Configuration](#configuration).
## Features in Detail
### Search vs. Chat mode
**Search** - Use when you know roughly what you're looking for. Full-text search finds items quickly by matching your query terms.
**Chat Mode** - Use when you want to explore concepts, find connections, or reason about your knowledge. The AI analyzes your query and finds relevant context across your entire knowledge base.
### Content Processing
Minne automatically processes content you save:
1. **Web scraping** extracts readable text from URLs
2. **Text analysis** identifies key concepts and relationships
3. **Graph creation** builds connections between related content
4. **Embedding generation** enables semantic search capabilities
### Visual Knowledge Graph
Explore your knowledge as an interactive network with flexible curation options:
**Manual Curation** - Create knowledge entities and relationships yourself with full control over your graph structure
**AI Automation** - Let AI automatically extract entities and discover relationships from your content
**Hybrid Approach** - Get AI-suggested relationships and entities that you can manually review, edit, or approve
The graph visualization shows:
- Knowledge entities as nodes (manually created or AI-extracted)
- Relationships as connections (manually defined, AI-discovered, or suggested)
- Interactive navigation for discovery and editing
### Optional FastEmbed Reranking
Minne ships with an opt-in reranking stage powered by [fastembed-rs](https://github.com/Anush008/fastembed-rs). When enabled, the hybrid retrieval results are rescored with a lightweight cross-encoder before being returned to chat or ingestion flows. In practice this often means more relevant results, boosting answer quality and downstream enrichment.
⚠️ **Resource notes**
- Enabling reranking downloads and caches ~1.1GB of model data on first startup (cached under `<data_dir>/fastembed/reranker` by default).
- Initialization takes longer while warming the cache, and each query consumes extra CPU. The default pool size (2) is tuned for a singe user setup, but could work with a pool size on 1 as well.
- The feature is disabled by default. Set `reranking_enabled: true` (or `RERANKING_ENABLED=true`) if youre comfortable with the additional footprint.
Example configuration:
```yaml
reranking_enabled: true
reranking_pool_size: 2
fastembed_cache_dir: "/var/lib/minne/fastembed" # optional override, defaults to .fastembed_cache
```
## Tech Stack
- **Backend:** Rust with Axum framework and Server-Side Rendering (SSR)
- **Frontend:** HTML with HTMX and minimal JavaScript for interactivity
- **Database:** SurrealDB (graph, document, and vector search)
- **AI Integration:** OpenAI-compatible API with structured outputs
- **Web Processing:** Headless Chrome for robust webpage content extraction
## Configuration
Minne can be configured using environment variables or a `config.yaml` file. Environment variables take precedence over `config.yaml`.
### Required Configuration
- `SURREALDB_ADDRESS`: WebSocket address of your SurrealDB instance (e.g., `ws://127.0.0.1:8000`)
- `SURREALDB_USERNAME`: Username for SurrealDB (e.g., `root_user`)
- `SURREALDB_PASSWORD`: Password for SurrealDB (e.g., `root_password`)
- `SURREALDB_DATABASE`: Database name in SurrealDB (e.g., `minne_db`)
- `SURREALDB_NAMESPACE`: Namespace in SurrealDB (e.g., `minne_ns`)
- `OPENAI_API_KEY`: Your API key for OpenAI compatible endpoint
- `HTTP_PORT`: Port for the Minne server (Default: `3000`)
### Optional Configuration
- `RUST_LOG`: Controls logging level (e.g., `minne=info,tower_http=debug`)
- `DATA_DIR`: Directory to store local data (e.g., `./data`)
- `OPENAI_BASE_URL`: Base URL for custom AI providers (like Ollama)
- `RERANKING_ENABLED` / `reranking_enabled`: Set to `true` to enable the FastEmbed reranking stage (default `false`)
- `RERANKING_POOL_SIZE` / `reranking_pool_size`: Maximum concurrent reranker workers (defaults to `2`)
- `FASTEMBED_CACHE_DIR` / `fastembed_cache_dir`: Directory for cached FastEmbed models (defaults to `<data_dir>/fastembed/reranker`)
- `FASTEMBED_SHOW_DOWNLOAD_PROGRESS` / `fastembed_show_download_progress`: Show model download progress when warming the cache (default `true`)
### Example config.yaml
```yaml
surrealdb_address: "ws://127.0.0.1:8000"
surrealdb_username: "root_user"
surrealdb_password: "root_password"
surrealdb_database: "minne_db"
surrealdb_namespace: "minne_ns"
openai_api_key: "sk-YourActualOpenAIKeyGoesHere"
data_dir: "./minne_app_data"
http_port: 3000
# rust_log: "info"
```
## Installation Options
### 1. Docker Compose (Recommended)
```bash
# Clone and run
git clone https://github.com/perstarkse/minne.git
cd minne
docker compose up -d
```
The included `docker-compose.yml` handles SurrealDB and Chromium dependencies automatically.
### 2. Nix
Or with Nix (with environment variables set):
```bash
nix run 'github:perstarkse/minne#main'
```
This fetches Minne and all dependencies, including Chromium.
Pre-built binaries for Windows, macOS, and Linux are available on the [Releases](https://github.com/perstarkse/minne/releases/latest) page.
### 3. Pre-built Binaries
## Features
Download binaries for Windows, macOS, and Linux from the [GitHub Releases](https://github.com/perstarkse/minne/releases/latest).
- **Fast** — Rust backend with server-side rendering and HTMX for snappy interactions
- **Search & Chat** — Search or use conversational AI to find and reason about content
- **Knowledge Graph** — Visual exploration with automatic or manual relationship curation
- **Hybrid Retrieval** — Vector similarity + full-text for relevant results
- **Multi-Format** — Ingest text, URLs, PDFs, audio, and images
- **Self-Hosted** — Your data, your server, any OpenAI-compatible API
**Requirements:** You'll need to provide SurrealDB and Chromium separately.
## Documentation
### 4. Build from Source
| Guide | Description |
|-------|-------------|
| [Installation](docs/installation.md) | Docker, Nix, binaries, source builds |
| [Configuration](docs/configuration.md) | Environment variables, config.yaml, AI setup |
| [Features](docs/features.md) | Search, Chat, Graph, Reranking, Ingestion |
| [Architecture](docs/architecture.md) | Tech stack, crate structure, data flow |
| [Vision](docs/vision.md) | Philosophy, roadmap, related projects |
```bash
git clone https://github.com/perstarkse/minne.git
cd minne
cargo run --release --bin main
```
## Tech Stack
**Requirements:** SurrealDB and Chromium must be installed and accessible in your PATH.
## Application Architecture
Minne offers flexible deployment options:
- **`main`**: Combined server and worker in one process (recommended for most users)
- **`server`**: Web interface and API only
- **`worker`**: Background processing only (for resource optimization)
## Usage
Once Minne is running at `http://localhost:3000`:
1. **Web Interface**: Full-featured experience for desktop and mobile
2. **iOS Shortcut**: Use the [Minne iOS Shortcut](https://www.icloud.com/shortcuts/e433fbd7602f4e2eaa70dca162323477) for quick content capture
3. **Content Types**: Save notes, URLs, audio files, and more
4. **Knowledge Graph**: Explore automatic connections between your content
5. **Chat Interface**: Query your knowledge base conversationally
## AI Configuration & Model Selection
### Setting Up AI Providers
Minne uses OpenAI-compatible APIs. Configure via environment variables or `config.yaml`:
- `OPENAI_API_KEY` (required): Your API key
- `OPENAI_BASE_URL` (optional): Custom provider URL (e.g., Ollama: `http://localhost:11434/v1`)
### Model Selection
1. Access the `/admin` page in your Minne instance
2. Select models for content processing and chat from your configured provider
3. **Content Processing Requirements**: The model must support structured outputs
4. **Embedding Dimensions**: Update this setting when changing embedding models (e.g., 1536 for `text-embedding-3-small`, 768 for `nomic-embed-text`)
## Roadmap
Current development focus:
- TUI frontend with system editor integration
- Enhanced reranking for improved retrieval recall
- Additional content type support
Feature requests and contributions are welcome!
## Development
```bash
# Run tests
cargo test
# Development build
cargo build
# Comprehensive linting
cargo clippy --workspace --all-targets --all-features
```
The codebase includes extensive unit tests. Integration tests and additional contributions are welcome.
Rust • Axum • HTMX • SurrealDB • FastEmbed
## Contributing
I've developed Minne primarily for my own use, but having been in the selfhosted space for a long time, and using the efforts by others, I thought I'd share with the community. Feature requests are welcome.
Feature requests and contributions welcome. See [Vision](docs/vision.md) for roadmap.
## License
Minne is licensed under the **GNU Affero General Public License v3.0 (AGPL-3.0)**. See the [LICENSE](LICENSE) file for details.
[AGPL-3.0](LICENSE)
-5
View File
@@ -20,8 +20,3 @@ futures = { workspace = true }
axum_typed_multipart = { workspace = true}
common = { path = "../common" }
[dev-dependencies]
common = { path = "../common", features = ["test-utils"] }
tower = "0.5"
uuid = { workspace = true }
+28
View File
@@ -11,3 +11,31 @@ pub struct ApiState {
pub config: AppConfig,
pub storage: StorageManager,
}
impl ApiState {
pub async fn new(
config: &AppConfig,
storage: StorageManager,
) -> Result<Self, Box<dyn std::error::Error>> {
let surreal_db_client = Arc::new(
SurrealDbClient::new(
&config.surrealdb_address,
&config.surrealdb_username,
&config.surrealdb_password,
&config.surrealdb_namespace,
&config.surrealdb_database,
)
.await?,
);
surreal_db_client.apply_migrations().await?;
let app_state = Self {
db: surreal_db_client.clone(),
config: config.clone(),
storage,
};
Ok(app_state)
}
}
+34 -59
View File
@@ -7,38 +7,36 @@ use common::error::AppError;
use serde::Serialize;
use thiserror::Error;
#[derive(Error, Debug)]
pub enum ApiErr {
#[error("internal server error")]
#[derive(Error, Debug, Serialize, Clone)]
pub enum ApiError {
#[error("Internal server error")]
InternalError(String),
#[error("validation error: {0}")]
#[error("Validation error: {0}")]
ValidationError(String),
#[error("not found: {0}")]
#[error("Not found: {0}")]
NotFound(String),
#[error("unauthorized: {0}")]
#[error("Unauthorized: {0}")]
Unauthorized(String),
#[error("payload too large: {0}")]
PayloadTooLarge(String),
}
impl From<AppError> for ApiErr {
impl From<AppError> for ApiError {
fn from(err: AppError) -> Self {
match err {
AppError::Database(_) | AppError::OpenAI(_) => {
tracing::error!("Internal error: {:?}", err);
Self::InternalError("Internal server error".to_string())
}
AppError::NotFound(msg) => Self::NotFound(msg),
AppError::Validation(msg) => Self::ValidationError(msg),
AppError::Auth(msg) => Self::Unauthorized(msg),
other => {
tracing::error!("internal API error: {other:?}");
Self::InternalError("Internal server error".to_string())
_ => Self::InternalError("Internal server error".to_string()),
}
}
}
}
impl IntoResponse for ApiErr {
impl IntoResponse for ApiError {
fn into_response(self) -> Response {
let (status, error_response) = match self {
Self::InternalError(message) => (
@@ -69,13 +67,6 @@ impl IntoResponse for ApiErr {
status: "error".to_string(),
},
),
Self::PayloadTooLarge(message) => (
StatusCode::PAYLOAD_TOO_LARGE,
ErrorResponse {
error: message,
status: "error".to_string(),
},
),
};
(status, Json(error_response)).into_response()
@@ -93,7 +84,6 @@ mod tests {
use super::*;
use common::error::AppError;
use std::fmt::Debug;
use std::io;
// Helper to check status code
fn assert_status_code<T: IntoResponse + Debug>(response: T, expected_status: StatusCode) {
@@ -105,58 +95,43 @@ mod tests {
fn test_app_error_to_api_error_conversion() {
// Test NotFound error conversion
let not_found = AppError::NotFound("resource not found".to_string());
let api_error = ApiErr::from(not_found);
assert!(matches!(api_error, ApiErr::NotFound(msg) if msg == "resource not found"));
let api_error = ApiError::from(not_found);
assert!(matches!(api_error, ApiError::NotFound(msg) if msg == "resource not found"));
// Test Validation error conversion
let validation = AppError::Validation("invalid input".to_string());
let api_error = ApiErr::from(validation);
assert!(matches!(api_error, ApiErr::ValidationError(msg) if msg == "invalid input"));
let api_error = ApiError::from(validation);
assert!(matches!(api_error, ApiError::ValidationError(msg) if msg == "invalid input"));
// Test Auth error conversion
let auth = AppError::Auth("unauthorized".to_string());
let api_error = ApiErr::from(auth);
assert!(matches!(api_error, ApiErr::Unauthorized(msg) if msg == "unauthorized"));
let api_error = ApiError::from(auth);
assert!(matches!(api_error, ApiError::Unauthorized(msg) if msg == "unauthorized"));
// Test for internal errors - create a mock error that doesn't require surrealdb
let internal_error = AppError::Io(io::Error::other("io error"));
let api_error = ApiErr::from(internal_error);
assert!(matches!(
api_error,
ApiErr::InternalError(msg) if msg == "Internal server error"
));
}
#[test]
fn test_app_error_internal_error_is_sanitized() {
let api_error = ApiErr::from(AppError::internal("db password incorrect"));
assert!(matches!(
api_error,
ApiErr::InternalError(msg) if msg == "Internal server error"
));
let internal_error =
AppError::Io(std::io::Error::new(std::io::ErrorKind::Other, "io error"));
let api_error = ApiError::from(internal_error);
assert!(matches!(api_error, ApiError::InternalError(_)));
}
#[test]
fn test_api_error_response_status_codes() {
// Test internal error status
let error = ApiErr::InternalError("server error".to_string());
let error = ApiError::InternalError("server error".to_string());
assert_status_code(error, StatusCode::INTERNAL_SERVER_ERROR);
// Test not found status
let error = ApiErr::NotFound("not found".to_string());
let error = ApiError::NotFound("not found".to_string());
assert_status_code(error, StatusCode::NOT_FOUND);
// Test validation error status
let error = ApiErr::ValidationError("invalid input".to_string());
let error = ApiError::ValidationError("invalid input".to_string());
assert_status_code(error, StatusCode::BAD_REQUEST);
// Test unauthorized status
let error = ApiErr::Unauthorized("not allowed".to_string());
let error = ApiError::Unauthorized("not allowed".to_string());
assert_status_code(error, StatusCode::UNAUTHORIZED);
// Test payload too large status
let error = ApiErr::PayloadTooLarge("too big".to_string());
assert_status_code(error, StatusCode::PAYLOAD_TOO_LARGE);
}
// Alternative approach that doesn't try to parse the response body
@@ -164,15 +139,15 @@ mod tests {
fn test_error_messages() {
// For validation errors
let message = "invalid data format";
let error = ApiErr::ValidationError(message.to_string());
let error = ApiError::ValidationError(message.to_string());
// Check that the error itself contains the message
assert_eq!(error.to_string(), format!("validation error: {message}"));
assert_eq!(error.to_string(), format!("Validation error: {}", message));
// For not found errors
let message = "user not found";
let error = ApiErr::NotFound(message.to_string());
assert_eq!(error.to_string(), format!("not found: {message}"));
let error = ApiError::NotFound(message.to_string());
assert_eq!(error.to_string(), format!("Not found: {}", message));
}
// Alternative approach for internal error test
@@ -181,11 +156,11 @@ mod tests {
// Create a sensitive error message
let sensitive_info = "db password incorrect";
// Create ApiErr with sensitive info
let api_error = ApiErr::InternalError(sensitive_info.to_string());
// Create ApiError with sensitive info
let api_error = ApiError::InternalError(sensitive_info.to_string());
// Check the error message is correctly set
assert_eq!(api_error.to_string(), "internal server error");
assert_eq!(api_error.to_string(), "Internal server error");
// Also verify correct status code
assert_status_code(api_error, StatusCode::INTERNAL_SERVER_ERROR);
+4 -8
View File
@@ -6,7 +6,7 @@ use axum::{
Router,
};
use middleware_api_auth::api_auth;
use routes::{categories::list, ingest::handle, liveness::live, readiness::ready};
use routes::{categories::get_categories, ingress::ingest_data, liveness::live, readiness::ready};
pub mod api_state;
pub mod error;
@@ -26,13 +26,9 @@ where
// Protected API endpoints (require auth)
let protected = Router::new()
.route(
"/ingest",
post(handle).layer(DefaultBodyLimit::max(
app_state.config.ingest_max_body_bytes,
)),
)
.route("/categories", get(list))
.route("/ingress", post(ingest_data))
.route("/categories", get(get_categories))
.layer(DefaultBodyLimit::max(1024 * 1024 * 1024))
.route_layer(from_fn_with_state(app_state.clone(), api_auth));
public.merge(protected)
+8 -68
View File
@@ -6,26 +6,26 @@ use axum::{
use common::storage::types::user::User;
use crate::{api_state::ApiState, error::ApiErr};
use crate::{api_state::ApiState, error::ApiError};
pub async fn api_auth(
State(state): State<ApiState>,
mut request: Request,
next: Next,
) -> Result<Response, ApiErr> {
) -> Result<Response, ApiError> {
let api_key = extract_api_key(&request)
.ok_or_else(|| ApiErr::Unauthorized("You have to be authenticated".to_string()))?;
.ok_or_else(|| ApiError::Unauthorized("You have to be authenticated".to_string()))?;
let user = User::find_by_api_key(api_key, &state.db).await?;
let user = User::find_by_api_key(&api_key, &state.db).await?;
let user =
user.ok_or_else(|| ApiErr::Unauthorized("You have to be authenticated".to_string()))?;
user.ok_or_else(|| ApiError::Unauthorized("You have to be authenticated".to_string()))?;
request.extensions_mut().insert(user);
Ok(next.run(request).await)
}
fn extract_api_key(request: &Request) -> Option<&str> {
fn extract_api_key(request: &Request) -> Option<String> {
request
.headers()
.get("X-API-Key")
@@ -35,67 +35,7 @@ fn extract_api_key(request: &Request) -> Option<&str> {
.headers()
.get("Authorization")
.and_then(|v| v.to_str().ok())
.and_then(|auth| auth.strip_prefix("Bearer "))
.map(str::trim)
.and_then(|auth| auth.strip_prefix("Bearer ").map(str::trim))
})
}
#[cfg(test)]
#[allow(clippy::expect_used)]
mod tests {
use axum::body::Body;
use axum::http::{HeaderValue, Request};
use super::extract_api_key;
fn request_with_headers(headers: &[(&str, &str)]) -> Request<Body> {
let mut builder = Request::builder().method("GET").uri("/");
for (name, value) in headers {
builder = builder.header(*name, *value);
}
builder.body(Body::empty()).expect("test request")
}
#[test]
fn extract_api_key_from_x_api_key_header() {
let request = request_with_headers(&[("X-API-Key", "sk_test_key")]);
assert_eq!(extract_api_key(&request), Some("sk_test_key"));
}
#[test]
fn extract_api_key_from_bearer_authorization() {
let request = request_with_headers(&[("Authorization", "Bearer sk_bearer_key")]);
assert_eq!(extract_api_key(&request), Some("sk_bearer_key"));
}
#[test]
fn extract_api_key_prefers_x_api_key_over_authorization() {
let request = request_with_headers(&[
("X-API-Key", "sk_header"),
("Authorization", "Bearer sk_bearer"),
]);
assert_eq!(extract_api_key(&request), Some("sk_header"));
}
#[test]
fn extract_api_key_returns_none_when_missing() {
let request = request_with_headers(&[]);
assert_eq!(extract_api_key(&request), None);
}
#[test]
fn extract_api_key_rejects_non_bearer_authorization() {
let request = request_with_headers(&[("Authorization", "Basic abc")]);
assert_eq!(extract_api_key(&request), None);
}
#[test]
fn extract_api_key_rejects_invalid_header_values() {
let mut request = request_with_headers(&[]);
request.headers_mut().insert(
"X-API-Key",
HeaderValue::from_bytes(&[0xFF]).expect("invalid header"),
);
assert_eq!(extract_api_key(&request), None);
}
.map(String::from)
}
+3 -3
View File
@@ -1,12 +1,12 @@
use axum::{extract::State, response::IntoResponse, Extension, Json};
use common::storage::types::user::User;
use crate::{api_state::ApiState, error::ApiErr};
use crate::{api_state::ApiState, error::ApiError};
pub async fn list(
pub async fn get_categories(
State(state): State<ApiState>,
Extension(user): Extension<User>,
) -> Result<impl IntoResponse, ApiErr> {
) -> Result<impl IntoResponse, ApiError> {
let categories = User::get_user_categories(&user.id, &state.db).await?;
Ok(Json(categories))
-79
View File
@@ -1,79 +0,0 @@
use axum::{extract::State, http::StatusCode, response::IntoResponse, Extension, Json};
use axum_typed_multipart::{FieldData, TryFromMultipart, TypedMultipart};
use common::{
error::AppError,
storage::types::{
file_info::FileInfo, ingestion_payload::IngestionPayload, ingestion_task::IngestionTask,
user::User,
},
utils::ingest_limits::{validate_ingest_input, IngestValidationError},
};
use futures::{future::try_join_all, TryFutureExt};
use serde_json::json;
use tempfile::NamedTempFile;
use tracing::info;
use crate::{api_state::ApiState, error::ApiErr};
#[derive(Debug, TryFromMultipart)]
pub struct Params {
pub content: Option<String>,
pub context: String,
pub category: String,
#[form_data(limit = "20000000")]
#[form_data(default)]
pub files: Vec<FieldData<NamedTempFile>>,
}
pub async fn handle(
State(state): State<ApiState>,
Extension(user): Extension<User>,
TypedMultipart(input): TypedMultipart<Params>,
) -> Result<impl IntoResponse, ApiErr> {
let user_id = user.id;
let has_content = input.content.as_ref().is_some_and(|c| !c.trim().is_empty());
match validate_ingest_input(
&state.config,
input.content.as_deref(),
&input.context,
&input.category,
input.files.len(),
) {
Ok(()) => {}
Err(IngestValidationError::PayloadTooLarge(message)) => {
return Err(ApiErr::PayloadTooLarge(message));
}
Err(IngestValidationError::BadRequest(message)) => {
return Err(ApiErr::ValidationError(message));
}
}
info!(
user_id = %user_id,
has_content,
content_len = input.content.as_ref().map_or(0, String::len),
context_len = input.context.len(),
category_len = input.category.len(),
file_count = input.files.len(),
"Received ingest request"
);
let file_infos = try_join_all(input.files.into_iter().map(|file| {
FileInfo::new_with_storage(file, &state.db, &user_id, &state.storage)
.map_err(AppError::from)
}))
.await?;
let payloads = IngestionPayload::create_ingestion_payload(
input.content,
input.context,
input.category,
file_infos,
user_id.clone(),
)?;
IngestionTask::create_all_and_add_to_db(payloads, &user_id, &state.db).await?;
Ok((StatusCode::OK, Json(json!({ "status": "success" }))))
}
+57
View File
@@ -0,0 +1,57 @@
use axum::{extract::State, http::StatusCode, response::IntoResponse, Extension, Json};
use axum_typed_multipart::{FieldData, TryFromMultipart, TypedMultipart};
use common::{
error::AppError,
storage::types::{
file_info::FileInfo, ingestion_payload::IngestionPayload, ingestion_task::IngestionTask,
user::User,
},
};
use futures::{future::try_join_all, TryFutureExt};
use serde_json::json;
use tempfile::NamedTempFile;
use tracing::info;
use crate::{api_state::ApiState, error::ApiError};
#[derive(Debug, TryFromMultipart)]
pub struct IngestParams {
pub content: Option<String>,
pub context: String,
pub category: String,
#[form_data(limit = "10000000")] // Adjust limit as needed
#[form_data(default)]
pub files: Vec<FieldData<NamedTempFile>>,
}
pub async fn ingest_data(
State(state): State<ApiState>,
Extension(user): Extension<User>,
TypedMultipart(input): TypedMultipart<IngestParams>,
) -> Result<impl IntoResponse, ApiError> {
info!("Received input: {:?}", input);
let user_id = user.id;
let file_infos = try_join_all(input.files.into_iter().map(|file| {
FileInfo::new_with_storage(file, &state.db, &user_id, &state.storage)
.map_err(AppError::from)
}))
.await?;
let payloads = IngestionPayload::create_ingestion_payload(
input.content,
input.context,
input.category,
file_infos,
&user_id,
)?;
let futures: Vec<_> = payloads
.into_iter()
.map(|object| IngestionTask::create_and_add_to_db(object, user_id.clone(), &state.db))
.collect();
try_join_all(futures).await?;
Ok((StatusCode::OK, Json(json!({ "status": "success" }))))
}
+1 -1
View File
@@ -1,4 +1,4 @@
pub mod categories;
pub mod ingest;
pub mod ingress;
pub mod liveness;
pub mod readiness;
+4 -7
View File
@@ -1,6 +1,5 @@
use axum::{extract::State, http::StatusCode, response::IntoResponse, Json};
use serde_json::json;
use tracing::error;
use crate::api_state::ApiState;
@@ -14,15 +13,13 @@ pub async fn ready(State(state): State<ApiState>) -> impl IntoResponse {
"checks": { "db": "ok" }
})),
),
Err(e) => {
error!("readiness check failed: {e:?}");
(
Err(e) => (
StatusCode::SERVICE_UNAVAILABLE,
Json(json!({
"status": "error",
"checks": { "db": "fail" }
"checks": { "db": "fail" },
"reason": e.to_string()
})),
)
}
),
}
}
-159
View File
@@ -1,159 +0,0 @@
#![allow(clippy::expect_used)]
use std::sync::Arc;
use api_router::{api_routes_v1, api_state::ApiState};
use axum::{
body::{to_bytes, Body},
http::{Request, StatusCode},
Router,
};
use common::{
storage::{db::SurrealDbClient, store::StorageManager, types::user::User},
utils::config::{AppConfig, StorageKind},
};
use tower::ServiceExt;
async fn build_test_app() -> (Router, Arc<SurrealDbClient>) {
let namespace = "api_router_test";
let database = uuid::Uuid::new_v4().to_string();
let db = Arc::new(
SurrealDbClient::memory(namespace, &database)
.await
.expect("in-memory db"),
);
db.apply_migrations()
.await
.expect("migrations should apply");
let config = AppConfig {
storage: StorageKind::Memory,
..Default::default()
};
let storage = StorageManager::new(&config).await.expect("storage manager");
let state = ApiState {
db: Arc::clone(&db),
config,
storage,
};
let router = api_routes_v1(&state).with_state(state);
(router, db)
}
async fn response_body(response: axum::response::Response) -> String {
let body = to_bytes(response.into_body(), usize::MAX)
.await
.expect("response body");
String::from_utf8(body.to_vec()).expect("utf-8 body")
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn live_probe_is_public() {
let (app, _db) = build_test_app().await;
let response = app
.clone()
.oneshot(
Request::builder()
.uri("/live")
.body(Body::empty())
.expect("live request"),
)
.await
.expect("live response");
assert_eq!(response.status(), StatusCode::OK);
assert!(response_body(response).await.contains("\"status\":\"ok\""));
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn ready_probe_is_public_and_reports_db_ok() {
let (app, _db) = build_test_app().await;
let response = app
.clone()
.oneshot(
Request::builder()
.uri("/ready")
.body(Body::empty())
.expect("ready request"),
)
.await
.expect("ready response");
assert_eq!(response.status(), StatusCode::OK);
let body = response_body(response).await;
assert!(body.contains("\"checks\":{\"db\":\"ok\"}") || body.contains("\"db\":\"ok\""));
assert!(!body.contains("reason"));
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn protected_route_requires_api_key() {
let (app, _db) = build_test_app().await;
let response = app
.clone()
.oneshot(
Request::builder()
.uri("/categories")
.body(Body::empty())
.expect("categories request"),
)
.await
.expect("categories response");
assert_eq!(response.status(), StatusCode::UNAUTHORIZED);
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn protected_route_rejects_invalid_api_key() {
let (app, _db) = build_test_app().await;
let response = app
.clone()
.oneshot(
Request::builder()
.uri("/categories")
.header("X-API-Key", "sk_invalid")
.body(Body::empty())
.expect("categories request"),
)
.await
.expect("categories response");
assert_eq!(response.status(), StatusCode::UNAUTHORIZED);
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn authenticated_user_can_list_categories() {
let (app, db) = build_test_app().await;
let user = User::create_new(
"api_router_test@example.com".to_string(),
"test_password".to_string(),
&db,
"UTC".to_string(),
"system".to_string(),
)
.await
.expect("test user");
let api_key = User::set_api_key(&user.id, &db).await.expect("api key");
let response = app
.clone()
.oneshot(
Request::builder()
.uri("/categories")
.header("X-API-Key", api_key)
.body(Body::empty())
.expect("categories request"),
)
.await
.expect("categories response");
assert_eq!(response.status(), StatusCode::OK);
}
+2 -5
View File
@@ -16,7 +16,7 @@ tracing = { workspace = true }
anyhow = { workspace = true }
thiserror = { workspace = true }
serde_json = { workspace = true }
surrealdb = { workspace = true }
surrealdb = { workspace = true, features = ["kv-mem"] }
async-openai = { workspace = true }
futures = { workspace = true }
tempfile = { workspace = true }
@@ -49,7 +49,4 @@ fastembed = { workspace = true }
[features]
test-utils = ["surrealdb/kv-mem"]
[dev-dependencies]
surrealdb = { workspace = true, features = ["kv-mem"] }
test-utils = []
@@ -1 +0,0 @@
DEFINE FIELD IF NOT EXISTS theme ON user TYPE string DEFAULT "system";
@@ -1,3 +0,0 @@
-- Per-user deduplication: same SHA256 may exist for different users.
REMOVE INDEX IF EXISTS file_sha256_idx ON file;
DEFINE INDEX IF NOT EXISTS file_user_sha256_idx ON file FIELDS user_id, sha256 UNIQUE;
@@ -1,33 +0,0 @@
-- Harden knowledge entity embeddings and graph storage invariants.
DEFINE FIELD IF NOT EXISTS source_id ON knowledge_entity_embedding TYPE string;
-- Backfill denormalized source_id from the linked entity.
FOR $emb IN (SELECT * FROM knowledge_entity_embedding WHERE source_id = NONE OR source_id = '') {
LET $entity = (SELECT source_id FROM $emb.entity_id)[0];
IF $entity != NONE {
UPDATE $emb.id SET source_id = $entity.source_id;
}
};
-- Re-key embeddings so record id matches entity id (stable 1:1 identity).
FOR $emb IN (SELECT * FROM knowledge_entity_embedding) {
LET $entity_key = record::id($emb.entity_id);
LET $canonical = type::thing('knowledge_entity_embedding', $entity_key);
IF $emb.id != $canonical {
UPSERT $canonical CONTENT {
entity_id: $emb.entity_id,
embedding: $emb.embedding,
user_id: $emb.user_id,
source_id: $emb.source_id,
created_at: $emb.created_at,
updated_at: $emb.updated_at
};
DELETE $emb.id;
}
};
REMOVE INDEX IF EXISTS knowledge_entity_embedding_entity_id_idx ON knowledge_entity_embedding;
DEFINE INDEX IF NOT EXISTS knowledge_entity_embedding_entity_id_idx ON knowledge_entity_embedding FIELDS entity_id UNIQUE;
DEFINE INDEX IF NOT EXISTS knowledge_entity_embedding_source_id_idx ON knowledge_entity_embedding FIELDS source_id;
DEFINE INDEX IF NOT EXISTS knowledge_entity_user_source_idx ON knowledge_entity FIELDS user_id, source_id;
@@ -1,21 +0,0 @@
-- Harden text chunk embeddings storage invariants.
-- Re-key embeddings so record id matches chunk id (stable 1:1 identity).
FOR $emb IN (SELECT * FROM text_chunk_embedding) {
LET $chunk_key = record::id($emb.chunk_id);
LET $canonical = type::thing('text_chunk_embedding', $chunk_key);
IF $emb.id != $canonical {
UPSERT $canonical CONTENT {
chunk_id: $emb.chunk_id,
embedding: $emb.embedding,
user_id: $emb.user_id,
source_id: $emb.source_id,
created_at: $emb.created_at,
updated_at: $emb.updated_at
};
DELETE $emb.id;
}
};
REMOVE INDEX IF EXISTS text_chunk_embedding_chunk_id_idx ON text_chunk_embedding;
DEFINE INDEX IF NOT EXISTS text_chunk_embedding_chunk_id_idx ON text_chunk_embedding FIELDS chunk_id UNIQUE;
@@ -1,8 +0,0 @@
-- Align persisted embedding settings when FastEmbed is the recorded backend but the model
-- name is still the OpenAI migration default (invalid for FastEmbed `from_str`).
UPDATE system_settings:current SET
embedding_model = 'Xenova/bge-small-en-v1.5',
embedding_dimensions = 384
WHERE embedding_backend = 'fastembed'
AND embedding_model = 'text-embedding-3-small';
@@ -1 +0,0 @@
{"schemas":"--- original\n+++ modified\n@@ -28,6 +28,7 @@\n # Add indexes based on query patterns (get_complete_conversation ownership check, get_user_conversations)\n DEFINE INDEX IF NOT EXISTS conversation_user_id_idx ON conversation FIELDS user_id;\n DEFINE INDEX IF NOT EXISTS conversation_created_at_idx ON conversation FIELDS created_at; # For get_user_conversations ORDER BY\n+DEFINE INDEX IF NOT EXISTS conversation_user_updated_at_idx ON conversation FIELDS user_id, updated_at; # For sidebar conversation projection ORDER BY\n\n # Defines the schema for the 'file' table (used by FileInfo).\n\n","events":null}
@@ -1 +0,0 @@
{"schemas":"--- original\n+++ modified\n@@ -45,9 +45,8 @@\n DEFINE FIELD IF NOT EXISTS mime_type ON file TYPE string;\n DEFINE FIELD IF NOT EXISTS user_id ON file TYPE string;\n\n-# Indexes based on usage (get_by_sha, potentially user lookups)\n-# Using UNIQUE based on the logic in FileInfo::new to prevent duplicates\n-DEFINE INDEX IF NOT EXISTS file_sha256_idx ON file FIELDS sha256 UNIQUE;\n+# Indexes based on usage (get_by_sha scoped by user_id, user lookups)\n+DEFINE INDEX IF NOT EXISTS file_user_sha256_idx ON file FIELDS user_id, sha256 UNIQUE;\n DEFINE INDEX IF NOT EXISTS file_user_id_idx ON file FIELDS user_id;\n\n # Defines the schema for the 'ingestion_task' table (used by IngestionTask).\n","events":null}
@@ -1 +0,0 @@
{"schemas":"--- original\n+++ modified\n@@ -68,7 +68,7 @@\n\n # Defines the schema for the 'knowledge_entity' table.\n\n-DEFINE TABLE IF NOT EXISTS knowledge_entity SCHEMALESS;\n+DEFINE TABLE IF NOT EXISTS knowledge_entity SCHEMAFULL;\n\n # Standard fields\n DEFINE FIELD IF NOT EXISTS created_at ON knowledge_entity TYPE datetime;\n@@ -90,6 +90,7 @@\n -- DEFINE INDEX IF NOT EXISTS idx_embedding_entities ON knowledge_entity FIELDS embedding HNSW DIMENSION 1536;\n DEFINE INDEX IF NOT EXISTS knowledge_entity_source_id_idx ON knowledge_entity FIELDS source_id;\n DEFINE INDEX IF NOT EXISTS knowledge_entity_user_id_idx ON knowledge_entity FIELDS user_id;\n+DEFINE INDEX IF NOT EXISTS knowledge_entity_user_source_idx ON knowledge_entity FIELDS user_id, source_id;\n DEFINE INDEX IF NOT EXISTS knowledge_entity_entity_type_idx ON knowledge_entity FIELDS entity_type;\n DEFINE INDEX IF NOT EXISTS knowledge_entity_created_at_idx ON knowledge_entity FIELDS created_at;\n\n@@ -102,6 +103,7 @@\n DEFINE FIELD IF NOT EXISTS created_at ON knowledge_entity_embedding TYPE datetime;\n DEFINE FIELD IF NOT EXISTS updated_at ON knowledge_entity_embedding TYPE datetime;\n DEFINE FIELD IF NOT EXISTS user_id ON knowledge_entity_embedding TYPE string;\n+DEFINE FIELD IF NOT EXISTS source_id ON knowledge_entity_embedding TYPE string;\n\n -- Custom fields\n DEFINE FIELD IF NOT EXISTS entity_id ON knowledge_entity_embedding TYPE record<knowledge_entity>;\n@@ -109,8 +111,9 @@\n\n -- Indexes\n -- DEFINE INDEX IF NOT EXISTS idx_embedding_knowledge_entity_embedding ON knowledge_entity_embedding FIELDS embedding HNSW DIMENSION 1536;\n-DEFINE INDEX IF NOT EXISTS knowledge_entity_embedding_entity_id_idx ON knowledge_entity_embedding FIELDS entity_id;\n+DEFINE INDEX IF NOT EXISTS knowledge_entity_embedding_entity_id_idx ON knowledge_entity_embedding FIELDS entity_id UNIQUE;\n DEFINE INDEX IF NOT EXISTS knowledge_entity_embedding_user_id_idx ON knowledge_entity_embedding FIELDS user_id;\n+DEFINE INDEX IF NOT EXISTS knowledge_entity_embedding_source_id_idx ON knowledge_entity_embedding FIELDS source_id;\n\n # Defines the schema for the 'message' table.\n\n@@ -135,19 +138,17 @@\n # Defines the 'relates_to' edge table for KnowledgeRelationships.\n # Edges connect nodes, in this case knowledge_entity records.\n\n-# Define the edge table itself, enforcing connections between knowledge_entity records\n-# SCHEMAFULL requires all fields to be defined, maybe start with SCHEMALESS if metadata might vary\n-DEFINE TABLE IF NOT EXISTS relates_to SCHEMALESS TYPE RELATION FROM knowledge_entity TO knowledge_entity;\n+DEFINE TABLE IF NOT EXISTS relates_to SCHEMAFULL TYPE RELATION FROM knowledge_entity TO knowledge_entity;\n+\n+DEFINE FIELD IF NOT EXISTS in ON relates_to TYPE record<knowledge_entity>;\n+DEFINE FIELD IF NOT EXISTS out ON relates_to TYPE record<knowledge_entity>;\n\n-# Define the metadata field within the edge\n # RelationshipMetadata is a struct, store as object\n DEFINE FIELD IF NOT EXISTS metadata ON relates_to TYPE object;\n+DEFINE FIELD IF NOT EXISTS metadata.user_id ON relates_to TYPE string;\n+DEFINE FIELD IF NOT EXISTS metadata.source_id ON relates_to TYPE string;\n+DEFINE FIELD IF NOT EXISTS metadata.relationship_type ON relates_to TYPE string;\n\n-# Optionally, define fields within the metadata object for stricter schema (requires SCHEMAFULL on table)\n-# DEFINE FIELD IF NOT EXISTS metadata.user_id ON relates_to TYPE string;\n-# DEFINE FIELD IF NOT EXISTS metadata.source_id ON relates_to TYPE string;\n-# DEFINE FIELD IF NOT EXISTS metadata.relationship_type ON relates_to TYPE string;\n-\n # Add indexes based on query patterns (delete_relationships_by_source_id, get_knowledge_relationships)\n DEFINE INDEX IF NOT EXISTS relates_to_metadata_source_id_idx ON relates_to FIELDS metadata.source_id;\n DEFINE INDEX IF NOT EXISTS relates_to_metadata_user_id_idx ON relates_to FIELDS metadata.user_id;\n","events":null}
@@ -1 +0,0 @@
{"schemas":"--- original\n+++ modified\n@@ -237,7 +237,7 @@\n\n -- Indexes\n -- DEFINE INDEX IF NOT EXISTS idx_embedding_text_chunk_embedding ON text_chunk_embedding FIELDS embedding HNSW DIMENSION 1536;\n-DEFINE INDEX IF NOT EXISTS text_chunk_embedding_chunk_id_idx ON text_chunk_embedding FIELDS chunk_id;\n+DEFINE INDEX IF NOT EXISTS text_chunk_embedding_chunk_id_idx ON text_chunk_embedding FIELDS chunk_id UNIQUE;\n DEFINE INDEX IF NOT EXISTS text_chunk_embedding_user_id_idx ON text_chunk_embedding FIELDS user_id;\n DEFINE INDEX IF NOT EXISTS text_chunk_embedding_source_id_idx ON text_chunk_embedding FIELDS source_id;\n\n","events":null}
-17
View File
@@ -1,17 +0,0 @@
# Defines the 'relates_to' edge table for KnowledgeRelationships.
# Edges connect nodes, in this case knowledge_entity records.
DEFINE TABLE IF NOT EXISTS relates_to SCHEMAFULL TYPE RELATION FROM knowledge_entity TO knowledge_entity;
DEFINE FIELD IF NOT EXISTS in ON relates_to TYPE record<knowledge_entity>;
DEFINE FIELD IF NOT EXISTS out ON relates_to TYPE record<knowledge_entity>;
# RelationshipMetadata is a struct, store as object
DEFINE FIELD IF NOT EXISTS metadata ON relates_to TYPE object;
DEFINE FIELD IF NOT EXISTS metadata.user_id ON relates_to TYPE string;
DEFINE FIELD IF NOT EXISTS metadata.source_id ON relates_to TYPE string;
DEFINE FIELD IF NOT EXISTS metadata.relationship_type ON relates_to TYPE string;
# Add indexes based on query patterns (delete_relationships_by_source_id, get_knowledge_relationships)
DEFINE INDEX IF NOT EXISTS relates_to_metadata_source_id_idx ON relates_to FIELDS metadata.source_id;
DEFINE INDEX IF NOT EXISTS relates_to_metadata_user_id_idx ON relates_to FIELDS metadata.user_id;
@@ -19,6 +19,6 @@ CREATE system_settings:current CONTENT {
image_processing_prompt: "Analyze this image and respond based on its primary content:\n - If the image is mainly text (document, screenshot, sign), transcribe the text verbatim.\n - If the image is mainly visual (photograph, art, landscape), provide a concise description of the scene.\n - For hybrid images (diagrams, ads), briefly describe the visual, then transcribe the text under a Text: heading.\n\n Respond directly with the analysis.",
embedding_dimensions: 1536,
query_system_prompt: "You are a knowledgeable assistant with access to a specialized knowledge base. You will be provided with relevant knowledge entities from the database as context. Each knowledge entity contains a name, description, and type, representing different concepts, ideas, and information.\nYour task is to:\n1. Carefully analyze the provided knowledge entities in the context\n2. Answer user questions based on this information\n3. Provide clear, concise, and accurate responses\n4. When referencing information, briefly mention which knowledge entity it came from\n5. If the provided context doesn't contain enough information to answer the question confidently, clearly state this\n6. If only partial information is available, explain what you can answer and what information is missing\n7. Avoid making assumptions or providing information not supported by the context\n8. Output the references to the documents. Use the UUIDs and make sure they are correct!\nRemember:\n- Be direct and honest about the limitations of your knowledge\n- Cite the relevant knowledge entities when providing information, but only provide the UUIDs in the reference array\n- If you need to combine information from multiple entities, explain how they connect\n- Don't speculate beyond what's provided in the context\nExample response formats:\n\"Based on [Entity Name], [answer...]\"\n\"I found relevant information in multiple entries: [explanation...]\"\n\"I apologize, but the provided context doesn't contain information about [topic]\"",
ingestion_system_prompt: "You are an AI assistant. You will receive a text content, along with user context and a category. Your task is to provide a structured JSON object representing the content in a graph format suitable for a graph database. You will also be presented with some existing knowledge_entities from the database, do not replicate these! Your task is to create meaningful knowledge entities from the submitted content. Try and infer as much as possible from the users context and category when creating these. If the user submits a large content, create more general entities. If the user submits a narrow and precise content, try and create precise knowledge entities.\nThe JSON should have the following structure:\n{\n\"knowledge_entities\": [\n{\n\"key\": \"unique-key-1\",\n\"name\": \"Entity Name\",\n\"description\": \"A detailed description of the entity.\",\n\"entity_type\": \"TypeOfEntity\"\n},\n// More entities...\n],\n\"relationships\": [\n{\n\"type\": \"RelationshipType\",\n\"source\": \"unique-key-1 or UUID from existing database\",\n\"target\": \"unique-key-1 or UUID from existing database\"\n},\n// More relationships...\n]\n}\nGuidelines:\n1. Do NOT generate any IDs or UUIDs. Use a unique `key` for each knowledge entity.\n2. Each KnowledgeEntity should have a unique `key`, a meaningful `name`, and a descriptive `description`.\n3. Define the type of each KnowledgeEntity using the following categories: Idea, Project, Document, Page, TextSnippet.\n4. Establish relationships between entities using types like RelatedTo, RelevantTo, SimilarTo.\n5. Use the `source` key to indicate the originating entity and the `target` key to indicate the related entity.\n6. You will be presented with a few existing KnowledgeEntities that are similar to the current ones. They will have an existing UUID. When creating relationships to these entities, use their UUID.\n7. Only create relationships between existing KnowledgeEntities.\n8. Entities that exist already in the database should NOT be created again. If there is only a minor overlap, skip creating a new entity.\n9. A new relationship MUST include a newly created KnowledgeEntity."
ingestion_system_prompt: "You are an AI assistant. You will receive a text content, along with user context and a category. Your task is to provide a structured JSON object representing the content in a graph format suitable for a graph database. You will also be presented with some existing knowledge_entities from the database, do not replicate these! Your task is to create meaningful knowledge entities from the submitted content. Try and infer as much as possible from the users context and category when creating these. If the user submits a large content, create more general entities. If the user submits a narrow and precise content, try and create precise knowledge entities.\nThe JSON should have the following structure:\n{\n\"knowledge_entities\": [\n{\n\"key\": \"unique-key-1\",\n\"name\": \"Entity Name\",\n\"description\": \"A detailed description of the entity.\",\n\"entity_type\": \"TypeOfEntity\"\n},\n// More entities...\n],\n\"relationships\": [\n{\n\"type\": \"RelationshipType\",\n\"source\": \"unique-key-1 or UUID from existing database\",\n\"target\": \"unique-key-1 or UUID from existing database\"\n},\n// More relationships...\n]\n}\nGuidelines:\n1. Do NOT generate any IDs or UUIDs. Use a unique `key` for each knowledge entity.\n2. Each KnowledgeEntity should have a unique `key`, a meaningful `name`, and a descriptive `description`.\n3. Define the type of each KnowledgeEntity using the following categories: Idea, Project, Document, Page, TextSnippet.\n4. Establish relationships between entities using types like RelatedTo, RelevantTo, SimilarTo.\n5. Use the `source` key to indicate the originating entity and the `target` key to indicate the related entity\"\n6. You will be presented with a few existing KnowledgeEntities that are similar to the current ones. They will have an existing UUID. When creating relationships to these entities, use their UUID.\n7. Only create relationships between existing KnowledgeEntities.\n8. Entities that exist already in the database should NOT be created again. If there is only a minor overlap, skip creating a new entity.\n9. A new relationship MUST include a newly created KnowledgeEntity."
};
END;
@@ -13,4 +13,3 @@ DEFINE FIELD IF NOT EXISTS title ON conversation TYPE string;
# Add indexes based on query patterns (get_complete_conversation ownership check, get_user_conversations)
DEFINE INDEX IF NOT EXISTS conversation_user_id_idx ON conversation FIELDS user_id;
DEFINE INDEX IF NOT EXISTS conversation_created_at_idx ON conversation FIELDS created_at; # For get_user_conversations ORDER BY
DEFINE INDEX IF NOT EXISTS conversation_user_updated_at_idx ON conversation FIELDS user_id, updated_at; # For sidebar conversation projection ORDER BY
@@ -13,6 +13,7 @@ DEFINE FIELD IF NOT EXISTS file_name ON file TYPE string;
DEFINE FIELD IF NOT EXISTS mime_type ON file TYPE string;
DEFINE FIELD IF NOT EXISTS user_id ON file TYPE string;
# Indexes based on usage (get_by_sha scoped by user_id, user lookups)
DEFINE INDEX IF NOT EXISTS file_user_sha256_idx ON file FIELDS user_id, sha256 UNIQUE;
# Indexes based on usage (get_by_sha, potentially user lookups)
# Using UNIQUE based on the logic in FileInfo::new to prevent duplicates
DEFINE INDEX IF NOT EXISTS file_sha256_idx ON file FIELDS sha256 UNIQUE;
DEFINE INDEX IF NOT EXISTS file_user_id_idx ON file FIELDS user_id;
@@ -1,6 +1,6 @@
# Defines the schema for the 'knowledge_entity' table.
DEFINE TABLE IF NOT EXISTS knowledge_entity SCHEMAFULL;
DEFINE TABLE IF NOT EXISTS knowledge_entity SCHEMALESS;
# Standard fields
DEFINE FIELD IF NOT EXISTS created_at ON knowledge_entity TYPE datetime;
@@ -22,6 +22,5 @@ DEFINE FIELD IF NOT EXISTS user_id ON knowledge_entity TYPE string;
-- DEFINE INDEX IF NOT EXISTS idx_embedding_entities ON knowledge_entity FIELDS embedding HNSW DIMENSION 1536;
DEFINE INDEX IF NOT EXISTS knowledge_entity_source_id_idx ON knowledge_entity FIELDS source_id;
DEFINE INDEX IF NOT EXISTS knowledge_entity_user_id_idx ON knowledge_entity FIELDS user_id;
DEFINE INDEX IF NOT EXISTS knowledge_entity_user_source_idx ON knowledge_entity FIELDS user_id, source_id;
DEFINE INDEX IF NOT EXISTS knowledge_entity_entity_type_idx ON knowledge_entity FIELDS entity_type;
DEFINE INDEX IF NOT EXISTS knowledge_entity_created_at_idx ON knowledge_entity FIELDS created_at;
@@ -7,7 +7,6 @@ DEFINE TABLE IF NOT EXISTS knowledge_entity_embedding SCHEMAFULL;
DEFINE FIELD IF NOT EXISTS created_at ON knowledge_entity_embedding TYPE datetime;
DEFINE FIELD IF NOT EXISTS updated_at ON knowledge_entity_embedding TYPE datetime;
DEFINE FIELD IF NOT EXISTS user_id ON knowledge_entity_embedding TYPE string;
DEFINE FIELD IF NOT EXISTS source_id ON knowledge_entity_embedding TYPE string;
-- Custom fields
DEFINE FIELD IF NOT EXISTS entity_id ON knowledge_entity_embedding TYPE record<knowledge_entity>;
@@ -15,6 +14,5 @@ DEFINE FIELD IF NOT EXISTS embedding ON knowledge_entity_embedding TYPE array<fl
-- Indexes
-- DEFINE INDEX IF NOT EXISTS idx_embedding_knowledge_entity_embedding ON knowledge_entity_embedding FIELDS embedding HNSW DIMENSION 1536;
DEFINE INDEX IF NOT EXISTS knowledge_entity_embedding_entity_id_idx ON knowledge_entity_embedding FIELDS entity_id UNIQUE;
DEFINE INDEX IF NOT EXISTS knowledge_entity_embedding_entity_id_idx ON knowledge_entity_embedding FIELDS entity_id;
DEFINE INDEX IF NOT EXISTS knowledge_entity_embedding_user_id_idx ON knowledge_entity_embedding FIELDS user_id;
DEFINE INDEX IF NOT EXISTS knowledge_entity_embedding_source_id_idx ON knowledge_entity_embedding FIELDS source_id;
+19
View File
@@ -0,0 +1,19 @@
# Defines the 'relates_to' edge table for KnowledgeRelationships.
# Edges connect nodes, in this case knowledge_entity records.
# Define the edge table itself, enforcing connections between knowledge_entity records
# SCHEMAFULL requires all fields to be defined, maybe start with SCHEMALESS if metadata might vary
DEFINE TABLE IF NOT EXISTS relates_to SCHEMALESS TYPE RELATION FROM knowledge_entity TO knowledge_entity;
# Define the metadata field within the edge
# RelationshipMetadata is a struct, store as object
DEFINE FIELD IF NOT EXISTS metadata ON relates_to TYPE object;
# Optionally, define fields within the metadata object for stricter schema (requires SCHEMAFULL on table)
# DEFINE FIELD IF NOT EXISTS metadata.user_id ON relates_to TYPE string;
# DEFINE FIELD IF NOT EXISTS metadata.source_id ON relates_to TYPE string;
# DEFINE FIELD IF NOT EXISTS metadata.relationship_type ON relates_to TYPE string;
# Add indexes based on query patterns (delete_relationships_by_source_id, get_knowledge_relationships)
DEFINE INDEX IF NOT EXISTS relates_to_metadata_source_id_idx ON relates_to FIELDS metadata.source_id;
DEFINE INDEX IF NOT EXISTS relates_to_metadata_user_id_idx ON relates_to FIELDS metadata.user_id;
@@ -15,6 +15,6 @@ DEFINE FIELD IF NOT EXISTS embedding ON text_chunk_embedding TYPE array<float>;
-- Indexes
-- DEFINE INDEX IF NOT EXISTS idx_embedding_text_chunk_embedding ON text_chunk_embedding FIELDS embedding HNSW DIMENSION 1536;
DEFINE INDEX IF NOT EXISTS text_chunk_embedding_chunk_id_idx ON text_chunk_embedding FIELDS chunk_id UNIQUE;
DEFINE INDEX IF NOT EXISTS text_chunk_embedding_chunk_id_idx ON text_chunk_embedding FIELDS chunk_id;
DEFINE INDEX IF NOT EXISTS text_chunk_embedding_user_id_idx ON text_chunk_embedding FIELDS user_id;
DEFINE INDEX IF NOT EXISTS text_chunk_embedding_source_id_idx ON text_chunk_embedding FIELDS source_id;
+20 -110
View File
@@ -4,128 +4,38 @@ use tokio::task::JoinError;
use crate::storage::types::file_info::FileError;
/// Errors from embedding provider operations.
#[allow(clippy::module_name_repetitions)]
#[derive(Error, Debug)]
pub enum EmbeddingError {
#[error("openai error: {0}")]
OpenAI(Box<OpenAIError>),
#[error("fastembed error: {0}")]
FastEmbed(String),
#[error("task join error: {0}")]
Join(#[from] JoinError),
#[error("fastembed model mutex poisoned: {0}")]
MutexPoisoned(String),
#[error("no embedding data received")]
NoData,
#[error("embedding configuration error: {0}")]
Config(String),
#[error("unknown fastembed model: {0}")]
UnknownModel(String),
}
impl From<OpenAIError> for EmbeddingError {
fn from(err: OpenAIError) -> Self {
Self::OpenAI(Box::new(err))
}
}
impl EmbeddingError {
pub(crate) fn fastembed(err: impl std::fmt::Display) -> Self {
Self::FastEmbed(err.to_string())
}
pub(crate) fn mutex_poisoned(err: impl std::fmt::Display) -> Self {
Self::MutexPoisoned(err.to_string())
}
}
// Core internal errors
#[allow(clippy::module_name_repetitions)]
#[derive(Error, Debug)]
pub enum AppError {
#[error("database error: {0}")]
Database(Box<surrealdb::Error>),
#[error("openai error: {0}")]
OpenAI(Box<OpenAIError>),
#[error("embedding error: {0}")]
Embedding(#[from] EmbeddingError),
#[error("file error: {0}")]
#[error("Database error: {0}")]
Database(#[from] surrealdb::Error),
#[error("OpenAI error: {0}")]
OpenAI(#[from] OpenAIError),
#[error("File error: {0}")]
File(#[from] FileError),
#[error("not found: {0}")]
#[error("Not found: {0}")]
NotFound(String),
#[error("validation error: {0}")]
#[error("Validation error: {0}")]
Validation(String),
#[error("authorization error: {0}")]
#[error("Authorization error: {0}")]
Auth(String),
#[error("llm parsing error: {0}")]
#[error("LLM parsing error: {0}")]
LLMParsing(String),
#[error("task join error: {0}")]
#[error("Task join error: {0}")]
Join(#[from] JoinError),
#[error("graph mapper error: {0}")]
#[error("Graph mapper error: {0}")]
GraphMapper(String),
#[error("io error: {0}")]
#[error("IoError: {0}")]
Io(#[from] std::io::Error),
#[error("reqwest error: {0}")]
Reqwest(Box<reqwest::Error>),
#[error("storage error: {0}")]
Storage(Box<object_store::Error>),
#[error("ingestion processing error: {0}")]
#[error("Reqwest error: {0}")]
Reqwest(#[from] reqwest::Error),
#[error("Anyhow error: {0}")]
Anyhow(#[from] anyhow::Error),
#[error("Ingestion Processing error: {0}")]
Processing(String),
#[error("dom smoothie error: {0}")]
DomSmoothie(Box<dom_smoothie::ReadabilityError>),
#[error("internal service error: {0}")]
#[error("DOM smoothie error: {0}")]
DomSmoothie(#[from] dom_smoothie::ReadabilityError),
#[error("Internal service error: {0}")]
InternalError(String),
}
impl From<surrealdb::Error> for AppError {
fn from(err: surrealdb::Error) -> Self {
Self::Database(Box::new(err))
}
}
impl From<OpenAIError> for AppError {
fn from(err: OpenAIError) -> Self {
Self::OpenAI(Box::new(err))
}
}
impl From<reqwest::Error> for AppError {
fn from(err: reqwest::Error) -> Self {
Self::Reqwest(Box::new(err))
}
}
impl From<object_store::Error> for AppError {
fn from(err: object_store::Error) -> Self {
Self::Storage(Box::new(err))
}
}
impl From<dom_smoothie::ReadabilityError> for AppError {
fn from(err: dom_smoothie::ReadabilityError) -> Self {
Self::DomSmoothie(Box::new(err))
}
}
impl AppError {
/// Builds an [`AppError::InternalError`] from a displayable message.
#[must_use]
pub fn internal(msg: impl std::fmt::Display) -> Self {
Self::InternalError(msg.to_string())
}
}
#[cfg(test)]
mod tests {
use super::AppError;
#[test]
fn app_error_is_reasonably_sized() {
assert!(
std::mem::size_of::<AppError>() <= 64,
"AppError is {} bytes",
std::mem::size_of::<AppError>()
);
}
}
-3
View File
@@ -3,6 +3,3 @@
pub mod error;
pub mod storage;
pub mod utils;
#[cfg(any(test, feature = "test-utils"))]
pub mod test_utils;
+58 -114
View File
@@ -13,8 +13,8 @@ use surrealdb::{
use surrealdb_migrations::MigrationRunner;
use tracing::debug;
/// Embedded SurrealDB project root (`migrations/`, `schemas/`, `.surrealdb`).
static MIGRATIONS_DIR: Dir<'_> = include_dir!("$CARGO_MANIFEST_DIR/db");
/// Embedded SurrealDB migration directory packaged with the crate.
static MIGRATIONS_DIR: Dir<'_> = include_dir!("$CARGO_MANIFEST_DIR/");
#[derive(Clone)]
pub struct SurrealDbClient {
@@ -26,20 +26,12 @@ pub trait ProvidesDb {
}
impl SurrealDbClient {
/// Initialize a new database client.
/// # Initialize a new datbase client
///
/// # Arguments
///
/// * `address` — Database connection string (e.g. `ws://localhost:8000` or `mem://`).
/// * `username` — Root username for authentication.
/// * `password` — Root password for authentication.
/// * `namespace` — SurrealDB namespace to use.
/// * `database` — SurrealDB database to use.
///
/// # Errors
///
/// Returns `Err` if the connection, authentication, or namespace/database selection fails.
/// In-memory (`mem://`) connections skip authentication.
/// # Returns
/// * `SurrealDbClient` initialized
pub async fn new(
address: &str,
username: &str,
@@ -49,10 +41,8 @@ impl SurrealDbClient {
) -> Result<Self, Error> {
let db = connect(address).await?;
// Skip sign-in for in-memory engine (no auth support)
if !address.starts_with("mem://") {
// Sign in to database
db.signin(Root { username, password }).await?;
}
// Set namespace
db.use_ns(namespace).use_db(database).await?;
@@ -60,19 +50,6 @@ impl SurrealDbClient {
Ok(SurrealDbClient { client: db })
}
/// Initialize a new database client using namespace-level authentication.
///
/// # Arguments
///
/// * `address` — Database connection string.
/// * `namespace` — SurrealDB namespace to use (also used for auth).
/// * `username` — Namespace username for authentication.
/// * `password` — Namespace password for authentication.
/// * `database` — SurrealDB database to use.
///
/// # Errors
///
/// Returns `Err` if the connection, namespace authentication, or namespace/database selection fails.
pub async fn new_with_namespace_user(
address: &str,
namespace: &str,
@@ -91,11 +68,6 @@ impl SurrealDbClient {
Ok(SurrealDbClient { client: db })
}
/// Create an Axum session store backed by SurrealDB.
///
/// # Errors
///
/// Returns `SessionError` if the session store configuration or table creation fails.
pub async fn create_session_store(
&self,
) -> Result<SessionStore<SessionSurrealPool<Any>>, SessionError> {
@@ -103,7 +75,7 @@ impl SurrealDbClient {
SessionStore::new(
Some(self.client.clone().into()),
SessionConfig::default()
.with_table_name("session")
.with_table_name("test_session_table")
.with_secure(true),
)
.await
@@ -114,63 +86,51 @@ impl SurrealDbClient {
/// This function should be called during application startup, after connecting to
/// the database and selecting the appropriate namespace and database, but before
/// the application starts performing operations that rely on the schema.
///
/// # Errors
///
/// Returns `AppError::InternalError` if the migration runner fails to apply any migration.
pub async fn apply_migrations(&self) -> Result<(), AppError> {
debug!("Applying migrations");
MigrationRunner::new(&self.client)
.load_files(&MIGRATIONS_DIR)
.up()
.await
.map_err(AppError::internal)?;
.map_err(|e| AppError::InternalError(e.to_string()))?;
Ok(())
}
/// Store an object in SurrealDB.
/// Operation to store a object in SurrealDB, requires the struct to implement StoredObject
///
/// # Arguments
/// * `item` - The item to be stored
///
/// * `item` — The item to store. Must implement `StoredObject`.
///
/// # Errors
///
/// Returns `Err` if the database create operation fails.
/// # Returns
/// * `Result` - Item or Error
pub async fn store_item<T>(&self, item: T) -> Result<Option<T>, Error>
where
T: StoredObject + Send + Sync + 'static,
{
self.client
.create((T::table_name(), item.id()))
.create((T::table_name(), item.get_id()))
.content(item)
.await
}
/// Upsert an object in SurrealDB, replacing any existing record with the same ID.
///
/// Useful for idempotent ingestion flows.
///
/// # Errors
///
/// Returns `Err` if the database upsert operation fails.
/// Operation to upsert an object in SurrealDB, replacing any existing record
/// with the same ID. Useful for idempotent ingestion flows.
pub async fn upsert_item<T>(&self, item: T) -> Result<Option<T>, Error>
where
T: StoredObject + Send + Sync + 'static,
{
let id = item.id().to_string();
let id = item.get_id().to_string();
self.client
.upsert((T::table_name(), id))
.content(item)
.await
}
/// Retrieve all objects from a table.
/// Operation to retrieve all objects from a certain table, requires the struct to implement StoredObject
///
/// # Errors
///
/// Returns `Err` if the database select operation fails.
/// # Returns
/// * `Result` - Vec<T> or Error
pub async fn get_all_stored_items<T>(&self) -> Result<Vec<T>, Error>
where
T: for<'de> StoredObject,
@@ -178,16 +138,13 @@ impl SurrealDbClient {
self.client.select(T::table_name()).await
}
/// Retrieve a single object by its ID.
/// Operation to retrieve a single object by its ID, requires the struct to implement StoredObject
///
/// # Arguments
/// * `id` - The ID of the item to retrieve
///
/// * `id` — The ID of the item to retrieve.
///
/// # Errors
///
/// Returns `Err` if the database select operation fails.
/// Returns `Ok(None)` if no record with the given ID exists.
/// # Returns
/// * `Result<Option<T>, Error>` - The found item or Error
pub async fn get_item<T>(&self, id: &str) -> Result<Option<T>, Error>
where
T: for<'de> StoredObject,
@@ -195,16 +152,13 @@ impl SurrealDbClient {
self.client.select((T::table_name(), id)).await
}
/// Delete a single object by its ID.
/// Operation to delete a single object by its ID, requires the struct to implement StoredObject
///
/// # Arguments
/// * `id` - The ID of the item to delete
///
/// * `id` — The ID of the item to delete.
///
/// # Errors
///
/// Returns `Err` if the database delete operation fails.
/// Returns `Ok(None)` if no record with the given ID exists.
/// # Returns
/// * `Result<Option<T>, Error>` - The deleted item or Error
pub async fn delete_item<T>(&self, id: &str) -> Result<Option<T>, Error>
where
T: for<'de> StoredObject,
@@ -212,11 +166,10 @@ impl SurrealDbClient {
self.client.delete((T::table_name(), id)).await
}
/// Listen to a table for real-time updates via a live query stream.
/// Operation to listen to a table for updates, requires the struct to implement StoredObject
///
/// # Errors
///
/// Returns `Err` if the database live query subscription fails.
/// # Returns
/// * `Result<Option<T>, Error>` - The deleted item or Error
pub async fn listen<T>(
&self,
) -> Result<impl Stream<Item = Result<Notification<T>, Error>>, Error>
@@ -249,9 +202,7 @@ impl SurrealDbClient {
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use crate::stored_object;
use anyhow::{self, Context};
use super::*;
use uuid::Uuid;
@@ -261,17 +212,19 @@ mod tests {
});
#[tokio::test]
async fn test_initialization_and_crud() -> anyhow::Result<()> {
async fn test_initialization_and_crud() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let database = &Uuid::new_v4().to_string(); // ensures isolation per test run
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
// Call your initialization
db.apply_migrations()
.await
.with_context(|| "Failed to initialize schema".to_string())?;
.expect("Failed to initialize schema");
// Test basic CRUD
let dummy = Dummy {
id: "abc".to_string(),
name: "first".to_string(),
@@ -279,50 +232,50 @@ mod tests {
updated_at: Utc::now(),
};
let stored = db
.store_item(dummy.clone())
.await
.with_context(|| "Failed to store".to_string())?;
// Store
let stored = db.store_item(dummy.clone()).await.expect("Failed to store");
assert!(stored.is_some());
// Read
let fetched = db
.get_item::<Dummy>(&dummy.id)
.await
.with_context(|| "Failed to fetch".to_string())?;
.expect("Failed to fetch");
assert_eq!(fetched, Some(dummy.clone()));
// Read all
let all = db
.get_all_stored_items::<Dummy>()
.await
.with_context(|| "Failed to fetch all".to_string())?;
.expect("Failed to fetch all");
assert!(all.contains(&dummy));
// Delete
let deleted = db
.delete_item::<Dummy>(&dummy.id)
.await
.with_context(|| "Failed to delete".to_string())?;
.expect("Failed to delete");
assert_eq!(deleted, Some(dummy));
// After delete, should not be present
let fetch_post = db
.get_item::<Dummy>("abc")
.await
.with_context(|| "Failed fetch post delete".to_string())?;
.expect("Failed fetch post delete");
assert!(fetch_post.is_none());
Ok(())
}
#[tokio::test]
async fn upsert_item_overwrites_existing_records() -> anyhow::Result<()> {
async fn upsert_item_overwrites_existing_records() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
db.apply_migrations()
.await
.with_context(|| "Failed to initialize schema".to_string())?;
.expect("Failed to initialize schema");
let mut dummy = Dummy {
id: "abc".to_string(),
@@ -333,22 +286,17 @@ mod tests {
db.store_item(dummy.clone())
.await
.with_context(|| "Failed to store initial record".to_string())?;
.expect("Failed to store initial record");
dummy.name = "updated".to_string();
let upserted = db
.upsert_item(dummy.clone())
.await
.with_context(|| "Failed to upsert record".to_string())?;
.expect("Failed to upsert record");
assert!(upserted.is_some());
let fetched: Option<Dummy> = db
.get_item(&dummy.id)
.await
.with_context(|| "fetch after upsert".to_string())?;
let fetched =
fetched.ok_or_else(|| anyhow::anyhow!("Expected record to exist after upsert"))?;
assert_eq!(fetched.name, "updated");
let fetched: Option<Dummy> = db.get_item(&dummy.id).await.expect("fetch after upsert");
assert_eq!(fetched.unwrap().name, "updated");
let new_record = Dummy {
id: "def".to_string(),
@@ -358,29 +306,25 @@ mod tests {
};
db.upsert_item(new_record.clone())
.await
.with_context(|| "Failed to upsert new record".to_string())?;
.expect("Failed to upsert new record");
let fetched_new: Option<Dummy> = db
.get_item(&new_record.id)
.await
.with_context(|| "fetch inserted via upsert".to_string())?;
.expect("fetch inserted via upsert");
assert_eq!(fetched_new, Some(new_record));
Ok(())
}
#[tokio::test]
async fn test_applying_migrations() -> anyhow::Result<()> {
async fn test_applying_migrations() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
db.apply_migrations()
.await
.with_context(|| "Failed to build indexes".to_string())?;
Ok(())
.expect("Failed to build indexes");
}
}
+123 -306
View File
@@ -9,38 +9,8 @@ use tracing::{debug, info, warn};
use crate::{error::AppError, storage::db::SurrealDbClient};
const INDEX_POLL_INTERVAL: Duration = Duration::from_millis(50);
const INDEX_BUILD_TIMEOUT: Duration = Duration::from_secs(30 * 60);
const FTS_ANALYZER_NAME: &str = "app_en_fts_analyzer";
/// HNSW index options used by runtime index creation (includes CONCURRENTLY).
pub const HNSW_INDEX_OPTIONS: &str = "DIST COSINE TYPE F32 EFC 100 M 8 CONCURRENTLY";
/// HNSW index options for use inside transactions (CONCURRENTLY not supported).
pub const HNSW_INDEX_OPTIONS_SYNC: &str = "DIST COSINE TYPE F32 EFC 100 M 8";
/// Builds a `DEFINE INDEX OVERWRITE ... HNSW` statement matching runtime index options.
#[must_use]
pub fn hnsw_index_overwrite_sql(index_name: &str, table: &str, dimension: usize) -> String {
format!(
"DEFINE INDEX OVERWRITE {index_name} ON TABLE {table} \
FIELDS embedding HNSW DIMENSION {dimension} {HNSW_INDEX_OPTIONS};"
)
}
/// Recreates an HNSW index inside a transaction (for tests and dimension migrations).
#[must_use]
pub fn hnsw_index_redefine_transaction_sql(
index_name: &str,
table: &str,
dimension: usize,
) -> String {
format!(
"BEGIN TRANSACTION;
REMOVE INDEX IF EXISTS {index_name} ON TABLE {table};
DEFINE INDEX {index_name} ON TABLE {table} FIELDS embedding HNSW DIMENSION {dimension} {HNSW_INDEX_OPTIONS_SYNC};
COMMIT TRANSACTION;"
)
}
#[derive(Clone, Copy)]
struct HnswIndexSpec {
index_name: &'static str,
@@ -53,12 +23,12 @@ const fn hnsw_index_specs() -> [HnswIndexSpec; 2] {
HnswIndexSpec {
index_name: "idx_embedding_text_chunk_embedding",
table: "text_chunk_embedding",
options: HNSW_INDEX_OPTIONS,
options: "DIST COSINE TYPE F32 EFC 100 M 8 CONCURRENTLY",
},
HnswIndexSpec {
index_name: "idx_embedding_knowledge_entity_embedding",
table: "knowledge_entity_embedding",
options: HNSW_INDEX_OPTIONS,
options: "DIST COSINE TYPE F32 EFC 100 M 8 CONCURRENTLY",
},
]
}
@@ -189,50 +159,26 @@ impl FtsIndexSpec {
/// Build runtime Surreal indexes (FTS + HNSW) using concurrent creation with readiness polling.
/// Idempotent: safe to call multiple times and will overwrite HNSW definitions when the dimension changes.
///
/// # Errors
///
/// Returns `AppError::InternalError` if any index definition or polling step fails.
pub async fn ensure_runtime(
pub async fn ensure_runtime_indexes(
db: &SurrealDbClient,
embedding_dimension: usize,
) -> Result<(), AppError> {
ensure_runtime_inner(db, embedding_dimension)
ensure_runtime_indexes_inner(db, embedding_dimension)
.await
.map_err(AppError::internal)
.map_err(|err| AppError::InternalError(err.to_string()))
}
/// Rebuild known FTS and HNSW indexes, skipping any that are not yet defined.
///
/// # Errors
///
/// Returns `AppError::InternalError` if any index rebuild operation fails.
pub async fn rebuild(db: &SurrealDbClient) -> Result<(), AppError> {
rebuild_inner(db).await.map_err(AppError::internal)
}
/// Returns the dimension of the currently defined chunk-embedding HNSW index, if any.
///
/// Stored embeddings always share this index's dimension because re-embedding rewrites the
/// vectors and the index together, so it acts as a persisted marker of the embedding space
/// actually present in the database. Returns `Ok(None)` when the index has not been created yet
/// (for example on a fresh database with no ingested data).
///
/// # Errors
///
/// Returns `AppError::InternalError` if the index metadata cannot be read.
pub async fn embedding_index_dimension(db: &SurrealDbClient) -> Result<Option<usize>, AppError> {
let spec = HnswIndexSpec {
index_name: "idx_embedding_text_chunk_embedding",
table: "text_chunk_embedding",
options: HNSW_INDEX_OPTIONS,
};
existing_hnsw_dimension(db, &spec)
pub async fn rebuild_indexes(db: &SurrealDbClient) -> Result<(), AppError> {
rebuild_indexes_inner(db)
.await
.map_err(AppError::internal)
.map_err(|err| AppError::InternalError(err.to_string()))
}
async fn ensure_runtime_inner(db: &SurrealDbClient, embedding_dimension: usize) -> Result<()> {
async fn ensure_runtime_indexes_inner(
db: &SurrealDbClient,
embedding_dimension: usize,
) -> Result<()> {
create_fts_analyzer(db).await?;
for spec in fts_index_specs() {
@@ -316,17 +262,22 @@ async fn get_index_status(db: &SurrealDbClient, index_name: &str, table: &str) -
.context("checking index status")?;
let info: Option<Value> = info_res.take(0).context("failed to take info result")?;
let Some(info) = info else {
return Ok("unknown".to_string());
let info = match info {
Some(i) => i,
None => return Ok("unknown".to_string()),
};
let parsed: IndexInfoForIndex =
serde_json::from_value(info).context("deserializing INFO FOR INDEX response")?;
let building = info.get("building");
let status = building
.and_then(|b| b.get("status"))
.and_then(|s| s.as_str())
.unwrap_or("ready")
.to_string();
Ok(parsed.building_status())
Ok(status)
}
async fn rebuild_inner(db: &SurrealDbClient) -> Result<()> {
async fn rebuild_indexes_inner(db: &SurrealDbClient) -> Result<()> {
debug!("Rebuilding indexes with concurrent definitions");
create_fts_analyzer(db).await?;
@@ -434,9 +385,10 @@ async fn create_fts_analyzer(db: &SurrealDbClient) -> Result<()> {
// is unavailable in the running Surreal build. Use IF NOT EXISTS to avoid clobbering
// an existing analyzer definition.
let snowball_query = format!(
"DEFINE ANALYZER IF NOT EXISTS {FTS_ANALYZER_NAME}
"DEFINE ANALYZER IF NOT EXISTS {analyzer}
TOKENIZERS class
FILTERS lowercase, ascii, snowball(english);"
FILTERS lowercase, ascii, snowball(english);",
analyzer = FTS_ANALYZER_NAME
);
match db.client.query(snowball_query).await {
@@ -458,9 +410,10 @@ async fn create_fts_analyzer(db: &SurrealDbClient) -> Result<()> {
}
let fallback_query = format!(
"DEFINE ANALYZER IF NOT EXISTS {FTS_ANALYZER_NAME}
"DEFINE ANALYZER IF NOT EXISTS {analyzer}
TOKENIZERS class
FILTERS lowercase, ascii;"
FILTERS lowercase, ascii;",
analyzer = FTS_ANALYZER_NAME
);
let res = db
@@ -493,7 +446,6 @@ async fn create_index_with_polling(
table: &str,
progress_table: Option<&str>,
) -> Result<()> {
const MAX_ATTEMPTS: usize = 3;
let expected_total = match progress_table {
Some(table) => Some(count_table_rows(db, table).await.with_context(|| {
format!("counting rows in {table} for index {index_name} progress")
@@ -501,9 +453,10 @@ async fn create_index_with_polling(
None => None,
};
let mut attempts: usize = 0;
let mut attempts = 0;
const MAX_ATTEMPTS: usize = 3;
loop {
attempts = attempts.saturating_add(1);
attempts += 1;
let res = db
.client
.query(definition.clone())
@@ -551,20 +504,8 @@ async fn poll_index_build_status(
poll_every: Duration,
) -> Result<()> {
let started_at = std::time::Instant::now();
let mut last_snapshot: Option<IndexBuildSnapshot> = None;
loop {
if started_at.elapsed() >= INDEX_BUILD_TIMEOUT {
return Err(anyhow::anyhow!(
"index build timed out after {:?} for {index_name} on {table} (last status: {})",
INDEX_BUILD_TIMEOUT,
last_snapshot
.as_ref()
.map_or("unknown", |snapshot| snapshot.status.as_str())
))
.with_context(|| format!("index {index_name} on table {table} did not become ready"));
}
tokio::time::sleep(poll_every).await;
let info_query = format!("INFO FOR INDEX {index_name} ON TABLE {table};");
@@ -578,15 +519,16 @@ async fn poll_index_build_status(
.context("failed to deserialize INFO FOR INDEX result")?;
let Some(snapshot) = parse_index_build_info(info, total_rows) else {
return Err(anyhow::anyhow!(
"INFO FOR INDEX returned no data for {index_name} on {table}"
));
warn!(
index = %index_name,
table = %table,
"INFO FOR INDEX returned no data; assuming index definition might be missing"
);
break;
};
last_snapshot = Some(snapshot.clone());
if let Some(pct) = snapshot.progress_pct {
debug!(
match snapshot.progress_pct {
Some(pct) => debug!(
index = %index_name,
table = %table,
status = snapshot.status,
@@ -597,9 +539,8 @@ async fn poll_index_build_status(
total = snapshot.total_rows,
progress_pct = format_args!("{pct:.1}"),
"Index build status"
);
} else {
debug!(
),
None => debug!(
index = %index_name,
table = %table,
status = snapshot.status,
@@ -608,7 +549,7 @@ async fn poll_index_build_status(
updated = snapshot.updated,
processed = snapshot.processed,
"Index build status"
);
),
}
if snapshot.is_ready() {
@@ -620,101 +561,31 @@ async fn poll_index_build_status(
total = snapshot.total_rows,
"Index is ready"
);
return Ok(());
break;
}
if snapshot.status.eq_ignore_ascii_case("error") {
return Err(anyhow::anyhow!(
"index build failed for {index_name} on {table}: status=error, processed={}, total={:?}",
snapshot.processed,
snapshot.total_rows
));
}
}
}
/// `building` block from SurrealDB `INFO FOR INDEX` (concurrent index builds).
#[derive(Debug, Clone, Deserialize, PartialEq, Eq)]
struct IndexBuildingProgress {
#[serde(default)]
initial: u64,
#[serde(default)]
pending: u64,
#[serde(default)]
updated: u64,
#[serde(default)]
status: String,
}
/// Top-level `INFO FOR INDEX` payload shape (SurrealDB v2.x).
#[derive(Debug, Clone, Deserialize, PartialEq, Eq, Default)]
struct IndexInfoForIndex {
#[serde(default)]
building: Option<IndexBuildingProgress>,
}
impl IndexInfoForIndex {
fn building_status(&self) -> String {
match &self.building {
None => "ready".to_string(),
Some(progress) if progress.status.is_empty() => "ready".to_string(),
Some(progress) => progress.status.clone(),
warn!(
index = %index_name,
table = %table,
status = snapshot.status,
"Index build reported error status; stopping polling"
);
break;
}
}
fn into_build_snapshot(self, total_rows: Option<u64>) -> IndexBuildSnapshot {
let (initial, pending, updated, status) = match self.building {
None => (0, 0, 0, "ready".to_string()),
Some(progress) => {
let status = if progress.status.is_empty() {
"ready".to_string()
} else {
progress.status
};
(progress.initial, progress.pending, progress.updated, status)
}
};
let processed = initial.saturating_add(updated);
let progress_pct = total_rows.map(|total| {
if total == 0 {
0.0
} else {
((f64::from(u32::try_from(processed).unwrap_or(u32::MAX))
/ f64::from(u32::try_from(total).unwrap_or(1)))
.min(1.0))
* 100.0
}
});
IndexBuildSnapshot {
status,
initial,
pending,
updated,
processed,
total_rows,
progress_pct,
}
}
Ok(())
}
/// Snapshot of an index build progress as reported by SurrealDB's `INFO FOR INDEX`.
#[derive(Debug, Clone, PartialEq)]
#[derive(Debug, PartialEq)]
struct IndexBuildSnapshot {
/// Current build status string (e.g., `"indexing"`, `"ready"`, `"error"`).
status: String,
/// Number of rows present when the build started.
initial: u64,
/// Number of rows still pending processing.
pending: u64,
/// Number of rows updated since the build started.
updated: u64,
/// Total rows processed so far (`initial + updated`).
processed: u64,
/// Total rows expected (from `SELECT count()` before the build), if available.
total_rows: Option<u64>,
/// Progress as a percentage of `processed / total_rows`, if `total_rows` is known.
progress_pct: Option<f64>,
}
@@ -729,8 +600,50 @@ fn parse_index_build_info(
total_rows: Option<u64>,
) -> Option<IndexBuildSnapshot> {
let info = info?;
let parsed: IndexInfoForIndex = serde_json::from_value(info).ok()?;
Some(parsed.into_build_snapshot(total_rows))
let building = info.get("building");
let status = building
.and_then(|b| b.get("status"))
.and_then(|s| s.as_str())
// If there's no `building` block at all, treat as "ready" (index not building anymore)
.unwrap_or("ready")
.to_string();
let initial = building
.and_then(|b| b.get("initial"))
.and_then(|v| v.as_u64())
.unwrap_or(0);
let pending = building
.and_then(|b| b.get("pending"))
.and_then(|v| v.as_u64())
.unwrap_or(0);
let updated = building
.and_then(|b| b.get("updated"))
.and_then(|v| v.as_u64())
.unwrap_or(0);
// `initial` is the number of rows seen when the build started; `updated` accounts for later writes.
let processed = initial.saturating_add(updated);
let progress_pct = total_rows.map(|total| {
if total == 0 {
0.0
} else {
((processed as f64 / total as f64).min(1.0)) * 100.0
}
});
Some(IndexBuildSnapshot {
status,
initial,
pending,
updated,
processed,
total_rows,
progress_pct,
})
}
#[derive(Debug, Deserialize)]
@@ -760,7 +673,7 @@ async fn table_index_definitions(
.client
.query(info_query)
.await
.with_context(|| format!("fetching table info for {table}"))?;
.with_context(|| format!("fetching table info for {}", table))?;
let info: surrealdb::Value = response
.take(0)
@@ -787,16 +700,12 @@ async fn index_exists(db: &SurrealDbClient, table: &str, index_name: &str) -> Re
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use crate::storage::db::SurrealDbClient;
use anyhow::{self, Context};
use super::*;
use serde_json::json;
use uuid::Uuid;
use super::*;
#[test]
fn parse_index_build_info_reports_progress() -> anyhow::Result<()> {
fn parse_index_build_info_reports_progress() {
let info = json!({
"building": {
"initial": 56894,
@@ -806,7 +715,7 @@ mod tests {
}
});
let snapshot = parse_index_build_info(Some(info), Some(61081)).context("snapshot")?;
let snapshot = parse_index_build_info(Some(info), Some(61081)).expect("snapshot");
assert_eq!(
snapshot,
IndexBuildSnapshot {
@@ -820,84 +729,16 @@ mod tests {
}
);
assert!(!snapshot.is_ready());
Ok(())
}
#[test]
fn parse_index_build_info_defaults_to_ready_when_no_building_block() -> anyhow::Result<()> {
fn parse_index_build_info_defaults_to_ready_when_no_building_block() {
// Surreal returns `{}` when the index exists but isn't building.
let info = json!({});
let snapshot = parse_index_build_info(Some(info), Some(10)).context("snapshot")?;
let snapshot = parse_index_build_info(Some(info), Some(10)).expect("snapshot");
assert!(snapshot.is_ready());
assert_eq!(snapshot.processed, 0);
assert_eq!(snapshot.progress_pct, Some(0.0));
Ok(())
}
#[test]
fn index_info_for_index_deserializes_ready_status_shape() -> anyhow::Result<()> {
let info = json!({
"building": {
"status": "ready"
}
});
let parsed: IndexInfoForIndex =
serde_json::from_value(info).context("deserialize ready shape")?;
assert_eq!(parsed.building_status(), "ready");
let snapshot = parse_index_build_info(
Some(json!({
"building": { "status": "ready" }
})),
None,
)
.context("snapshot")?;
assert!(snapshot.is_ready());
assert_eq!(snapshot.initial, 0);
Ok(())
}
#[test]
fn index_info_for_index_deserializes_indexing_shape_from_surreal_docs() -> anyhow::Result<()> {
let info = json!({
"building": {
"initial": 8143,
"pending": 19,
"status": "indexing",
"updated": 80
}
});
let parsed: IndexInfoForIndex =
serde_json::from_value(info.clone()).context("deserialize indexing shape")?;
assert_eq!(parsed.building_status(), "indexing");
let snapshot = parse_index_build_info(Some(info), None).context("snapshot")?;
assert_eq!(snapshot.status, "indexing");
assert_eq!(snapshot.initial, 8143);
assert_eq!(snapshot.pending, 19);
assert_eq!(snapshot.updated, 80);
assert_eq!(snapshot.processed, 8223);
assert!(!snapshot.is_ready());
Ok(())
}
#[test]
fn parse_index_build_info_reports_error_status() -> anyhow::Result<()> {
let info = json!({
"building": {
"initial": 100,
"pending": 5,
"status": "error",
"updated": 10
}
});
let snapshot = parse_index_build_info(Some(info), Some(200)).context("snapshot")?;
assert_eq!(snapshot.status, "error");
assert!(!snapshot.is_ready());
Ok(())
}
#[test]
@@ -907,72 +748,48 @@ mod tests {
}
#[tokio::test]
async fn ensure_runtime_is_idempotent() -> anyhow::Result<()> {
async fn ensure_runtime_indexes_is_idempotent() {
let namespace = "indexes_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.context("in-memory db")?;
.expect("in-memory db");
db.apply_migrations()
.await
.context("migrations should succeed")?;
.expect("migrations should succeed");
ensure_runtime(&db, 1536)
// First run creates everything
ensure_runtime_indexes(&db, 1536)
.await
.context("first call should succeed")?;
ensure_runtime(&db, 1536)
.expect("initial index creation");
// Second run should be a no-op and still succeed
ensure_runtime_indexes(&db, 1536)
.await
.context("second index creation")?;
Ok(())
.expect("second index creation");
}
#[tokio::test]
async fn embedding_index_dimension_reflects_runtime_state() -> anyhow::Result<()> {
let namespace = "indexes_marker";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.context("in-memory db")?;
db.apply_migrations()
.await
.context("migrations should succeed")?;
// Before any index exists, there is no stored embedding dimension to detect.
assert_eq!(embedding_index_dimension(&db).await?, None);
ensure_runtime(&db, 1536)
.await
.context("initial index creation")?;
assert_eq!(embedding_index_dimension(&db).await?, Some(1536));
// After a dimension change the marker tracks the new index dimension.
ensure_runtime(&db, 256)
.await
.context("overwritten index creation")?;
assert_eq!(embedding_index_dimension(&db).await?, Some(256));
Ok(())
}
#[tokio::test]
async fn ensure_hnsw_index_overwrites_dimension() -> anyhow::Result<()> {
async fn ensure_hnsw_index_overwrites_dimension() {
let namespace = "indexes_dim";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.context("in-memory db")?;
.expect("in-memory db");
db.apply_migrations()
.await
.context("migrations should succeed")?;
.expect("migrations should succeed");
ensure_runtime(&db, 1536)
// Create initial index with default dimension
ensure_runtime_indexes(&db, 1536)
.await
.context("initial index creation")?;
ensure_runtime(&db, 128)
.expect("initial index creation");
// Change dimension and ensure overwrite path is exercised
ensure_runtime_indexes(&db, 128)
.await
.context("overwritten index creation")?;
Ok(())
.expect("overwritten index creation");
}
}
+97 -488
View File
@@ -2,24 +2,23 @@ use std::io::ErrorKind;
use std::path::{Component, Path, PathBuf};
use std::sync::Arc;
use anyhow::{anyhow, Context, Result as AnyResult};
use anyhow::{anyhow, Result as AnyResult};
use bytes::Bytes;
use futures::stream::BoxStream;
use futures::{StreamExt, TryStreamExt};
use object_store::aws::AmazonS3Builder;
use object_store::local::LocalFileSystem;
use object_store::memory::InMemory;
use object_store::{path::Path as ObjPath, ObjectStore};
use crate::utils::config::{AppConfig, StorageKind};
pub type DynStorage = Arc<dyn ObjectStore>;
pub type DynStore = Arc<dyn ObjectStore>;
/// Storage manager with persistent state and proper lifecycle management.
#[derive(Clone)]
pub struct StorageManager {
// Store from objectstore wrapped as dyn
store: DynStorage,
store: DynStore,
// Simple enum to track which kind
backend_kind: StorageKind,
// Where on disk
@@ -31,13 +30,8 @@ impl StorageManager {
///
/// This method validates the configuration and creates the appropriate
/// storage backend with proper initialization.
///
/// # Errors
///
/// Returns `Err` if the storage backend cannot be created or initialised
/// (e.g. missing S3 bucket, local filesystem permission error).
pub async fn new(cfg: &AppConfig) -> object_store::Result<Self> {
let backend_kind = cfg.storage;
let backend_kind = cfg.storage.clone();
let (store, local_base) = create_storage_backend(cfg).await?;
Ok(Self {
@@ -51,8 +45,7 @@ impl StorageManager {
///
/// This method is useful for testing scenarios where you want to inject
/// a specific storage backend.
#[must_use]
pub fn with_backend(store: DynStorage, backend_kind: StorageKind) -> Self {
pub fn with_backend(store: DynStore, backend_kind: StorageKind) -> Self {
Self {
store,
backend_kind,
@@ -61,13 +54,11 @@ impl StorageManager {
}
/// Get the storage backend kind.
#[must_use]
pub fn backend_kind(&self) -> &StorageKind {
&self.backend_kind
}
/// Access the resolved local base directory when using the local backend.
#[must_use]
pub fn local_base_path(&self) -> Option<&Path> {
self.local_base.as_deref()
}
@@ -76,7 +67,6 @@ impl StorageManager {
///
/// Returns `None` when the backend is not local or when the provided location includes
/// unsupported components (absolute paths or parent traversals).
#[must_use]
pub fn resolve_local_path(&self, location: &str) -> Option<PathBuf> {
let base = self.local_base_path()?;
let relative = Path::new(location);
@@ -95,10 +85,6 @@ impl StorageManager {
///
/// This operation persists data using the underlying storage backend.
/// For memory backends, data persists for the lifetime of the StorageManager.
///
/// # Errors
///
/// Returns `Err` if the underlying storage backend fails to persist the data.
pub async fn put(&self, location: &str, data: Bytes) -> object_store::Result<()> {
let path = ObjPath::from(location);
let payload = object_store::PutPayload::from_bytes(data);
@@ -107,27 +93,16 @@ impl StorageManager {
/// Retrieve bytes from the specified location.
///
/// Reads via [`Self::get_stream`] and buffers the full object in memory.
///
/// # Errors
///
/// Returns `Err` if the location does not exist or the underlying backend fails.
/// Returns the full contents buffered in memory.
pub async fn get(&self, location: &str) -> object_store::Result<Bytes> {
let mut stream = self.get_stream(location).await?;
let mut collected = Vec::new();
while let Some(chunk) = stream.next().await {
collected.extend_from_slice(&chunk?);
}
Ok(Bytes::from(collected))
let path = ObjPath::from(location);
let result = self.store.get(&path).await?;
result.bytes().await
}
/// Get a streaming handle for large objects.
///
/// Returns a fallible stream of Bytes chunks suitable for large file processing.
///
/// # Errors
///
/// Returns `Err` if the location does not exist or the underlying backend fails.
pub async fn get_stream(
&self,
location: &str,
@@ -140,10 +115,6 @@ impl StorageManager {
/// Delete all objects below the specified prefix.
///
/// For local filesystem backends, this also attempts to clean up empty directories.
///
/// # Errors
///
/// Returns `Err` if the underlying backend fails during deletion.
pub async fn delete_prefix(&self, prefix: &str) -> object_store::Result<()> {
let prefix_path = ObjPath::from(prefix);
let locations = self
@@ -165,10 +136,6 @@ impl StorageManager {
}
/// List all objects below the specified prefix.
///
/// # Errors
///
/// Returns `Err` if the underlying backend fails to list objects.
pub async fn list(
&self,
prefix: Option<&str>,
@@ -178,10 +145,6 @@ impl StorageManager {
}
/// Check if an object exists at the specified location.
///
/// # Errors
///
/// Returns `Err` if the underlying backend returns a non-NotFound error.
pub async fn exists(&self, location: &str) -> object_store::Result<bool> {
let path = ObjPath::from(location);
self.store
@@ -252,13 +215,10 @@ impl StorageManager {
/// storage backends with proper error handling and validation.
async fn create_storage_backend(
cfg: &AppConfig,
) -> object_store::Result<(DynStorage, Option<PathBuf>)> {
) -> object_store::Result<(DynStore, Option<PathBuf>)> {
match cfg.storage {
StorageKind::Local => {
let base = resolve_base_dir(cfg).map_err(|err| object_store::Error::Generic {
store: "LocalFileSystem",
source: err.into(),
})?;
let base = resolve_base_dir(cfg);
if !base.exists() {
tokio::fs::create_dir_all(&base).await.map_err(|e| {
object_store::Error::Generic {
@@ -274,37 +234,6 @@ async fn create_storage_backend(
let store = InMemory::new();
Ok((Arc::new(store), None))
}
StorageKind::S3 => {
let bucket = cfg
.s3_bucket
.as_ref()
.ok_or_else(|| object_store::Error::Generic {
store: "S3",
source: anyhow!("s3_bucket is required for S3 storage").into(),
})?;
let mut builder = AmazonS3Builder::new()
.with_bucket_name(bucket)
.with_allow_http(true);
if let (Ok(key), Ok(secret)) = (
std::env::var("AWS_ACCESS_KEY_ID"),
std::env::var("AWS_SECRET_ACCESS_KEY"),
) {
builder = builder
.with_access_key_id(key)
.with_secret_access_key(secret);
}
if let Some(endpoint) = &cfg.s3_endpoint {
builder = builder.with_endpoint(endpoint);
}
builder = builder.with_region(&cfg.s3_region);
let store = builder.build()?;
Ok((Arc::new(store), None))
}
}
}
@@ -314,38 +243,10 @@ async fn create_storage_backend(
/// automatic memory backend setup and proper test isolation.
#[cfg(test)]
pub mod testing {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use super::*;
use crate::utils::config::{AppConfig, PdfIngestMode};
use uuid;
const DEFAULT_TEST_S3_BUCKET: &str = "minne-tests";
const DEFAULT_TEST_S3_ENDPOINT: &str = "http://127.0.0.1:19000";
fn configured_test_s3_bucket() -> String {
std::env::var("MINNE_TEST_S3_BUCKET")
.ok()
.filter(|value| !value.trim().is_empty())
.or_else(|| {
std::env::var("S3_BUCKET")
.ok()
.filter(|value| !value.trim().is_empty())
})
.unwrap_or_else(|| DEFAULT_TEST_S3_BUCKET.to_string())
}
fn configured_test_s3_endpoint() -> String {
std::env::var("MINNE_TEST_S3_ENDPOINT")
.ok()
.filter(|value| !value.trim().is_empty())
.or_else(|| {
std::env::var("S3_ENDPOINT")
.ok()
.filter(|value| !value.trim().is_empty())
})
.unwrap_or_else(|| DEFAULT_TEST_S3_ENDPOINT.to_string())
}
/// Create a test configuration with memory storage.
///
/// This provides a ready-to-use configuration for testing scenarios
@@ -380,7 +281,7 @@ pub mod testing {
surrealdb_password: "test".into(),
surrealdb_namespace: "test".into(),
surrealdb_database: "test".into(),
data_dir: base,
data_dir: base.into(),
http_port: 0,
openai_base_url: "..".into(),
storage: StorageKind::Local,
@@ -389,30 +290,6 @@ pub mod testing {
}
}
/// Create a test configuration with S3 storage (MinIO).
///
/// Uses `MINNE_TEST_S3_ENDPOINT` / `S3_ENDPOINT` and
/// `MINNE_TEST_S3_BUCKET` / `S3_BUCKET` when provided.
pub fn test_config_s3() -> AppConfig {
AppConfig {
openai_api_key: "test".into(),
surrealdb_address: "test".into(),
surrealdb_username: "test".into(),
surrealdb_password: "test".into(),
surrealdb_namespace: "test".into(),
surrealdb_database: "test".into(),
data_dir: "/tmp/unused".into(),
http_port: 0,
openai_base_url: "..".into(),
storage: StorageKind::S3,
s3_bucket: Some(configured_test_s3_bucket()),
s3_endpoint: Some(configured_test_s3_endpoint()),
s3_region: "us-east-1".into(),
pdf_ingest_mode: PdfIngestMode::LlmFirst,
..Default::default()
}
}
/// A specialized StorageManager for testing scenarios.
///
/// This provides automatic setup for memory storage with proper isolation
@@ -420,7 +297,7 @@ pub mod testing {
#[derive(Clone)]
pub struct TestStorageManager {
storage: StorageManager,
temp_dir: Option<(String, std::path::PathBuf)>, // For local storage cleanup
_temp_dir: Option<(String, std::path::PathBuf)>, // For local storage cleanup
}
impl TestStorageManager {
@@ -434,7 +311,7 @@ pub mod testing {
Ok(Self {
storage,
temp_dir: None,
_temp_dir: None,
})
}
@@ -451,31 +328,7 @@ pub mod testing {
Ok(Self {
storage,
temp_dir: resolved,
})
}
/// Create a new TestStorageManager with S3 backend (MinIO).
///
/// This requires a reachable MinIO endpoint and an existing test bucket.
pub async fn new_s3() -> object_store::Result<Self> {
// Ensure credentials are set for MinIO
// We set these env vars for the process, which AmazonS3Builder will pick up
std::env::set_var("AWS_ACCESS_KEY_ID", "minioadmin");
std::env::set_var("AWS_SECRET_ACCESS_KEY", "minioadmin");
std::env::set_var("AWS_REGION", "us-east-1");
let cfg = test_config_s3();
let storage = StorageManager::new(&cfg).await?;
// Probe the bucket so tests can cleanly skip when the endpoint is unreachable
// or the test bucket is not provisioned.
let probe_prefix = format!("__minne_s3_probe__/{}", uuid::Uuid::new_v4());
storage.list(Some(&probe_prefix)).await?;
Ok(Self {
storage,
temp_dir: None,
_temp_dir: resolved,
})
}
@@ -490,7 +343,10 @@ pub mod testing {
None
};
Ok(Self { storage, temp_dir })
Ok(Self {
storage,
_temp_dir: temp_dir,
})
}
/// Get a reference to the underlying StorageManager.
@@ -513,14 +369,6 @@ pub mod testing {
self.storage.get(location).await
}
/// Get a streaming handle for test data.
pub async fn get_stream(
&self,
location: &str,
) -> object_store::Result<BoxStream<'static, object_store::Result<Bytes>>> {
self.storage.get_stream(location).await
}
/// Delete test data below the specified prefix.
pub async fn delete_prefix(&self, prefix: &str) -> object_store::Result<()> {
self.storage.delete_prefix(prefix).await
@@ -543,7 +391,7 @@ pub mod testing {
impl Drop for TestStorageManager {
fn drop(&mut self) {
// Clean up temporary directories for local storage
if let Some((_, path)) = &self.temp_dir {
if let Some((_, path)) = &self._temp_dir {
if path.exists() {
let _ = std::fs::remove_dir_all(path);
}
@@ -582,22 +430,14 @@ pub mod testing {
/// Resolve the absolute base directory used for local storage from config.
///
/// If `data_dir` is relative, it is resolved against the process current working directory.
///
/// # Errors
///
/// Returns `Err` when `data_dir` is relative and the current working directory cannot be read.
pub fn resolve_base_dir(cfg: &AppConfig) -> AnyResult<PathBuf> {
/// If `data_dir` is relative, it is resolved against the current working directory.
pub fn resolve_base_dir(cfg: &AppConfig) -> PathBuf {
if cfg.data_dir.starts_with('/') {
Ok(PathBuf::from(&cfg.data_dir))
PathBuf::from(&cfg.data_dir)
} else {
let cwd = std::env::current_dir().with_context(|| {
format!(
"failed to resolve relative data_dir '{}' against the current working directory",
cfg.data_dir
)
})?;
Ok(cwd.join(&cfg.data_dir))
std::env::current_dir()
.unwrap_or_else(|_| PathBuf::from("."))
.join(&cfg.data_dir)
}
}
@@ -626,10 +466,8 @@ pub fn split_object_path(path: &str) -> AnyResult<(String, String)> {
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use super::*;
use crate::utils::config::{PdfIngestMode::LlmFirst, StorageKind};
use anyhow::Context;
use bytes::Bytes;
use uuid::Uuid;
@@ -668,11 +506,11 @@ mod tests {
}
#[tokio::test]
async fn test_storage_manager_memory_basic_operations() -> anyhow::Result<()> {
async fn test_storage_manager_memory_basic_operations() {
let cfg = test_config_memory();
let storage = StorageManager::new(&cfg)
.await
.with_context(|| "create storage manager".to_string())?;
.expect("create storage manager");
assert!(storage.local_base_path().is_none());
let location = "test/data/file.txt";
@@ -682,42 +520,31 @@ mod tests {
storage
.put(location, Bytes::from(data.to_vec()))
.await
.with_context(|| "put".to_string())?;
let retrieved = storage
.get(location)
.await
.with_context(|| "get".to_string())?;
.expect("put");
let retrieved = storage.get(location).await.expect("get");
assert_eq!(retrieved.as_ref(), data);
// Test exists
assert!(storage
.exists(location)
.await
.with_context(|| "exists check".to_string())?);
assert!(storage.exists(location).await.expect("exists check"));
// Test delete
storage
.delete_prefix("test/data/")
.await
.with_context(|| "delete".to_string())?;
storage.delete_prefix("test/data/").await.expect("delete");
assert!(!storage
.exists(location)
.await
.with_context(|| "exists check after delete".to_string())?);
Ok(())
.expect("exists check after delete"));
}
#[tokio::test]
async fn test_storage_manager_local_basic_operations() -> anyhow::Result<()> {
async fn test_storage_manager_local_basic_operations() {
let base = format!("/tmp/minne_storage_test_{}", Uuid::new_v4());
let cfg = test_config(&base);
let storage = StorageManager::new(&cfg)
.await
.with_context(|| "create storage manager".to_string())?;
.expect("create storage manager");
let resolved_base = storage
.local_base_path()
.with_context(|| "resolved base dir".to_string())?
.expect("resolved base dir")
.to_path_buf();
assert_eq!(resolved_base, PathBuf::from(&base));
@@ -728,53 +555,42 @@ mod tests {
storage
.put(location, Bytes::from(data.to_vec()))
.await
.with_context(|| "put".to_string())?;
let retrieved = storage
.get(location)
.await
.with_context(|| "get".to_string())?;
.expect("put");
let retrieved = storage.get(location).await.expect("get");
assert_eq!(retrieved.as_ref(), data);
let object_dir = resolved_base.join("test/data");
tokio::fs::metadata(&object_dir)
.await
.with_context(|| "object directory exists after write".to_string())?;
.expect("object directory exists after write");
// Test exists
assert!(storage
.exists(location)
.await
.with_context(|| "exists check".to_string())?);
assert!(storage.exists(location).await.expect("exists check"));
// Test delete
storage
.delete_prefix("test/data/")
.await
.with_context(|| "delete".to_string())?;
storage.delete_prefix("test/data/").await.expect("delete");
assert!(!storage
.exists(location)
.await
.with_context(|| "exists check after delete".to_string())?);
.expect("exists check after delete"));
assert!(
tokio::fs::metadata(&object_dir).await.is_err(),
"object directory should be removed"
);
tokio::fs::metadata(&resolved_base)
.await
.with_context(|| "base directory remains intact".to_string())?;
.expect("base directory remains intact");
// Clean up
let _ = tokio::fs::remove_dir_all(&base).await;
Ok(())
}
#[tokio::test]
async fn test_storage_manager_memory_persistence() -> anyhow::Result<()> {
async fn test_storage_manager_memory_persistence() {
let cfg = test_config_memory();
let storage = StorageManager::new(&cfg)
.await
.with_context(|| "create storage manager".to_string())?;
.expect("create storage manager");
let location = "persistence/test.txt";
let data1 = b"first data";
@@ -784,40 +600,32 @@ mod tests {
storage
.put(location, Bytes::from(data1.to_vec()))
.await
.with_context(|| "put first".to_string())?;
.expect("put first");
// Retrieve and verify first data
let retrieved1 = storage
.get(location)
.await
.with_context(|| "get first".to_string())?;
let retrieved1 = storage.get(location).await.expect("get first");
assert_eq!(retrieved1.as_ref(), data1);
// Overwrite with second data
storage
.put(location, Bytes::from(data2.to_vec()))
.await
.with_context(|| "put second".to_string())?;
.expect("put second");
// Retrieve and verify second data
let retrieved2 = storage
.get(location)
.await
.with_context(|| "get second".to_string())?;
let retrieved2 = storage.get(location).await.expect("get second");
assert_eq!(retrieved2.as_ref(), data2);
// Data persists across multiple operations using the same StorageManager
assert_ne!(retrieved1.as_ref(), retrieved2.as_ref());
Ok(())
}
#[tokio::test]
async fn test_storage_manager_list_operations() -> anyhow::Result<()> {
async fn test_storage_manager_list_operations() {
let cfg = test_config_memory();
let storage = StorageManager::new(&cfg)
.await
.with_context(|| "create storage manager".to_string())?;
.expect("create storage manager");
// Create multiple files
let files = vec![
@@ -830,21 +638,15 @@ mod tests {
storage
.put(location, Bytes::from(data.to_vec()))
.await
.with_context(|| "put".to_string())?;
.expect("put");
}
// Test listing without prefix
let all_files = storage
.list(None)
.await
.with_context(|| "list all".to_string())?;
let all_files = storage.list(None).await.expect("list all");
assert_eq!(all_files.len(), 3);
// Test listing with prefix
let dir1_files = storage
.list(Some("dir1/"))
.await
.with_context(|| "list dir1".to_string())?;
let dir1_files = storage.list(Some("dir1/")).await.expect("list dir1");
assert_eq!(dir1_files.len(), 2);
assert!(dir1_files
.iter()
@@ -857,18 +659,16 @@ mod tests {
let empty_files = storage
.list(Some("nonexistent/"))
.await
.with_context(|| "list nonexistent".to_string())?;
.expect("list nonexistent");
assert_eq!(empty_files.len(), 0);
Ok(())
}
#[tokio::test]
async fn test_storage_manager_stream_operations() -> anyhow::Result<()> {
async fn test_storage_manager_stream_operations() {
let cfg = test_config_memory();
let storage = StorageManager::new(&cfg)
.await
.with_context(|| "create storage manager".to_string())?;
.expect("create storage manager");
let location = "stream/test.bin";
let content = vec![42u8; 1024 * 64]; // 64KB of data
@@ -877,27 +677,22 @@ mod tests {
storage
.put(location, Bytes::from(content.clone()))
.await
.with_context(|| "put large data".to_string())?;
.expect("put large data");
// Get as stream
let mut stream = storage
.get_stream(location)
.await
.with_context(|| "get stream".to_string())?;
let mut stream = storage.get_stream(location).await.expect("get stream");
let mut collected = Vec::new();
while let Some(chunk) = stream.next().await {
let chunk = chunk.with_context(|| "stream chunk".to_string())?;
let chunk = chunk.expect("stream chunk");
collected.extend_from_slice(&chunk);
}
assert_eq!(collected, content);
Ok(())
}
#[tokio::test]
async fn test_storage_manager_with_custom_backend() -> anyhow::Result<()> {
async fn test_storage_manager_with_custom_backend() {
use object_store::memory::InMemory;
// Create custom memory backend
@@ -911,28 +706,20 @@ mod tests {
storage
.put(location, Bytes::from(data.to_vec()))
.await
.with_context(|| "put".to_string())?;
let retrieved = storage
.get(location)
.await
.with_context(|| "get".to_string())?;
.expect("put");
let retrieved = storage.get(location).await.expect("get");
assert_eq!(retrieved.as_ref(), data);
assert!(storage
.exists(location)
.await
.with_context(|| "exists".to_string())?);
assert!(storage.exists(location).await.expect("exists"));
assert_eq!(*storage.backend_kind(), StorageKind::Memory);
Ok(())
}
#[tokio::test]
async fn test_storage_manager_error_handling() -> anyhow::Result<()> {
async fn test_storage_manager_error_handling() {
let cfg = test_config_memory();
let storage = StorageManager::new(&cfg)
.await
.with_context(|| "create storage manager".to_string())?;
.expect("create storage manager");
// Test getting non-existent file
let result = storage.get("nonexistent.txt").await;
@@ -942,290 +729,112 @@ mod tests {
let exists = storage
.exists("nonexistent.txt")
.await
.with_context(|| "exists check".to_string())?;
.expect("exists check");
assert!(!exists);
// Test listing with invalid location (should not panic)
let _result = storage.get("").await;
// This may or may not error depending on the backend implementation
// The important thing is that it doesn't panic
Ok(())
}
// TestStorageManager tests
#[tokio::test]
async fn test_test_storage_manager_memory() -> anyhow::Result<()> {
async fn test_test_storage_manager_memory() {
let test_storage = testing::TestStorageManager::new_memory()
.await
.with_context(|| "create test storage".to_string())?;
.expect("create test storage");
let location = "test/storage/file.txt";
let data = b"test data with TestStorageManager";
// Test put and get
test_storage
.put(location, data)
.await
.with_context(|| "put".to_string())?;
let retrieved = test_storage
.get(location)
.await
.with_context(|| "get".to_string())?;
test_storage.put(location, data).await.expect("put");
let retrieved = test_storage.get(location).await.expect("get");
assert_eq!(retrieved.as_ref(), data);
// Test existence check
assert!(test_storage
.exists(location)
.await
.with_context(|| "exists".to_string())?);
assert!(test_storage.exists(location).await.expect("exists"));
// Test list
let files = test_storage
.list(Some("test/storage/"))
.await
.with_context(|| "list".to_string())?;
.expect("list");
assert_eq!(files.len(), 1);
// Test delete
test_storage
.delete_prefix("test/storage/")
.await
.with_context(|| "delete".to_string())?;
.expect("delete");
assert!(!test_storage
.exists(location)
.await
.with_context(|| "exists after delete".to_string())?);
Ok(())
.expect("exists after delete"));
}
#[tokio::test]
async fn test_test_storage_manager_local() -> anyhow::Result<()> {
async fn test_test_storage_manager_local() {
let test_storage = testing::TestStorageManager::new_local()
.await
.with_context(|| "create test storage".to_string())?;
.expect("create test storage");
let location = "test/local/file.txt";
let data = b"test data with local TestStorageManager";
test_storage
.put(location, data)
.await
.with_context(|| "put".to_string())?;
let retrieved = test_storage
.get(location)
.await
.with_context(|| "get".to_string())?;
// Test put and get
test_storage.put(location, data).await.expect("put");
let retrieved = test_storage.get(location).await.expect("get");
assert_eq!(retrieved.as_ref(), data);
assert!(test_storage
.exists(location)
.await
.with_context(|| "exists".to_string())?);
// Test existence check
assert!(test_storage.exists(location).await.expect("exists"));
Ok(())
// The storage should be automatically cleaned up when test_storage is dropped
}
#[tokio::test]
async fn test_test_storage_manager_isolation() -> anyhow::Result<()> {
async fn test_test_storage_manager_isolation() {
let storage1 = testing::TestStorageManager::new_memory()
.await
.with_context(|| "create test storage 1".to_string())?;
.expect("create test storage 1");
let storage2 = testing::TestStorageManager::new_memory()
.await
.with_context(|| "create test storage 2".to_string())?;
.expect("create test storage 2");
let location = "isolation/test.txt";
let data1 = b"storage 1 data";
let data2 = b"storage 2 data";
storage1
.put(location, data1)
.await
.with_context(|| "put storage 1".to_string())?;
storage2
.put(location, data2)
.await
.with_context(|| "put storage 2".to_string())?;
// Put different data in each storage
storage1.put(location, data1).await.expect("put storage 1");
storage2.put(location, data2).await.expect("put storage 2");
let retrieved1 = storage1
.get(location)
.await
.with_context(|| "get storage 1".to_string())?;
let retrieved2 = storage2
.get(location)
.await
.with_context(|| "get storage 2".to_string())?;
// Verify isolation
let retrieved1 = storage1.get(location).await.expect("get storage 1");
let retrieved2 = storage2.get(location).await.expect("get storage 2");
assert_eq!(retrieved1.as_ref(), data1);
assert_eq!(retrieved2.as_ref(), data2);
assert_ne!(retrieved1.as_ref(), retrieved2.as_ref());
Ok(())
}
#[tokio::test]
async fn test_test_storage_manager_config() -> anyhow::Result<()> {
async fn test_test_storage_manager_config() {
let cfg = testing::test_config_memory();
let test_storage = testing::TestStorageManager::with_config(&cfg)
.await
.with_context(|| "create test storage with config".to_string())?;
.expect("create test storage with config");
let location = "config/test.txt";
let data = b"test data with custom config";
test_storage
.put(location, data)
.await
.with_context(|| "put".to_string())?;
let retrieved = test_storage
.get(location)
.await
.with_context(|| "get".to_string())?;
test_storage.put(location, data).await.expect("put");
let retrieved = test_storage.get(location).await.expect("get");
assert_eq!(retrieved.as_ref(), data);
// Verify it's using memory backend
assert_eq!(*test_storage.storage().backend_kind(), StorageKind::Memory);
Ok(())
}
// S3 Tests - Require a reachable MinIO endpoint and test bucket.
// `TestStorageManager::new_s3()` probes connectivity and these tests auto-skip when unavailable.
#[tokio::test]
async fn test_storage_manager_s3_basic_operations() -> anyhow::Result<()> {
// Skip if S3 connection fails (e.g. no MinIO)
let Ok(storage) = testing::TestStorageManager::new_s3().await else {
eprintln!("Skipping S3 test (setup failed)");
return Ok(());
};
let prefix = format!("test-basic-{}", Uuid::new_v4());
let location = format!("{prefix}/file.txt");
let data = b"test data for S3";
// Test put
if let Err(e) = storage.put(&location, data).await {
eprintln!("Skipping S3 test (put failed - bucket missing?): {e}");
return Ok(());
}
// Test get
let retrieved = storage
.get(&location)
.await
.with_context(|| "get".to_string())?;
assert_eq!(retrieved.as_ref(), data);
// Test exists
assert!(storage
.exists(&location)
.await
.with_context(|| "exists".to_string())?);
// Test delete
storage
.delete_prefix(&format!("{prefix}/"))
.await
.with_context(|| "delete".to_string())?;
assert!(!storage
.exists(&location)
.await
.with_context(|| "exists after delete".to_string())?);
Ok(())
}
#[tokio::test]
async fn test_storage_manager_s3_list_operations() -> anyhow::Result<()> {
let Ok(storage) = testing::TestStorageManager::new_s3().await else {
return Ok(());
};
let prefix = format!("test-list-{}", Uuid::new_v4());
let files = vec![
(format!("{prefix}/file1.txt"), b"content1"),
(format!("{prefix}/file2.txt"), b"content2"),
(format!("{prefix}/sub/file3.txt"), b"content3"),
];
for (loc, data) in &files {
if storage.put(loc, *data).await.is_err() {
return Ok(()); // Abort if put fails
}
}
// List with prefix
let list_prefix = format!("{prefix}/");
let items = storage
.list(Some(&list_prefix))
.await
.with_context(|| "list".to_string())?;
assert_eq!(items.len(), 3);
// Cleanup
storage
.delete_prefix(&list_prefix)
.await
.with_context(|| "cleanup".to_string())?;
Ok(())
}
#[tokio::test]
async fn test_storage_manager_s3_stream_operations() -> anyhow::Result<()> {
let Ok(storage) = testing::TestStorageManager::new_s3().await else {
return Ok(());
};
let prefix = format!("test-stream-{}", Uuid::new_v4());
let location = format!("{prefix}/large.bin");
let content = vec![42u8; 1024 * 10]; // 10KB
if storage.put(&location, &content).await.is_err() {
return Ok(());
}
let mut stream = storage
.get_stream(&location)
.await
.with_context(|| "get stream".to_string())?;
let mut collected = Vec::new();
while let Some(chunk) = stream.next().await {
collected.extend_from_slice(&chunk.with_context(|| "chunk".to_string())?);
}
assert_eq!(collected, content);
storage
.delete_prefix(&format!("{prefix}/"))
.await
.with_context(|| "cleanup".to_string())?;
Ok(())
}
#[tokio::test]
async fn test_storage_manager_s3_backend_kind() -> anyhow::Result<()> {
let Ok(storage) = testing::TestStorageManager::new_s3().await else {
return Ok(());
};
assert_eq!(*storage.storage().backend_kind(), StorageKind::S3);
Ok(())
}
#[tokio::test]
async fn test_storage_manager_s3_error_handling() -> anyhow::Result<()> {
let Ok(storage) = testing::TestStorageManager::new_s3().await else {
return Ok(());
};
let location = format!("nonexistent-{}/file.txt", Uuid::new_v4());
assert!(storage.get(&location).await.is_err());
// exists may fail if S3 is unavailable; treat error as false
assert!(!storage.exists(&location).await.unwrap_or(false));
Ok(())
}
}
+95 -149
View File
@@ -1,5 +1,4 @@
use crate::storage::types::{user::User, StoredObject};
use crate::utils::serde_helpers::deserialize_flexible_id;
use crate::storage::types::{file_info::deserialize_flexible_id, user::User, StoredObject};
use serde::{Deserialize, Serialize};
use crate::{error::AppError, storage::db::SurrealDbClient};
@@ -17,71 +16,55 @@ impl StoredObject for Analytics {
"analytics"
}
fn id(&self) -> &str {
fn get_id(&self) -> &str {
&self.id
}
}
impl Analytics {
const RECORD_ID: &'static str = "current";
/// Ensures the singleton analytics record exists (idempotent).
///
/// Production databases are also seeded by `20250503_215025_initial_setup.surql`;
/// this uses an atomic `UPSERT` for tests and recovery.
pub async fn ensure_initialized(db: &SurrealDbClient) -> Result<Self, AppError> {
let analytics: Option<Self> = db
.client
.query(
"UPSERT type::thing('analytics', $id) SET visitors = visitors ?? 0, page_loads = page_loads ?? 0 RETURN AFTER",
)
.bind(("id", Self::RECORD_ID))
.await?
.take(0)?;
let analytics = db.get_item::<Self>("current").await?;
if analytics.is_none() {
let created_analytics = Analytics {
id: "current".to_string(),
visitors: 0,
page_loads: 0,
};
let stored: Option<Self> = db.store_item(created_analytics).await?;
return stored.ok_or(AppError::Validation(
"Failed to initialize analytics".into(),
));
}
analytics.ok_or(AppError::Validation(
"failed to initialize analytics".into(),
"Failed to initialize analytics".into(),
))
}
pub async fn get_current(db: &SurrealDbClient) -> Result<Self, AppError> {
let analytics: Option<Self> = db.get_item("current").await?;
analytics.ok_or(AppError::NotFound("analytics not found".into()))
analytics.ok_or(AppError::NotFound("Analytics not found".into()))
}
pub async fn increment_visitors(db: &SurrealDbClient) -> Result<Self, AppError> {
let updated: Option<Self> = db
.client
.query(
"UPSERT type::thing('analytics', $id) SET visitors = (visitors ?? 0) + 1, page_loads = page_loads ?? 0 RETURN AFTER",
)
.bind(("id", Self::RECORD_ID))
.query("UPDATE type::thing('analytics', 'current') SET visitors += 1 RETURN AFTER")
.await?
.take(0)?;
updated.ok_or(AppError::Validation("failed to update analytics".into()))
updated.ok_or(AppError::Validation("Failed to update analytics".into()))
}
pub async fn increment_page_loads(db: &SurrealDbClient) -> Result<Self, AppError> {
Self::record_page_view(db, false).await
}
/// Records a page view, optionally counting the visitor as new.
pub async fn record_page_view(
db: &SurrealDbClient,
is_new_visitor: bool,
) -> Result<Self, AppError> {
let visitor_delta = i64::from(is_new_visitor);
let updated: Option<Self> = db
.client
.query(
"UPSERT type::thing('analytics', $id) SET page_loads = (page_loads ?? 0) + 1, visitors = (visitors ?? 0) + $visitor_delta RETURN AFTER",
)
.bind(("id", Self::RECORD_ID))
.bind(("visitor_delta", visitor_delta))
.query("UPDATE type::thing('analytics', 'current') SET page_loads += 1 RETURN AFTER")
.await?
.take(0)?;
updated.ok_or(AppError::Validation("failed to update analytics".into()))
updated.ok_or(AppError::Validation("Failed to update analytics".into()))
}
pub async fn get_users_amount(db: &SurrealDbClient) -> Result<i64, AppError> {
@@ -105,10 +88,8 @@ impl Analytics {
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use super::*;
use crate::stored_object;
use anyhow::{self};
use uuid::Uuid;
stored_object!(TestUser, "user", {
@@ -118,14 +99,18 @@ mod tests {
});
#[tokio::test]
async fn test_analytics_initialization() -> anyhow::Result<()> {
async fn test_analytics_initialization() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("Failed to start in-memory surrealdb");
// Test initialization of analytics
let analytics = Analytics::ensure_initialized(&db).await?;
let analytics = Analytics::ensure_initialized(&db)
.await
.expect("Failed to initialize analytics");
// Verify initial state after initialization
assert_eq!(analytics.id, "current");
@@ -133,198 +118,159 @@ mod tests {
assert_eq!(analytics.visitors, 0);
// Test idempotency - ensure calling it again doesn't change anything
let analytics_again = Analytics::ensure_initialized(&db).await?;
let analytics_again = Analytics::ensure_initialized(&db)
.await
.expect("Failed to get analytics after initialization");
assert_eq!(analytics.id, analytics_again.id);
assert_eq!(analytics.page_loads, analytics_again.page_loads);
assert_eq!(analytics.visitors, analytics_again.visitors);
Ok(())
}
#[tokio::test]
async fn test_get_current_analytics() -> anyhow::Result<()> {
async fn test_get_current_analytics() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("Failed to start in-memory surrealdb");
// Initialize analytics
Analytics::ensure_initialized(&db).await?;
Analytics::ensure_initialized(&db)
.await
.expect("Failed to initialize analytics");
// Test get_current method
let analytics = Analytics::get_current(&db).await?;
let analytics = Analytics::get_current(&db)
.await
.expect("Failed to get current analytics");
assert_eq!(analytics.id, "current");
assert_eq!(analytics.page_loads, 0);
assert_eq!(analytics.visitors, 0);
Ok(())
}
#[tokio::test]
async fn test_increment_visitors() -> anyhow::Result<()> {
async fn test_increment_visitors() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("Failed to start in-memory surrealdb");
// Initialize analytics
Analytics::ensure_initialized(&db).await?;
Analytics::ensure_initialized(&db)
.await
.expect("Failed to initialize analytics");
// Test increment_visitors method
let analytics = Analytics::increment_visitors(&db).await?;
let analytics = Analytics::increment_visitors(&db)
.await
.expect("Failed to increment visitors");
assert_eq!(analytics.visitors, 1);
assert_eq!(analytics.page_loads, 0);
// Increment again and check
let analytics = Analytics::increment_visitors(&db).await?;
let analytics = Analytics::increment_visitors(&db)
.await
.expect("Failed to increment visitors again");
assert_eq!(analytics.visitors, 2);
assert_eq!(analytics.page_loads, 0);
Ok(())
}
#[tokio::test]
async fn test_increment_page_loads() -> anyhow::Result<()> {
async fn test_increment_page_loads() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("Failed to start in-memory surrealdb");
// Initialize analytics
Analytics::ensure_initialized(&db).await?;
Analytics::ensure_initialized(&db)
.await
.expect("Failed to initialize analytics");
// Test increment_page_loads method
let analytics = Analytics::increment_page_loads(&db).await?;
let analytics = Analytics::increment_page_loads(&db)
.await
.expect("Failed to increment page loads");
assert_eq!(analytics.visitors, 0);
assert_eq!(analytics.page_loads, 1);
// Increment again and check
let analytics = Analytics::increment_page_loads(&db).await?;
let analytics = Analytics::increment_page_loads(&db)
.await
.expect("Failed to increment page loads again");
assert_eq!(analytics.visitors, 0);
assert_eq!(analytics.page_loads, 2);
Ok(())
}
#[tokio::test]
async fn test_get_users_amount() -> anyhow::Result<()> {
async fn test_get_users_amount() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("Failed to start in-memory surrealdb");
// Test with no users
let count = Analytics::get_users_amount(&db).await?;
let count = Analytics::get_users_amount(&db)
.await
.expect("Failed to get users amount");
assert_eq!(count, 0);
// Create a few test users
for i in 0..3 {
let user = TestUser {
id: format!("user{i}"),
email: format!("user{i}@example.com"),
id: format!("user{}", i),
email: format!("user{}@example.com", i),
password: "password".to_string(),
user_id: format!("uid{i}"),
user_id: format!("uid{}", i),
created_at: Utc::now(),
updated_at: Utc::now(),
};
db.store_item(user).await?;
db.store_item(user)
.await
.expect("Failed to create test user");
}
// Test users amount after adding users
let count = Analytics::get_users_amount(&db).await?;
let count = Analytics::get_users_amount(&db)
.await
.expect("Failed to get users amount after adding users");
assert_eq!(count, 3);
Ok(())
}
#[tokio::test]
async fn test_increment_visitors_without_prior_init() -> anyhow::Result<()> {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let analytics = Analytics::increment_visitors(&db).await?;
assert_eq!(analytics.visitors, 1);
assert_eq!(analytics.page_loads, 0);
Ok(())
}
#[tokio::test]
async fn test_increment_page_loads_without_prior_init() -> anyhow::Result<()> {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let analytics = Analytics::increment_page_loads(&db).await?;
assert_eq!(analytics.page_loads, 1);
assert_eq!(analytics.visitors, 0);
Ok(())
}
#[tokio::test]
async fn test_visitor_and_page_load_increments_are_independent() -> anyhow::Result<()> {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let after_visitors = Analytics::increment_visitors(&db).await?;
assert_eq!(after_visitors.visitors, 1);
assert_eq!(after_visitors.page_loads, 0);
let after_page_load = Analytics::increment_page_loads(&db).await?;
assert_eq!(after_page_load.visitors, 1);
assert_eq!(after_page_load.page_loads, 1);
let after_second_visitor = Analytics::increment_visitors(&db).await?;
assert_eq!(after_second_visitor.visitors, 2);
assert_eq!(after_second_visitor.page_loads, 1);
Ok(())
}
#[tokio::test]
async fn test_record_page_view() -> anyhow::Result<()> {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let first_view = Analytics::record_page_view(&db, true).await?;
assert_eq!(first_view.visitors, 1);
assert_eq!(first_view.page_loads, 1);
let returning_view = Analytics::record_page_view(&db, false).await?;
assert_eq!(returning_view.visitors, 1);
assert_eq!(returning_view.page_loads, 2);
Ok(())
}
#[tokio::test]
async fn test_get_current_nonexistent() -> anyhow::Result<()> {
async fn test_get_current_nonexistent() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("Failed to start in-memory surrealdb");
// Don't initialize analytics and try to get it
let result = Analytics::get_current(&db).await;
assert!(result.is_err());
match result {
Ok(_) => anyhow::bail!("Expected NotFound error, got success"),
Err(AppError::NotFound(_)) => {}
Err(err) => anyhow::bail!("Expected NotFound error, got: {err:?}"),
if let Err(err) = result {
match err {
AppError::NotFound(_) => {
// Expected error
}
_ => panic!("Expected NotFound error, got: {:?}", err),
}
}
Ok(())
}
}
+67 -407
View File
@@ -10,57 +10,7 @@ stored_object!(Conversation, "conversation", {
title: String
});
#[derive(Debug, Clone, serde::Serialize, serde::Deserialize, PartialEq, Eq)]
#[allow(clippy::module_name_repetitions)]
pub struct SidebarConversation {
#[serde(deserialize_with = "deserialize_sidebar_id")]
pub id: String,
pub title: String,
}
struct SidebarIdVisitor;
impl<'de> serde::de::Visitor<'de> for SidebarIdVisitor {
type Value = String;
fn expecting(&self, formatter: &mut std::fmt::Formatter) -> std::fmt::Result {
formatter.write_str("a string id or a SurrealDB Thing")
}
fn visit_str<E>(self, value: &str) -> Result<Self::Value, E>
where
E: serde::de::Error,
{
Ok(value.to_string())
}
fn visit_string<E>(self, value: String) -> Result<Self::Value, E>
where
E: serde::de::Error,
{
Ok(value)
}
fn visit_map<A>(self, map: A) -> Result<Self::Value, A::Error>
where
A: serde::de::MapAccess<'de>,
{
let thing = <surrealdb::sql::Thing as serde::Deserialize>::deserialize(
serde::de::value::MapAccessDeserializer::new(map),
)?;
Ok(thing.id.to_raw())
}
}
fn deserialize_sidebar_id<'de, D>(deserializer: D) -> Result<String, D::Error>
where
D: serde::Deserializer<'de>,
{
deserializer.deserialize_any(SidebarIdVisitor)
}
impl Conversation {
#[must_use]
pub fn new(user_id: String, title: String) -> Self {
let now = Utc::now();
Self {
@@ -80,7 +30,7 @@ impl Conversation {
let conversation: Conversation = db
.get_item(conversation_id)
.await?
.ok_or_else(|| AppError::NotFound("conversation not found".to_string()))?;
.ok_or_else(|| AppError::NotFound("Conversation not found".to_string()))?;
if conversation.user_id != user_id {
return Err(AppError::Auth(
@@ -88,15 +38,10 @@ impl Conversation {
));
}
let messages: Vec<Message> = db
.client
.query(
"SELECT * FROM type::table($message_table) WHERE conversation_id = $conversation_id AND type::thing($conversation_table, $conversation_id).user_id = $user_id ORDER BY updated_at",
)
.bind(("message_table", Message::table_name()))
.bind(("conversation_table", Self::table_name()))
.bind(("conversation_id", conversation_id.to_string()))
.bind(("user_id", user_id.to_string()))
let messages:Vec<Message> = db.client.
query("SELECT * FROM type::table($table_name) WHERE conversation_id = $conversation_id ORDER BY updated_at").
bind(("table_name", Message::table_name())).
bind(("conversation_id", conversation_id.to_string()))
.await?
.take(0)?;
@@ -111,7 +56,7 @@ impl Conversation {
// First verify ownership by getting conversation user_id
let conversation: Option<Conversation> = db.get_item(id).await?;
let conversation =
conversation.ok_or_else(|| AppError::NotFound("conversation not found".to_string()))?;
conversation.ok_or_else(|| AppError::NotFound("Conversation not found".to_string()))?;
if conversation.user_id != user_id {
return Err(AppError::Auth(
@@ -119,7 +64,7 @@ impl Conversation {
));
}
let updated: Option<Self> = db
let _updated: Option<Self> = db
.update((Self::table_name(), id))
.patch(PatchOp::replace("/title", new_title.to_string()))
.patch(PatchOp::replace(
@@ -128,118 +73,82 @@ impl Conversation {
))
.await?;
if updated.is_none() {
return Err(AppError::NotFound("conversation not found".to_string()));
}
Ok(())
}
pub async fn get_user_sidebar_conversations(
user_id: &str,
db: &SurrealDbClient,
) -> Result<Vec<SidebarConversation>, AppError> {
let conversations: Vec<SidebarConversation> = db
.client
.query(
"SELECT id, title, updated_at FROM type::table($table_name) WHERE user_id = $user_id ORDER BY updated_at DESC",
)
.bind(("table_name", Self::table_name()))
.bind(("user_id", user_id.to_string()))
.await?
.take(0)?;
Ok(conversations)
}
}
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use crate::storage::types::message::MessageRole;
use anyhow::{self, Context};
use super::*;
const MESSAGE_QUERY_FOR_OWNER: &str = "SELECT * FROM type::table($message_table) WHERE conversation_id = $conversation_id AND type::thing($conversation_table, $conversation_id).user_id = $user_id ORDER BY updated_at";
async fn fetch_messages_for_owner(
db: &SurrealDbClient,
conversation_id: &str,
user_id: &str,
) -> Result<Vec<Message>, AppError> {
db.client
.query(MESSAGE_QUERY_FOR_OWNER)
.bind(("message_table", Message::table_name()))
.bind(("conversation_table", Conversation::table_name()))
.bind(("conversation_id", conversation_id.to_string()))
.bind(("user_id", user_id.to_string()))
.await?
.take(0)
.map_err(AppError::from)
}
#[tokio::test]
async fn test_create_conversation() -> anyhow::Result<()> {
async fn test_create_conversation() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
// Create a new conversation
let user_id = "test_user";
let title = "Test Conversation";
let conversation = Conversation::new(user_id.to_string(), title.to_string());
// Verify conversation properties
assert_eq!(conversation.user_id, user_id);
assert_eq!(conversation.title, title);
assert!(!conversation.id.is_empty());
// Store the conversation
let result = db.store_item(conversation.clone()).await;
assert!(result.is_ok());
// Verify it can be retrieved
let retrieved: Option<Conversation> = db
.get_item(&conversation.id)
.await
.with_context(|| "Failed to retrieve conversation".to_string())?;
.expect("Failed to retrieve conversation");
assert!(retrieved.is_some());
let retrieved =
retrieved.ok_or_else(|| anyhow::anyhow!("Expected conversation to exist"))?;
let retrieved = retrieved.unwrap();
assert_eq!(retrieved.id, conversation.id);
assert_eq!(retrieved.user_id, user_id);
assert_eq!(retrieved.title, title);
Ok(())
}
#[tokio::test]
async fn test_get_complete_conversation_not_found() -> anyhow::Result<()> {
async fn test_get_complete_conversation_not_found() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
// Try to get a conversation that doesn't exist
let result =
Conversation::get_complete_conversation("nonexistent_id", "test_user", &db).await;
assert!(result.is_err());
match result {
Err(AppError::NotFound(_)) => {}
_ => anyhow::bail!("Expected NotFound error"),
Err(AppError::NotFound(_)) => { /* expected error */ }
_ => panic!("Expected NotFound error"),
}
Ok(())
}
#[tokio::test]
async fn test_get_complete_conversation_unauthorized() -> anyhow::Result<()> {
async fn test_get_complete_conversation_unauthorized() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
// Create and store a conversation for user_id_1
let user_id_1 = "user_1";
let conversation =
Conversation::new(user_id_1.to_string(), "Private Conversation".to_string());
@@ -247,28 +156,27 @@ mod tests {
db.store_item(conversation)
.await
.with_context(|| "Failed to store conversation".to_string())?;
.expect("Failed to store conversation");
// Try to access with a different user
let user_id_2 = "user_2";
let result =
Conversation::get_complete_conversation(&conversation_id, user_id_2, &db).await;
assert!(result.is_err());
match result {
Err(AppError::Auth(_)) => {}
_ => anyhow::bail!("Expected Auth error"),
Err(AppError::Auth(_)) => { /* expected error */ }
_ => panic!("Expected Auth error"),
}
Ok(())
}
#[tokio::test]
async fn test_patch_title_success() -> anyhow::Result<()> {
async fn test_patch_title_success() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
let user_id = "user_1";
let original_title = "Original Title";
@@ -277,50 +185,49 @@ mod tests {
db.store_item(conversation)
.await
.with_context(|| "Failed to store conversation".to_string())?;
.expect("Failed to store conversation");
let new_title = "Updated Title";
// Patch title successfully
let result = Conversation::patch_title(&conversation_id, user_id, new_title, &db).await;
assert!(result.is_ok());
// Retrieve from DB to verify
let updated_conversation = db
.get_item::<Conversation>(&conversation_id)
.await
.with_context(|| "Failed to get conversation".to_string())?
.ok_or_else(|| anyhow::anyhow!("Conversation missing"))?;
.expect("Failed to get conversation")
.expect("Conversation missing");
assert_eq!(updated_conversation.title, new_title);
assert_eq!(updated_conversation.user_id, user_id);
Ok(())
}
#[tokio::test]
async fn test_patch_title_not_found() -> anyhow::Result<()> {
async fn test_patch_title_not_found() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
// Try to patch non-existing conversation
let result = Conversation::patch_title("nonexistent", "user_x", "New Title", &db).await;
assert!(result.is_err());
match result {
Err(AppError::NotFound(_)) => {}
_ => anyhow::bail!("Expected NotFound error"),
_ => panic!("Expected NotFound error"),
}
Ok(())
}
#[tokio::test]
async fn test_patch_title_unauthorized() -> anyhow::Result<()> {
async fn test_patch_title_unauthorized() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
let owner_id = "owner";
let other_user_id = "intruder";
@@ -329,131 +236,38 @@ mod tests {
db.store_item(conversation)
.await
.with_context(|| "Failed to store conversation".to_string())?;
.expect("Failed to store conversation");
// Attempt patch with unauthorized user
let result =
Conversation::patch_title(&conversation_id, other_user_id, "Hacked Title", &db).await;
assert!(result.is_err());
match result {
Err(AppError::Auth(_)) => {}
_ => anyhow::bail!("Expected Auth error"),
_ => panic!("Expected Auth error"),
}
Ok(())
}
#[tokio::test]
async fn test_get_user_sidebar_conversations_filters_and_orders_by_updated_at_desc() {
async fn test_get_complete_conversation_with_messages() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("Failed to start in-memory surrealdb");
let user_id = "sidebar_user";
let other_user_id = "other_user";
let base = Utc::now();
let mut oldest = Conversation::new(user_id.to_string(), "Oldest".to_string());
oldest.updated_at = base - chrono::Duration::minutes(30);
let mut newest = Conversation::new(user_id.to_string(), "Newest".to_string());
newest.updated_at = base - chrono::Duration::minutes(5);
let mut middle = Conversation::new(user_id.to_string(), "Middle".to_string());
middle.updated_at = base - chrono::Duration::minutes(15);
let mut other_user = Conversation::new(other_user_id.to_string(), "Other".to_string());
other_user.updated_at = base;
db.store_item(oldest.clone())
.await
.expect("Failed to store oldest conversation");
db.store_item(newest.clone())
.await
.expect("Failed to store newest conversation");
db.store_item(middle.clone())
.await
.expect("Failed to store middle conversation");
db.store_item(other_user)
.await
.expect("Failed to store other-user conversation");
let sidebar_items = Conversation::get_user_sidebar_conversations(user_id, &db)
.await
.expect("Failed to get sidebar conversations");
assert_eq!(sidebar_items.len(), 3);
let s0 = sidebar_items.first().expect("expected 3 items");
let s1 = sidebar_items.get(1).expect("expected 3 items");
let s2 = sidebar_items.get(2).expect("expected 3 items");
assert_eq!(s0.id, newest.id);
assert_eq!(s0.title, "Newest");
assert_eq!(s1.id, middle.id);
assert_eq!(s1.title, "Middle");
assert_eq!(s2.id, oldest.id);
assert_eq!(s2.title, "Oldest");
}
#[tokio::test]
async fn test_sidebar_projection_reflects_patch_title_and_updated_at_reorder() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("Failed to start in-memory surrealdb");
let user_id = "sidebar_patch_user";
let base = Utc::now();
let mut first = Conversation::new(user_id.to_string(), "First".to_string());
first.updated_at = base - chrono::Duration::minutes(20);
let mut second = Conversation::new(user_id.to_string(), "Second".to_string());
second.updated_at = base - chrono::Duration::minutes(10);
db.store_item(first.clone())
.await
.expect("Failed to store first conversation");
db.store_item(second.clone())
.await
.expect("Failed to store second conversation");
let before_patch = Conversation::get_user_sidebar_conversations(user_id, &db)
.await
.expect("Failed to get sidebar conversations before patch");
let before = before_patch.first().expect("expected at least 1 item");
assert_eq!(before.id, second.id);
Conversation::patch_title(&first.id, user_id, "First (renamed)", &db)
.await
.expect("Failed to patch conversation title");
let after_patch = Conversation::get_user_sidebar_conversations(user_id, &db)
.await
.expect("Failed to get sidebar conversations after patch");
let after = after_patch.first().expect("expected at least 1 item");
assert_eq!(after.id, first.id);
assert_eq!(after.title, "First (renamed)");
}
#[tokio::test]
async fn test_get_complete_conversation_with_messages() -> anyhow::Result<()> {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
// Create and store a conversation for user_id_1
let user_id_1 = "user_1";
let conversation = Conversation::new(user_id_1.to_string(), "Conversation".to_string());
let conversation_id = conversation.id.clone();
db.store_item(conversation)
.await
.with_context(|| "Failed to store conversation".to_string())?;
.expect("Failed to store conversation");
// Create messages
let message1 = Message::new(
conversation_id.clone(),
MessageRole::User,
@@ -473,200 +287,46 @@ mod tests {
None,
);
// Store messages
db.store_item(message1)
.await
.with_context(|| "Failed to store message1".to_string())?;
.expect("Failed to store message1");
db.store_item(message2)
.await
.with_context(|| "Failed to store message2".to_string())?;
.expect("Failed to store message2");
db.store_item(message3)
.await
.with_context(|| "Failed to store message3".to_string())?;
.expect("Failed to store message3");
// Retrieve the complete conversation
let result =
Conversation::get_complete_conversation(&conversation_id, user_id_1, &db).await;
assert!(result.is_ok(), "Failed to retrieve complete conversation");
let (retrieved_conversation, retrieved_messages) =
result.with_context(|| "Failed to retrieve complete conversation".to_string())?;
let (retrieved_conversation, messages) = result.unwrap();
// Verify conversation data
assert_eq!(retrieved_conversation.id, conversation_id);
assert_eq!(retrieved_conversation.user_id, user_id_1);
assert_eq!(retrieved_conversation.title, "Conversation");
assert_eq!(retrieved_messages.len(), 3);
// Verify messages
assert_eq!(messages.len(), 3);
let message_contents: Vec<&str> = retrieved_messages
.iter()
.map(|m| m.content.as_str())
.collect();
// Verify messages are sorted by updated_at
let message_contents: Vec<&str> = messages.iter().map(|m| m.content.as_str()).collect();
assert!(message_contents.contains(&"Hello, AI!"));
assert!(message_contents.contains(&"Hello, human! How can I help you today?"));
assert!(message_contents.contains(&"Tell me about Rust programming."));
// Make sure we can't access with different user
let user_id_2 = "user_2";
let unauthorized_result =
Conversation::get_complete_conversation(&conversation_id, user_id_2, &db).await;
assert!(unauthorized_result.is_err());
match unauthorized_result {
Err(AppError::Auth(_)) => {}
_ => anyhow::bail!("Expected Auth error"),
}
Ok(())
}
#[test]
fn test_sidebar_conversation_deserializes_plain_string_id() {
let item: SidebarConversation =
serde_json::from_str(r#"{"id":"conv-plain","title":"My chat"}"#)
.expect("valid sidebar conversation json");
assert_eq!(item.id, "conv-plain");
assert_eq!(item.title, "My chat");
}
#[tokio::test]
async fn test_sidebar_conversation_deserializes_id_from_db_record() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("Failed to start in-memory surrealdb");
let owner = "sidebar_owner";
let conversation = Conversation::new(owner.to_string(), "Sidebar title".to_string());
let expected_id = conversation.id.clone();
db.store_item(conversation)
.await
.expect("Failed to store conversation");
let items = Conversation::get_user_sidebar_conversations(owner, &db)
.await
.expect("Failed to load sidebar");
assert_eq!(items.len(), 1);
let item = items.first().expect("expected one sidebar item");
assert_eq!(item.id, expected_id);
assert_eq!(item.title, "Sidebar title");
}
#[tokio::test]
async fn test_message_query_filters_by_owner_user_id_in_sql() -> anyhow::Result<()> {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let owner = "owner_user";
let intruder = "intruder_user";
let conversation = Conversation::new(owner.to_string(), "Private".to_string());
let conversation_id = conversation.id.clone();
db.store_item(conversation).await?;
db.store_item(Message::new(
conversation_id.clone(),
MessageRole::User,
"secret message".to_string(),
None,
))
.await?;
let owner_messages = fetch_messages_for_owner(&db, &conversation_id, owner).await?;
assert_eq!(owner_messages.len(), 1);
assert_eq!(
owner_messages
.first()
.expect("expected owner message")
.content,
"secret message"
);
let intruder_messages = fetch_messages_for_owner(&db, &conversation_id, intruder).await?;
assert!(
intruder_messages.is_empty(),
"SQL owner filter must not return messages for a non-owner user_id"
);
Ok(())
}
#[tokio::test]
async fn test_get_complete_conversation_orders_messages_by_updated_at() -> anyhow::Result<()> {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let user_id = "order_user";
let conversation = Conversation::new(user_id.to_string(), "Ordered".to_string());
let conversation_id = conversation.id.clone();
db.store_item(conversation).await?;
let base = Utc::now();
let mut first = Message::new(
conversation_id.clone(),
MessageRole::User,
"first".to_string(),
None,
);
first.updated_at = base - chrono::Duration::minutes(20);
let mut second = Message::new(
conversation_id.clone(),
MessageRole::AI,
"second".to_string(),
None,
);
second.updated_at = base - chrono::Duration::minutes(5);
db.store_item(first).await?;
db.store_item(second).await?;
let (_, messages) =
Conversation::get_complete_conversation(&conversation_id, user_id, &db).await?;
assert_eq!(messages.len(), 2);
assert_eq!(
messages.first().expect("expected first message").content,
"first"
);
assert_eq!(
messages.get(1).expect("expected second message").content,
"second"
);
Ok(())
}
#[tokio::test]
async fn test_patch_title_not_found_when_conversation_deleted() -> anyhow::Result<()> {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database).await?;
let owner = "owner";
let conversation = Conversation::new(owner.to_string(), "To delete".to_string());
let conversation_id = conversation.id.clone();
db.store_item(conversation).await?;
db.delete_item::<Conversation>(&conversation_id).await?;
let result = Conversation::patch_title(&conversation_id, owner, "New title", &db).await;
assert!(result.is_err());
match result {
Err(AppError::NotFound(_)) => {}
other => anyhow::bail!("expected NotFound, got {other:?}"),
}
Ok(())
}
#[tokio::test]
async fn test_conversation_new_initializes_timestamps_and_id() {
let before = Utc::now();
let conversation = Conversation::new("user".to_string(), "Title".to_string());
let after = Utc::now();
assert!(!conversation.id.is_empty());
assert!(conversation.created_at >= before && conversation.created_at <= after);
assert_eq!(conversation.created_at, conversation.updated_at);
assert_eq!(conversation.user_id, "user");
assert_eq!(conversation.title, "Title");
Err(AppError::Auth(_)) => { /* expected error */ }
_ => panic!("Expected Auth error"),
}
}
}
File diff suppressed because it is too large Load Diff
+130 -316
View File
@@ -1,4 +1,9 @@
#![allow(clippy::result_large_err)]
#![allow(
clippy::result_large_err,
clippy::needless_pass_by_value,
clippy::implicit_clone,
clippy::semicolon_if_nothing_returned
)]
use crate::{error::AppError, storage::types::file_info::FileInfo};
use serde::{Deserialize, Serialize};
use tracing::info;
@@ -26,150 +31,78 @@ pub enum IngestionPayload {
},
}
impl Default for IngestionPayload {
/// An empty text payload, used as a cheap placeholder when the real content
/// has been moved out of a task (see [`crate::storage::types::ingestion_task::IngestionTask::take_content`]).
fn default() -> Self {
Self::Text {
text: String::new(),
context: String::new(),
category: String::new(),
user_id: String::new(),
}
}
}
/// Shared ingest metadata moved or cloned into each payload variant.
struct IngestFields {
context: String,
category: String,
user_id: String,
}
/// Result of parsing optional ingest content before file payloads are built.
#[derive(Debug)]
enum ParsedContent {
/// No URL or text payload should be appended.
Skip,
Url(String),
Text(String),
}
impl ParsedContent {
#[must_use]
fn follows(&self) -> bool {
!matches!(self, Self::Skip)
}
}
impl IngestionPayload {
/// Creates ingestion payloads from the provided content, context, and files.
///
/// Files are emitted first. When both files and content are present, shared
/// metadata is cloned per file; otherwise the last file-only payload moves
/// `context`, `category`, and `user_id` without cloning.
/// # Arguments
/// * `content` - Optional textual content to be ingressed
/// * `context` - context for processing the ingress content
/// * `category` - Category to classify the ingressed content
/// * `files` - Vector of `FileInfo` objects containing information about uploaded files
/// * `user_id` - Identifier of the user performing the ingress operation
///
/// # Errors
///
/// Returns [`AppError::NotFound`] when no valid files or content are provided.
/// # Returns
/// * `Result<Vec<IngestionPayload>, AppError>` - On success, returns a vector of ingress objects
/// (one per file/content type). On failure, returns an `AppError`.
#[allow(clippy::similar_names)]
pub fn create_ingestion_payload(
content: Option<String>,
context: String,
category: String,
files: Vec<FileInfo>,
user_id: String,
user_id: &str,
) -> Result<Vec<IngestionPayload>, AppError> {
let parsed = Self::parse_content(content);
let content_follows = parsed.follows();
let file_count = files.len();
#[allow(clippy::arithmetic_side_effects)]
let capacity = file_count + usize::from(content_follows);
let mut object_list = Vec::with_capacity(capacity);
let mut fields = Some(IngestFields {
context,
category,
user_id,
});
// Initialize list
let mut object_list = Vec::new();
for (index, file) in files.into_iter().enumerate() {
let is_last_file = index.saturating_add(1) == file_count;
if content_follows || !is_last_file {
let Some(shared) = fields.as_ref() else {
return Err(AppError::internal("shared ingest fields consumed early"));
};
object_list.push(Self::File {
file_info: file,
context: shared.context.clone(),
category: shared.category.clone(),
user_id: shared.user_id.clone(),
});
} else {
let Some(shared) = fields.take() else {
return Err(AppError::internal("shared ingest fields missing for file"));
};
object_list.push(Self::File {
file_info: file,
context: shared.context,
category: shared.category,
user_id: shared.user_id,
// Create a IngestionPayload from content if it exists, checking for URL or text
if let Some(input_content) = content {
match Url::parse(&input_content) {
Ok(url) => {
info!("Detected URL: {}", url);
object_list.push(IngestionPayload::Url {
url: url.to_string(),
context: context.clone(),
category: category.clone(),
user_id: user_id.into(),
});
}
}
if let ParsedContent::Url(url) = parsed {
info!("Detected URL: {url}");
let Some(shared) = fields.take() else {
return Err(AppError::internal("shared ingest fields missing for url"));
};
object_list.push(Self::Url {
url,
context: shared.context,
category: shared.category,
user_id: shared.user_id,
});
} else if let ParsedContent::Text(text) = parsed {
Err(_) => {
if input_content.len() > 2 {
info!("Treating input as plain text");
let Some(shared) = fields.take() else {
return Err(AppError::internal("shared ingest fields missing for text"));
};
object_list.push(Self::Text {
text,
context: shared.context,
category: shared.category,
user_id: shared.user_id,
object_list.push(IngestionPayload::Text {
text: input_content.to_string(),
context: context.clone(),
category: category.clone(),
user_id: user_id.into(),
});
}
}
}
}
for file in files {
object_list.push(IngestionPayload::File {
file_info: file,
context: context.clone(),
category: category.clone(),
user_id: user_id.into(),
})
}
// If no objects are constructed, we return Err
if object_list.is_empty() {
return Err(AppError::NotFound(
"no valid content or files provided".into(),
"No valid content or files provided".into(),
));
}
Ok(object_list)
}
fn parse_content(content: Option<String>) -> ParsedContent {
let Some(input_content) = content else {
return ParsedContent::Skip;
};
if input_content.len() <= 2 {
return ParsedContent::Skip;
}
match Url::parse(&input_content) {
Ok(url) => ParsedContent::Url(url.to_string()),
Err(_) => ParsedContent::Text(input_content),
}
}
}
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use anyhow::{self, Context};
use chrono::Utc;
use super::*;
@@ -198,23 +131,24 @@ mod tests {
}
#[test]
fn test_create_ingestion_payload_with_url() -> anyhow::Result<()> {
fn test_create_ingestion_payload_with_url() {
let url = "https://example.com";
let context = "Process this URL";
let category = "websites";
let user_id = "user123";
let files = vec![];
let result = IngestionPayload::create_ingestion_payload(
Some(url.to_string()),
context.to_string(),
category.to_string(),
vec![],
user_id.to_string(),
files,
user_id,
)
.with_context(|| "create_ingestion_payload".to_string())?;
.unwrap();
assert_eq!(result.len(), 1);
match result.first().context("expected one result")? {
match &result[0] {
IngestionPayload::Url {
url: payload_url,
context: payload_context,
@@ -222,34 +156,34 @@ mod tests {
user_id: payload_user_id,
} => {
// URL parser may normalize the URL by adding a trailing slash
assert!(payload_url == &url.to_string() || payload_url == &format!("{url}/"));
assert!(payload_url == &url.to_string() || payload_url == &format!("{}/", url));
assert_eq!(payload_context, &context);
assert_eq!(payload_category, &category);
assert_eq!(payload_user_id, &user_id);
}
_ => anyhow::bail!("Expected Url variant"),
_ => panic!("Expected Url variant"),
}
Ok(())
}
#[test]
fn test_create_ingestion_payload_with_text() -> anyhow::Result<()> {
fn test_create_ingestion_payload_with_text() {
let text = "This is some text content";
let context = "Process this text";
let category = "notes";
let user_id = "user123";
let files = vec![];
let result = IngestionPayload::create_ingestion_payload(
Some(text.to_string()),
context.to_string(),
category.to_string(),
vec![],
user_id.to_string(),
files,
user_id,
)
.with_context(|| "create_ingestion_payload".to_string())?;
.unwrap();
assert_eq!(result.len(), 1);
match result.first().context("expected one result")? {
match &result[0] {
IngestionPayload::Text {
text: payload_text,
context: payload_context,
@@ -261,13 +195,12 @@ mod tests {
assert_eq!(payload_category, category);
assert_eq!(payload_user_id, user_id);
}
_ => anyhow::bail!("Expected Text variant"),
_ => panic!("Expected Text variant"),
}
Ok(())
}
#[test]
fn test_create_ingestion_payload_with_file() -> anyhow::Result<()> {
fn test_create_ingestion_payload_with_file() {
let context = "Process this file";
let category = "documents";
let user_id = "user123";
@@ -278,36 +211,36 @@ mod tests {
};
let file_info: FileInfo = mock_file.into();
let file_id = file_info.id.clone();
let files = vec![file_info.clone()];
let result = IngestionPayload::create_ingestion_payload(
None,
context.to_string(),
category.to_string(),
vec![file_info],
user_id.to_string(),
)?;
files,
user_id,
)
.unwrap();
assert_eq!(result.len(), 1);
match result.first().context("expected one result")? {
match &result[0] {
IngestionPayload::File {
file_info: payload_file_info,
context: payload_context,
category: payload_category,
user_id: payload_user_id,
} => {
assert_eq!(payload_file_info.id, file_id);
assert_eq!(payload_file_info.id, file_info.id);
assert_eq!(payload_context, context);
assert_eq!(payload_category, category);
assert_eq!(payload_user_id, user_id);
}
_ => anyhow::bail!("Expected File variant"),
_ => panic!("Expected File variant"),
}
Ok(())
}
#[test]
fn test_create_ingestion_payload_with_url_and_file() -> anyhow::Result<()> {
fn test_create_ingestion_payload_with_url_and_file() {
let url = "https://example.com";
let context = "Process this data";
let category = "mixed";
@@ -319,207 +252,88 @@ mod tests {
};
let file_info: FileInfo = mock_file.into();
let file_id = file_info.id.clone();
let files = vec![file_info.clone()];
let result = IngestionPayload::create_ingestion_payload(
Some(url.to_string()),
context.to_string(),
category.to_string(),
vec![file_info],
user_id.to_string(),
)?;
files,
user_id,
)
.unwrap();
assert_eq!(result.len(), 2);
// Check first item is File (files processed first to minimize clones)
match result.first().context("expected first item")? {
IngestionPayload::File {
file_info: payload_file_info,
..
} => {
assert_eq!(payload_file_info.id, file_id);
}
_ => anyhow::bail!("Expected first item to be File variant"),
}
// Check second item is URL
match result.get(1).context("expected second item")? {
// Check first item is URL
match &result[0] {
IngestionPayload::Url {
url: payload_url, ..
} => {
// URL parser may normalize the URL by adding a trailing slash
assert!(payload_url == &url.to_string() || payload_url == &format!("{url}/"));
assert!(payload_url == &url.to_string() || payload_url == &format!("{}/", url));
}
_ => anyhow::bail!("Expected second item to be Url variant"),
_ => panic!("Expected first item to be Url variant"),
}
// Check second item is File
match &result[1] {
IngestionPayload::File {
file_info: payload_file_info,
..
} => {
assert_eq!(payload_file_info.id, file_info.id);
}
_ => panic!("Expected second item to be File variant"),
}
Ok(())
}
#[test]
fn test_create_ingestion_payload_empty_input() -> anyhow::Result<()> {
fn test_create_ingestion_payload_empty_input() {
let context = "Process something";
let category = "empty";
let user_id = "user123";
let result = IngestionPayload::create_ingestion_payload(
None,
context.to_string(),
category.to_string(),
vec![],
user_id.to_string(),
);
assert!(result.is_err());
match result {
Err(AppError::NotFound(msg)) => {
assert_eq!(msg, "no valid content or files provided");
}
_ => anyhow::bail!("Expected NotFound error"),
}
Ok(())
}
#[test]
fn test_create_ingestion_payload_with_empty_text() -> anyhow::Result<()> {
let text = ""; // Empty text
let context = "Process this";
let category = "notes";
let user_id = "user123";
let result = IngestionPayload::create_ingestion_payload(
Some(text.to_string()),
context.to_string(),
category.to_string(),
vec![],
user_id.to_string(),
);
assert!(result.is_err());
match result {
Err(AppError::NotFound(msg)) => {
assert_eq!(msg, "no valid content or files provided");
}
_ => anyhow::bail!("Expected NotFound error"),
}
Ok(())
}
#[test]
fn test_create_ingestion_payload_with_file_and_text() -> anyhow::Result<()> {
let text = "plain notes";
let context = "ctx";
let category = "cat";
let user_id = "user123";
let file_info: FileInfo = MockFileInfo {
id: "file1".to_string(),
}
.into();
let result = IngestionPayload::create_ingestion_payload(
Some(text.to_string()),
context.to_string(),
category.to_string(),
vec![file_info],
user_id.to_string(),
)?;
assert_eq!(result.len(), 2);
let first = result.first().expect("expected first payload");
let second = result.get(1).expect("expected second payload");
match (first, second) {
(
IngestionPayload::File {
file_info: payload_file,
context: file_context,
..
},
IngestionPayload::Text {
text: payload_text,
context: text_context,
category: text_category,
user_id: text_user_id,
},
) => {
assert_eq!(payload_file.id, "file1");
assert_eq!(file_context, context);
assert_eq!(payload_text, text);
assert_eq!(text_context, context);
assert_eq!(text_category, category);
assert_eq!(text_user_id, user_id);
}
_ => anyhow::bail!("expected File then Text"),
}
Ok(())
}
#[test]
fn test_create_ingestion_payload_short_content_with_file_only_yields_file() -> anyhow::Result<()>
{
let context = "ctx";
let category = "cat";
let user_id = "user123";
let file_info: FileInfo = MockFileInfo {
id: "file1".to_string(),
}
.into();
let result = IngestionPayload::create_ingestion_payload(
Some("ab".to_string()),
context.to_string(),
category.to_string(),
vec![file_info],
user_id.to_string(),
)?;
assert_eq!(result.len(), 1);
match result.first().context("expected one file payload")? {
IngestionPayload::File {
file_info,
context: payload_context,
category: payload_category,
user_id: payload_user_id,
} => {
assert_eq!(file_info.id, "file1");
assert_eq!(payload_context, context);
assert_eq!(payload_category, category);
assert_eq!(payload_user_id, user_id);
}
_ => anyhow::bail!("expected File variant only"),
}
Ok(())
}
#[test]
fn test_create_ingestion_payload_two_files_without_content() -> anyhow::Result<()> {
let context = "ctx";
let category = "cat";
let user_id = "user123";
let files = vec![
MockFileInfo {
id: "file1".to_string(),
}
.into(),
MockFileInfo {
id: "file2".to_string(),
}
.into(),
];
let files = vec![];
let result = IngestionPayload::create_ingestion_payload(
None,
context.to_string(),
category.to_string(),
files,
user_id.to_string(),
)?;
user_id,
);
assert_eq!(result.len(), 2);
assert!(matches!(
result.first(),
Some(IngestionPayload::File { .. })
));
assert!(matches!(result.get(1), Some(IngestionPayload::File { .. })));
Ok(())
assert!(result.is_err());
match result {
Err(AppError::NotFound(msg)) => {
assert_eq!(msg, "No valid content or files provided");
}
_ => panic!("Expected NotFound error"),
}
}
#[test]
fn test_create_ingestion_payload_with_empty_text() {
let text = ""; // Empty text
let context = "Process this";
let category = "notes";
let user_id = "user123";
let files = vec![];
let result = IngestionPayload::create_ingestion_payload(
Some(text.to_string()),
context.to_string(),
category.to_string(),
files,
user_id,
);
assert!(result.is_err());
match result {
Err(AppError::NotFound(msg)) => {
assert_eq!(msg, "No valid content or files provided");
}
_ => panic!("Expected NotFound error"),
}
}
}
+100 -276
View File
@@ -1,7 +1,15 @@
use std::{sync::Arc, time::Duration};
#![allow(
clippy::cast_possible_wrap,
clippy::items_after_statements,
clippy::arithmetic_side_effects,
clippy::cast_sign_loss,
clippy::missing_docs_in_private_items,
clippy::trivially_copy_pass_by_ref,
clippy::expect_used
)]
use std::time::Duration;
use chrono::Duration as ChronoDuration;
use futures::future::try_join_all;
use state_machines::state_machine;
use surrealdb::sql::Datetime as SurrealDatetime;
use uuid::Uuid;
@@ -14,7 +22,7 @@ pub const MAX_ATTEMPTS: u32 = 3;
pub const DEFAULT_LEASE_SECS: i64 = 300;
pub const DEFAULT_PRIORITY: i32 = 0;
#[derive(Debug, Default, Clone, Copy, serde::Serialize, serde::Deserialize, PartialEq, Eq)]
#[derive(Debug, Default, Clone, serde::Serialize, serde::Deserialize, PartialEq, Eq)]
pub enum TaskState {
#[serde(rename = "Pending")]
#[default]
@@ -34,7 +42,6 @@ pub enum TaskState {
}
impl TaskState {
#[must_use]
pub fn as_str(&self) -> &'static str {
match self {
TaskState::Pending => "Pending",
@@ -47,7 +54,6 @@ impl TaskState {
}
}
#[must_use]
pub fn is_terminal(&self) -> bool {
matches!(
self,
@@ -55,7 +61,6 @@ impl TaskState {
)
}
#[must_use]
pub fn display_label(&self) -> &'static str {
match self {
TaskState::Pending => "Pending",
@@ -69,16 +74,12 @@ impl TaskState {
}
}
/// Information about an error that occurred during task processing.
#[derive(Debug, Clone, serde::Serialize, serde::Deserialize, PartialEq, Eq, Default)]
pub struct TaskErrorInfo {
/// Machine-readable error code (e.g., `"pipeline_error"`).
pub code: Option<String>,
/// Human-readable error description.
pub message: String,
}
/// Internal events that drive the task state machine transitions.
#[derive(Debug, Clone, Copy)]
enum TaskTransition {
StartProcessing,
@@ -90,7 +91,7 @@ enum TaskTransition {
}
impl TaskTransition {
fn as_str(self) -> &'static str {
fn as_str(&self) -> &'static str {
match self {
TaskTransition::StartProcessing => "start_processing",
TaskTransition::Succeed => "succeed",
@@ -140,20 +141,34 @@ mod lifecycle {
pub(super) fn pending() -> TaskLifecycleMachine<(), Pending> {
TaskLifecycleMachine::new(())
}
pub(super) fn reserved() -> TaskLifecycleMachine<(), Reserved> {
pending()
.reserve()
.expect("reserve transition from Pending should exist")
}
fn invalid_transition(state: TaskState, event: TaskTransition) -> AppError {
pub(super) fn processing() -> TaskLifecycleMachine<(), Processing> {
reserved()
.start_processing()
.expect("start_processing transition from Reserved should exist")
}
pub(super) fn failed() -> TaskLifecycleMachine<(), Failed> {
processing()
.fail()
.expect("fail transition from Processing should exist")
}
}
fn invalid_transition(state: &TaskState, event: TaskTransition) -> AppError {
AppError::Validation(format!(
"invalid task transition: {} -> {}",
"Invalid task transition: {} -> {}",
state.as_str(),
event.as_str()
))
}
fn worker_id_for_bind(worker_id: Option<&String>) -> String {
worker_id.cloned().unwrap_or_default()
}
stored_object!(IngestionTask, "ingestion_task", {
content: IngestionPayload,
state: TaskState,
@@ -182,7 +197,6 @@ stored_object!(IngestionTask, "ingestion_task", {
});
impl IngestionTask {
#[must_use]
pub fn new(content: IngestionPayload, user_id: String) -> Self {
let now = chrono::Utc::now();
@@ -206,85 +220,33 @@ impl IngestionTask {
}
}
#[must_use]
pub fn can_retry(&self) -> bool {
self.attempts < self.max_attempts
}
/// Moves the payload out of the task, leaving an empty placeholder behind.
///
/// The task's `content` is only needed while driving the pipeline; the
/// terminal `user_id`, `state`, and bookkeeping fields are stored separately,
/// so replacing it with the default placeholder avoids cloning large payloads.
#[must_use]
pub fn take_content(&mut self) -> IngestionPayload {
std::mem::take(&mut self.content)
}
#[must_use]
pub fn lease_duration(&self) -> Duration {
Duration::from_secs(u64::try_from(self.lease_duration_secs.max(0)).unwrap_or(0))
Duration::from_secs(self.lease_duration_secs.max(0) as u64)
}
/// Create a new task and immediately persist it to the database.
///
/// # Errors
///
/// Returns `AppError::Database` if the store operation fails.
/// Returns `AppError::internal` if the database returns no stored record.
pub async fn create_and_add_to_db(
content: IngestionPayload,
user_id: impl AsRef<str>,
user_id: String,
db: &SurrealDbClient,
) -> Result<IngestionTask, AppError> {
let task = Self::new(content, user_id.as_ref().to_string());
db.store_item(task)
.await?
.ok_or_else(|| AppError::internal("ingestion task store returned no record"))
let task = Self::new(content, user_id);
db.store_item(task.clone()).await?;
Ok(task)
}
/// Create and persist multiple tasks concurrently (one `CREATE` per payload).
///
/// Use this when ingest produces several payloads (files plus URL/text). For a
/// single payload, call [`Self::create_and_add_to_db`] instead.
///
/// # Errors
///
/// Returns the first [`AppError`] from any failed store, same as [`try_join_all`].
pub async fn create_all_and_add_to_db(
contents: Vec<IngestionPayload>,
user_id: impl AsRef<str>,
db: &SurrealDbClient,
) -> Result<Vec<IngestionTask>, AppError> {
if contents.is_empty() {
return Ok(Vec::new());
}
let user_id = Arc::new(user_id.as_ref().to_string());
let db = db.clone();
try_join_all(contents.into_iter().map(|content| {
let user_id = Arc::clone(&user_id);
let db = db.clone();
async move { Self::create_and_add_to_db(content, user_id.as_ref(), &db).await }
}))
.await
}
/// Claim the next ready task for processing.
///
/// Atomically reserves a task by transitioning it from a candidate state to `Reserved`.
/// Returns `Ok(None)` if no task is ready to claim.
///
/// # Errors
///
/// Returns `AppError::Database` if the update query fails.
pub async fn claim_next_ready(
db: &SurrealDbClient,
worker_id: &str,
now: chrono::DateTime<chrono::Utc>,
lease_duration: Duration,
) -> Result<Option<IngestionTask>, AppError> {
debug_assert!(lifecycle::pending().reserve().is_ok());
debug_assert!(lifecycle::failed().reserve().is_ok());
const CLAIM_QUERY: &str = r#"
UPDATE (
SELECT * FROM type::table($table)
@@ -314,11 +276,6 @@ impl IngestionTask {
RETURN *;
"#;
debug_assert!(lifecycle::pending().reserve().is_ok());
debug_assert!(lifecycle::pending().reserve().is_ok_and(|m| m
.start_processing()
.is_ok_and(|m| m.fail().is_ok_and(|m| m.reserve().is_ok()))));
let mut result = db
.client
.query(CLAIM_QUERY)
@@ -343,22 +300,13 @@ impl IngestionTask {
.bind(("reserved_state", TaskState::Reserved.as_str()))
.bind(("now", SurrealDatetime::from(now)))
.bind(("worker_id", worker_id.to_string()))
.bind((
"lease_secs",
i64::try_from(lease_duration.as_secs()).unwrap_or(i64::MAX),
))
.bind(("lease_secs", lease_duration.as_secs() as i64))
.await?;
let task: Option<IngestionTask> = result.take(0)?;
Ok(task)
}
/// Transition this task from `Reserved` to `Processing`.
///
/// # Errors
///
/// Returns `AppError::Validation` if the task is not in `Reserved` state
/// or belongs to a different worker. Returns `AppError::Database` on DB failure.
pub async fn mark_processing(&self, db: &SurrealDbClient) -> Result<IngestionTask, AppError> {
const START_PROCESSING_QUERY: &str = r#"
UPDATE type::thing($table, $id)
@@ -370,7 +318,6 @@ impl IngestionTask {
"#;
let now = chrono::Utc::now();
let worker_id = worker_id_for_bind(self.worker_id.as_ref());
let mut result = db
.client
.query(START_PROCESSING_QUERY)
@@ -379,19 +326,13 @@ impl IngestionTask {
.bind(("processing", TaskState::Processing.as_str()))
.bind(("reserved", TaskState::Reserved.as_str()))
.bind(("now", SurrealDatetime::from(now)))
.bind(("worker_id", worker_id))
.bind(("worker_id", self.worker_id.clone().unwrap_or_default()))
.await?;
let updated: Option<IngestionTask> = result.take(0)?;
updated.ok_or_else(|| invalid_transition(self.state, TaskTransition::StartProcessing))
updated.ok_or_else(|| invalid_transition(&self.state, TaskTransition::StartProcessing))
}
/// Transition this task from `Processing` to `Succeeded`.
///
/// # Errors
///
/// Returns `AppError::Validation` if the task is not in `Processing` state
/// or belongs to a different worker. Returns `AppError::Database` on DB failure.
pub async fn mark_succeeded(&self, db: &SurrealDbClient) -> Result<IngestionTask, AppError> {
const COMPLETE_QUERY: &str = r#"
UPDATE type::thing($table, $id)
@@ -408,7 +349,6 @@ impl IngestionTask {
"#;
let now = chrono::Utc::now();
let worker_id = worker_id_for_bind(self.worker_id.as_ref());
let mut result = db
.client
.query(COMPLETE_QUERY)
@@ -417,27 +357,23 @@ impl IngestionTask {
.bind(("succeeded", TaskState::Succeeded.as_str()))
.bind(("processing", TaskState::Processing.as_str()))
.bind(("now", SurrealDatetime::from(now)))
.bind(("worker_id", worker_id))
.bind(("worker_id", self.worker_id.clone().unwrap_or_default()))
.await?;
let updated: Option<IngestionTask> = result.take(0)?;
updated.ok_or_else(|| invalid_transition(self.state, TaskTransition::Succeed))
updated.ok_or_else(|| invalid_transition(&self.state, TaskTransition::Succeed))
}
/// Transition this task from `Processing` to `Failed`.
///
/// The task will be rescheduled for retry after `retry_delay`.
///
/// # Errors
///
/// Returns `AppError::Validation` if the task is not in `Processing` state
/// or belongs to a different worker. Returns `AppError::Database` on DB failure.
pub async fn mark_failed(
&self,
error: TaskErrorInfo,
retry_delay: Duration,
db: &SurrealDbClient,
) -> Result<IngestionTask, AppError> {
let now = chrono::Utc::now();
let retry_at = now
+ ChronoDuration::from_std(retry_delay).unwrap_or_else(|_| ChronoDuration::seconds(30));
const FAIL_QUERY: &str = r#"
UPDATE type::thing($table, $id)
SET state = $failed,
@@ -452,15 +388,6 @@ impl IngestionTask {
RETURN *;
"#;
let now = chrono::Utc::now();
let retry_at = now
.checked_add_signed(
ChronoDuration::from_std(retry_delay)
.unwrap_or_else(|_| ChronoDuration::seconds(30)),
)
.unwrap_or(now);
let worker_id = worker_id_for_bind(self.worker_id.as_ref());
let mut result = db
.client
.query(FAIL_QUERY)
@@ -472,19 +399,13 @@ impl IngestionTask {
.bind(("retry_at", SurrealDatetime::from(retry_at)))
.bind(("error_code", error.code.clone()))
.bind(("error_message", error.message.clone()))
.bind(("worker_id", worker_id))
.bind(("worker_id", self.worker_id.clone().unwrap_or_default()))
.await?;
let updated: Option<IngestionTask> = result.take(0)?;
updated.ok_or_else(|| invalid_transition(self.state, TaskTransition::Fail))
updated.ok_or_else(|| invalid_transition(&self.state, TaskTransition::Fail))
}
/// Transition this task from `Failed` to `DeadLetter`.
///
/// # Errors
///
/// Returns `AppError::Validation` if the task is not in `Failed` state.
/// Returns `AppError::Database` on DB failure.
pub async fn mark_dead_letter(
&self,
error: TaskErrorInfo,
@@ -518,15 +439,9 @@ impl IngestionTask {
.await?;
let updated: Option<IngestionTask> = result.take(0)?;
updated.ok_or_else(|| invalid_transition(self.state, TaskTransition::DeadLetter))
updated.ok_or_else(|| invalid_transition(&self.state, TaskTransition::DeadLetter))
}
/// Transition this task to `Cancelled` from any non-terminal state.
///
/// # Errors
///
/// Returns `AppError::Validation` if the task is in a terminal state.
/// Returns `AppError::Database` on DB failure.
pub async fn mark_cancelled(&self, db: &SurrealDbClient) -> Result<IngestionTask, AppError> {
const CANCEL_QUERY: &str = r#"
UPDATE type::thing($table, $id)
@@ -557,15 +472,9 @@ impl IngestionTask {
.await?;
let updated: Option<IngestionTask> = result.take(0)?;
updated.ok_or_else(|| invalid_transition(self.state, TaskTransition::Cancel))
updated.ok_or_else(|| invalid_transition(&self.state, TaskTransition::Cancel))
}
/// Release a reserved task back to `Pending` state.
///
/// # Errors
///
/// Returns `AppError::Validation` if the task is not in `Reserved` state.
/// Returns `AppError::Database` on DB failure.
pub async fn release(&self, db: &SurrealDbClient) -> Result<IngestionTask, AppError> {
const RELEASE_QUERY: &str = r#"
UPDATE type::thing($table, $id)
@@ -589,14 +498,9 @@ impl IngestionTask {
.await?;
let updated: Option<IngestionTask> = result.take(0)?;
updated.ok_or_else(|| invalid_transition(self.state, TaskTransition::Release))
updated.ok_or_else(|| invalid_transition(&self.state, TaskTransition::Release))
}
/// Retrieve all non-terminal tasks across active states.
///
/// # Errors
///
/// Returns `AppError::Database` if the query fails.
pub async fn get_unfinished_tasks(
db: &SurrealDbClient,
) -> Result<Vec<IngestionTask>, AppError> {
@@ -625,9 +529,6 @@ impl IngestionTask {
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use anyhow::{self, Context};
use super::*;
use crate::storage::types::ingestion_payload::IngestionPayload;
@@ -640,16 +541,16 @@ mod tests {
}
}
async fn memory_db() -> anyhow::Result<SurrealDbClient> {
async fn memory_db() -> SurrealDbClient {
let namespace = "test_ns";
let database = Uuid::new_v4().to_string();
SurrealDbClient::memory(namespace, &database)
.await
.with_context(|| "in-memory surrealdb".to_string())
.expect("in-memory surrealdb")
}
#[tokio::test]
async fn test_new_task_defaults() -> anyhow::Result<()> {
async fn test_new_task_defaults() {
let user_id = "user123";
let payload = create_payload(user_id);
let task = IngestionTask::new(payload.clone(), user_id.to_string());
@@ -661,140 +562,73 @@ mod tests {
assert_eq!(task.max_attempts, MAX_ATTEMPTS);
assert!(task.locked_at.is_none());
assert!(task.worker_id.is_none());
Ok(())
}
#[test]
fn test_take_content_moves_payload_and_leaves_default() {
let user_id = "user123";
let payload = create_payload(user_id);
let mut task = IngestionTask::new(payload.clone(), user_id.to_string());
let taken = task.take_content();
assert_eq!(taken, payload);
assert_eq!(task.content, IngestionPayload::default());
}
#[tokio::test]
async fn test_create_all_and_add_to_db_empty() -> anyhow::Result<()> {
let db = memory_db().await?;
let tasks = IngestionTask::create_all_and_add_to_db(vec![], "user123", &db).await?;
assert!(tasks.is_empty());
Ok(())
}
#[tokio::test]
async fn test_create_all_and_add_to_db_stores_multiple() -> anyhow::Result<()> {
let db = memory_db().await?;
let user_id = "user123";
let payloads = vec![
create_payload(user_id),
IngestionPayload::Text {
text: "second payload".to_string(),
context: "ctx".to_string(),
category: "cat".to_string(),
user_id: user_id.to_string(),
},
];
let created = IngestionTask::create_all_and_add_to_db(payloads, user_id, &db).await?;
assert_eq!(created.len(), 2);
let first = created.first().expect("expected first task");
let second = created.get(1).expect("expected second task");
assert_ne!(first.id, second.id);
for task in &created {
let stored: Option<IngestionTask> = db.get_item::<IngestionTask>(&task.id).await?;
let stored = stored.with_context(|| format!("task {} should exist", task.id))?;
assert_eq!(stored.id, task.id);
assert_eq!(stored.state, TaskState::Pending);
assert_eq!(stored.user_id, user_id);
}
Ok(())
}
#[tokio::test]
async fn test_create_and_store_task() -> anyhow::Result<()> {
let db = memory_db().await?;
async fn test_create_and_store_task() {
let db = memory_db().await;
let user_id = "user123";
let payload = create_payload(user_id);
let created =
IngestionTask::create_and_add_to_db(payload.clone(), user_id.to_string(), &db)
.await
.with_context(|| "store".to_string())?;
.expect("store");
let stored: Option<IngestionTask> = db
.get_item::<IngestionTask>(&created.id)
.await
.with_context(|| "fetch".to_string())?;
.expect("fetch");
let stored = stored.with_context(|| "task exists".to_string())?;
let stored = stored.expect("task exists");
assert_eq!(stored.id, created.id);
assert_eq!(stored.state, TaskState::Pending);
assert_eq!(stored.attempts, 0);
Ok(())
}
#[tokio::test]
async fn test_claim_and_transition() -> anyhow::Result<()> {
let db = memory_db().await?;
async fn test_claim_and_transition() {
let db = memory_db().await;
let user_id = "user123";
let payload = create_payload(user_id);
let task = IngestionTask::new(payload, user_id.to_string());
db.store_item(task.clone())
.await
.with_context(|| "store".to_string())?;
db.store_item(task.clone()).await.expect("store");
let worker_id = "worker-1";
let now = chrono::Utc::now();
let claimed = IngestionTask::claim_next_ready(&db, worker_id, now, Duration::from_secs(60))
.await
.with_context(|| "claim".to_string())?
.with_context(|| "task claimed".to_string())?;
.expect("claim");
let claimed = claimed.expect("task claimed");
assert_eq!(claimed.state, TaskState::Reserved);
assert_eq!(claimed.worker_id.as_deref(), Some(worker_id));
let processing = claimed
.mark_processing(&db)
.await
.with_context(|| "processing".to_string())?;
let processing = claimed.mark_processing(&db).await.expect("processing");
assert_eq!(processing.state, TaskState::Processing);
let succeeded = processing
.mark_succeeded(&db)
.await
.with_context(|| "succeeded".to_string())?;
let succeeded = processing.mark_succeeded(&db).await.expect("succeeded");
assert_eq!(succeeded.state, TaskState::Succeeded);
assert!(succeeded.worker_id.is_none());
assert!(succeeded.locked_at.is_none());
Ok(())
}
#[tokio::test]
async fn test_fail_and_dead_letter() -> anyhow::Result<()> {
let db = memory_db().await?;
async fn test_fail_and_dead_letter() {
let db = memory_db().await;
let user_id = "user123";
let payload = create_payload(user_id);
let task = IngestionTask::new(payload, user_id.to_string());
db.store_item(task.clone())
.await
.with_context(|| "store".to_string())?;
db.store_item(task.clone()).await.expect("store");
let worker_id = "worker-dead";
let now = chrono::Utc::now();
let claimed = IngestionTask::claim_next_ready(&db, worker_id, now, Duration::from_secs(60))
.await
.with_context(|| "claim".to_string())?
.with_context(|| "claimed".to_string())?;
.expect("claim")
.expect("claimed");
let processing = claimed
.mark_processing(&db)
.await
.with_context(|| "processing".to_string())?;
let processing = claimed.mark_processing(&db).await.expect("processing");
let error_info = TaskErrorInfo {
code: Some("pipeline_error".into()),
@@ -804,7 +638,7 @@ mod tests {
let failed = processing
.mark_failed(error_info.clone(), Duration::from_secs(30), &db)
.await
.with_context(|| "failed update".to_string())?;
.expect("failed update");
assert_eq!(failed.state, TaskState::Failed);
assert_eq!(failed.error_message.as_deref(), Some("failed"));
assert!(failed.worker_id.is_none());
@@ -814,26 +648,24 @@ mod tests {
let dead = failed
.mark_dead_letter(error_info.clone(), &db)
.await
.with_context(|| "dead letter".to_string())?;
.expect("dead letter");
assert_eq!(dead.state, TaskState::DeadLetter);
assert_eq!(dead.error_message.as_deref(), Some("failed"));
Ok(())
}
#[tokio::test]
async fn test_mark_processing_requires_reservation() -> anyhow::Result<()> {
let db = memory_db().await?;
async fn test_mark_processing_requires_reservation() {
let db = memory_db().await;
let user_id = "user123";
let payload = create_payload(user_id);
let task = IngestionTask::new(payload.clone(), user_id.to_string());
db.store_item(task.clone())
.await
.with_context(|| "store".to_string())?;
db.store_item(task.clone()).await.expect("store");
let Err(err) = task.mark_processing(&db).await else {
anyhow::bail!("processing should fail without reservation")
};
let err = task
.mark_processing(&db)
.await
.expect_err("processing should fail without reservation");
match err {
AppError::Validation(message) => {
@@ -842,23 +674,20 @@ mod tests {
"unexpected message: {message}"
);
}
other => anyhow::bail!("expected validation error, got {other:?}"),
other => panic!("expected validation error, got {other:?}"),
}
Ok(())
}
#[tokio::test]
async fn test_mark_failed_requires_processing() -> anyhow::Result<()> {
let db = memory_db().await?;
async fn test_mark_failed_requires_processing() {
let db = memory_db().await;
let user_id = "user123";
let payload = create_payload(user_id);
let task = IngestionTask::new(payload.clone(), user_id.to_string());
db.store_item(task.clone())
.await
.with_context(|| "store".to_string())?;
db.store_item(task.clone()).await.expect("store");
let Err(err) = task
let err = task
.mark_failed(
TaskErrorInfo {
code: None,
@@ -868,9 +697,7 @@ mod tests {
&db,
)
.await
else {
anyhow::bail!("failing should require processing state")
};
.expect_err("failing should require processing state");
match err {
AppError::Validation(message) => {
@@ -879,25 +706,23 @@ mod tests {
"unexpected message: {message}"
);
}
other => anyhow::bail!("expected validation error, got {other:?}"),
other => panic!("expected validation error, got {other:?}"),
}
Ok(())
}
#[tokio::test]
async fn test_release_requires_reservation() -> anyhow::Result<()> {
let db = memory_db().await?;
async fn test_release_requires_reservation() {
let db = memory_db().await;
let user_id = "user123";
let payload = create_payload(user_id);
let task = IngestionTask::new(payload.clone(), user_id.to_string());
db.store_item(task.clone())
.await
.with_context(|| "store".to_string())?;
db.store_item(task.clone()).await.expect("store");
let Err(err) = task.release(&db).await else {
anyhow::bail!("release should require reserved state")
};
let err = task
.release(&db)
.await
.expect_err("release should require reserved state");
match err {
AppError::Validation(message) => {
@@ -906,8 +731,7 @@ mod tests {
"unexpected message: {message}"
);
}
other => anyhow::bail!("expected validation error, got {other:?}"),
}
Ok(())
other => panic!("expected validation error, got {other:?}"),
}
}
}
File diff suppressed because it is too large Load Diff
@@ -2,17 +2,11 @@ use std::collections::HashMap;
use surrealdb::RecordId;
use crate::{
error::AppError,
storage::{db::SurrealDbClient, indexes::hnsw_index_redefine_transaction_sql},
stored_object,
};
use crate::{error::AppError, storage::db::SurrealDbClient, stored_object};
stored_object!(KnowledgeEntityEmbedding, "knowledge_entity_embedding", {
entity_id: RecordId,
embedding: Vec<f32>,
/// Denormalized source id for bulk deletes
source_id: String,
/// Denormalized user id for query scoping
user_id: String
});
@@ -23,43 +17,29 @@ impl KnowledgeEntityEmbedding {
db: &SurrealDbClient,
dimension: usize,
) -> Result<(), AppError> {
let query = hnsw_index_redefine_transaction_sql(
"idx_embedding_knowledge_entity_embedding",
Self::table_name(),
dimension,
let query = format!(
"BEGIN TRANSACTION;
REMOVE INDEX IF EXISTS idx_embedding_knowledge_entity_embedding ON TABLE {table};
DEFINE INDEX idx_embedding_knowledge_entity_embedding ON TABLE {table} FIELDS embedding HNSW DIMENSION {dimension};
COMMIT TRANSACTION;",
table = Self::table_name(),
);
let res = db.client.query(query).await.map_err(AppError::from)?;
res.check().map_err(AppError::from)?;
let res = db.client.query(query).await.map_err(AppError::Database)?;
res.check().map_err(AppError::Database)?;
Ok(())
}
/// Validates that an embedding vector matches the configured HNSW dimension.
#[allow(clippy::result_large_err)]
pub fn validate_dimension(embedding: &[f32], expected: usize) -> Result<(), AppError> {
if embedding.len() != expected {
return Err(AppError::Validation(format!(
"embedding dimension mismatch: got {}, expected {expected}",
embedding.len()
)));
}
Ok(())
}
/// Create a new knowledge entity embedding.
///
/// The embedding record id equals `entity_id` so each entity has at most one embedding row.
#[must_use]
pub fn new(entity_id: &str, source_id: String, embedding: Vec<f32>, user_id: String) -> Self {
/// Create a new knowledge entity embedding
pub fn new(entity_id: &str, embedding: Vec<f32>, user_id: String) -> Self {
let now = Utc::now();
Self {
id: entity_id.to_owned(),
id: uuid::Uuid::new_v4().to_string(),
created_at: now,
updated_at: now,
entity_id: RecordId::from_table_key("knowledge_entity", entity_id),
embedding,
source_id,
user_id,
}
}
@@ -78,8 +58,8 @@ impl KnowledgeEntityEmbedding {
.query(query)
.bind(("entity_id", entity_id.clone()))
.await
.map_err(AppError::from)?;
let embeddings: Vec<Self> = result.take(0).map_err(AppError::from)?;
.map_err(AppError::Database)?;
let embeddings: Vec<Self> = result.take(0).map_err(AppError::Database)?;
Ok(embeddings.into_iter().next())
}
@@ -92,6 +72,8 @@ impl KnowledgeEntityEmbedding {
return Ok(HashMap::new());
}
let ids_list: Vec<RecordId> = entity_ids.to_vec();
let query = format!(
"SELECT * FROM {} WHERE entity_id INSIDE $entity_ids",
Self::table_name()
@@ -99,10 +81,10 @@ impl KnowledgeEntityEmbedding {
let mut result = db
.client
.query(query)
.bind(("entity_ids", entity_ids.to_vec()))
.bind(("entity_ids", ids_list))
.await
.map_err(AppError::from)?;
let embeddings: Vec<Self> = result.take(0).map_err(AppError::from)?;
.map_err(AppError::Database)?;
let embeddings: Vec<Self> = result.take(0).map_err(AppError::Database)?;
Ok(embeddings
.into_iter()
@@ -123,41 +105,59 @@ impl KnowledgeEntityEmbedding {
.query(query)
.bind(("entity_id", entity_id.clone()))
.await
.map_err(AppError::from)?
.check()
.map_err(AppError::from)?;
.map_err(AppError::Database)?;
Ok(())
}
/// Delete all embeddings with the given denormalized `source_id`.
/// Delete embeddings by source_id (via joining to knowledge_entity table)
#[allow(clippy::items_after_statements)]
pub async fn delete_by_source_id(
source_id: &str,
db: &SurrealDbClient,
) -> Result<(), AppError> {
let query = format!(
"DELETE FROM {} WHERE source_id = $source_id",
Self::table_name()
);
db.client
let query = "SELECT id FROM knowledge_entity WHERE source_id = $source_id";
let mut res = db
.client
.query(query)
.bind(("source_id", source_id.to_owned()))
.await
.map_err(AppError::from)?
.check()
.map_err(AppError::from)?;
.map_err(AppError::Database)?;
#[allow(clippy::missing_docs_in_private_items)]
#[derive(Deserialize)]
struct IdRow {
id: RecordId,
}
let ids: Vec<IdRow> = res.take(0).map_err(AppError::Database)?;
for row in ids {
Self::delete_by_entity_id(&row.id, db).await?;
}
Ok(())
}
}
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use super::*;
use crate::storage::db::SurrealDbClient;
use crate::storage::types::knowledge_entity::{KnowledgeEntity, KnowledgeEntityType};
use crate::test_utils::{prepare_knowledge_entity_test_db, setup_test_db};
use anyhow::{self, Context};
use chrono::Utc;
use surrealdb::Value as SurrealValue;
use uuid::Uuid;
async fn setup_test_db() -> SurrealDbClient {
let namespace = "test_ns";
let database = Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, &database)
.await
.expect("Failed to start in-memory surrealdb");
db.apply_migrations()
.await
.expect("Failed to apply migrations");
db
}
fn build_knowledge_entity_with_id(
key: &str,
@@ -177,27 +177,12 @@ mod tests {
}
}
#[test]
fn new_uses_entity_id_as_record_id() {
let emb = KnowledgeEntityEmbedding::new(
"entity-abc",
"source-1".to_owned(),
vec![0.1, 0.2],
"user-1".to_owned(),
);
assert_eq!(emb.id, "entity-abc");
}
#[test]
fn validate_dimension_rejects_mismatch() {
let err = KnowledgeEntityEmbedding::validate_dimension(&[0.1, 0.2, 0.3], 2)
.expect_err("expected dimension mismatch");
assert!(matches!(err, AppError::Validation(_)));
}
#[tokio::test]
async fn test_create_and_get_by_entity_id() -> anyhow::Result<()> {
let db = prepare_knowledge_entity_test_db(3).await?;
async fn test_create_and_get_by_entity_id() {
let db = setup_test_db().await;
KnowledgeEntityEmbedding::redefine_hnsw_index(&db, 3)
.await
.expect("set test index dimension");
let user_id = "user_ke";
let entity_key = "entity-1";
let source_id = "source-ke";
@@ -207,27 +192,26 @@ mod tests {
KnowledgeEntity::store_with_embedding(entity.clone(), embedding_vec.clone(), &db)
.await
.with_context(|| "Failed to store entity with embedding".to_string())?;
.expect("Failed to store entity with embedding");
let entity_rid = RecordId::from_table_key(KnowledgeEntity::table_name(), &entity.id);
let fetched = KnowledgeEntityEmbedding::get_by_entity_id(&entity_rid, &db)
.await
.with_context(|| "Failed to get embedding by entity_id".to_string())?
.ok_or_else(|| anyhow::anyhow!("Expected embedding to exist"))?;
.expect("Failed to get embedding by entity_id")
.expect("Expected embedding to exist");
assert_eq!(fetched.id, entity_key);
assert_eq!(fetched.user_id, user_id);
assert_eq!(fetched.source_id, source_id);
assert_eq!(fetched.entity_id, entity_rid);
assert_eq!(fetched.embedding, embedding_vec);
Ok(())
}
#[tokio::test]
async fn test_delete_by_entity_id() -> anyhow::Result<()> {
let db = prepare_knowledge_entity_test_db(3).await?;
async fn test_delete_by_entity_id() {
let db = setup_test_db().await;
KnowledgeEntityEmbedding::redefine_hnsw_index(&db, 3)
.await
.expect("set test index dimension");
let user_id = "user_ke";
let entity_key = "entity-delete";
let source_id = "source-del";
@@ -236,75 +220,61 @@ mod tests {
KnowledgeEntity::store_with_embedding(entity.clone(), vec![0.5_f32, 0.6, 0.7], &db)
.await
.with_context(|| "Failed to store entity with embedding".to_string())?;
.expect("Failed to store entity with embedding");
let entity_rid = RecordId::from_table_key(KnowledgeEntity::table_name(), &entity.id);
let existing = KnowledgeEntityEmbedding::get_by_entity_id(&entity_rid, &db)
.await
.with_context(|| "Failed to get embedding before delete".to_string())?;
.expect("Failed to get embedding before delete");
assert!(existing.is_some());
KnowledgeEntityEmbedding::delete_by_entity_id(&entity_rid, &db)
.await
.with_context(|| "Failed to delete by entity_id".to_string())?;
.expect("Failed to delete by entity_id");
let after = KnowledgeEntityEmbedding::get_by_entity_id(&entity_rid, &db)
.await
.with_context(|| "Failed to get embedding after delete".to_string())?;
.expect("Failed to get embedding after delete");
assert!(after.is_none());
Ok(())
}
#[tokio::test]
async fn test_store_with_embedding_creates_entity_and_embedding() -> anyhow::Result<()> {
let db = prepare_knowledge_entity_test_db(3).await?;
async fn test_store_with_embedding_creates_entity_and_embedding() {
let db = setup_test_db().await;
let user_id = "user_store";
let source_id = "source_store";
let embedding = vec![0.2_f32, 0.3, 0.4];
KnowledgeEntityEmbedding::redefine_hnsw_index(&db, embedding.len())
.await
.expect("set test index dimension");
let entity = build_knowledge_entity_with_id("entity-store", source_id, user_id);
KnowledgeEntity::store_with_embedding(entity.clone(), embedding.clone(), &db)
.await
.with_context(|| "Failed to store entity with embedding".to_string())?;
.expect("Failed to store entity with embedding");
let stored_entity: Option<KnowledgeEntity> = db
.get_item(&entity.id)
.await
.with_context(|| "Failed to get entity".to_string())?;
let stored_entity: Option<KnowledgeEntity> = db.get_item(&entity.id).await.unwrap();
assert!(stored_entity.is_some());
let entity_rid = RecordId::from_table_key(KnowledgeEntity::table_name(), &entity.id);
let stored_embedding = KnowledgeEntityEmbedding::get_by_entity_id(&entity_rid, &db)
.await
.with_context(|| "Failed to fetch embedding".to_string())?;
let stored_embedding =
stored_embedding.ok_or_else(|| anyhow::anyhow!("Expected embedding to exist"))?;
assert_eq!(stored_embedding.id, entity.id);
.expect("Failed to fetch embedding");
assert!(stored_embedding.is_some());
let stored_embedding = stored_embedding.unwrap();
assert_eq!(stored_embedding.user_id, user_id);
assert_eq!(stored_embedding.source_id, source_id);
assert_eq!(stored_embedding.entity_id, entity_rid);
Ok(())
}
#[tokio::test]
async fn test_store_with_embedding_rejects_wrong_dimension() -> anyhow::Result<()> {
let db = prepare_knowledge_entity_test_db(3).await?;
let entity = build_knowledge_entity_with_id("entity-dim", "source-dim", "user-dim");
let result = KnowledgeEntity::store_with_embedding(entity, vec![0.1, 0.2], &db).await;
assert!(matches!(result, Err(AppError::Validation(_))));
Ok(())
}
#[tokio::test]
async fn test_delete_by_source_id() -> anyhow::Result<()> {
let db = prepare_knowledge_entity_test_db(3).await?;
async fn test_delete_by_source_id() {
let db = setup_test_db().await;
KnowledgeEntityEmbedding::redefine_hnsw_index(&db, 3)
.await
.expect("set test index dimension");
let user_id = "user_ke";
let source_id = "shared-ke";
let other_source = "other-ke";
@@ -315,13 +285,13 @@ mod tests {
KnowledgeEntity::store_with_embedding(entity1.clone(), vec![1.0_f32, 1.1, 1.2], &db)
.await
.with_context(|| "Failed to store entity with embedding".to_string())?;
.expect("Failed to store entity with embedding");
KnowledgeEntity::store_with_embedding(entity2.clone(), vec![2.0_f32, 2.1, 2.2], &db)
.await
.with_context(|| "Failed to store entity with embedding".to_string())?;
.expect("Failed to store entity with embedding");
KnowledgeEntity::store_with_embedding(entity_other.clone(), vec![3.0_f32, 3.1, 3.2], &db)
.await
.with_context(|| "Failed to store entity with embedding".to_string())?;
.expect("Failed to store entity with embedding");
let entity1_rid = RecordId::from_table_key(KnowledgeEntity::table_name(), &entity1.id);
let entity2_rid = RecordId::from_table_key(KnowledgeEntity::table_name(), &entity2.id);
@@ -329,75 +299,59 @@ mod tests {
KnowledgeEntityEmbedding::delete_by_source_id(source_id, &db)
.await
.with_context(|| "Failed to delete by source_id".to_string())?;
.expect("Failed to delete by source_id");
assert!(
KnowledgeEntityEmbedding::get_by_entity_id(&entity1_rid, &db)
.await
.with_context(|| "get entity1 embedding after delete".to_string())?
.unwrap()
.is_none()
);
assert!(
KnowledgeEntityEmbedding::get_by_entity_id(&entity2_rid, &db)
.await
.with_context(|| "get entity2 embedding after delete".to_string())?
.unwrap()
.is_none()
);
assert!(KnowledgeEntityEmbedding::get_by_entity_id(&other_rid, &db)
.await
.with_context(|| "get other embedding after delete".to_string())?
.unwrap()
.is_some());
Ok(())
}
#[tokio::test]
async fn test_redefine_hnsw_index_updates_dimension() -> anyhow::Result<()> {
let db = setup_test_db().await?;
async fn test_redefine_hnsw_index_updates_dimension() {
let db = setup_test_db().await;
KnowledgeEntityEmbedding::redefine_hnsw_index(&db, 16)
.await
.with_context(|| "failed to redefine index".to_string())?;
.expect("failed to redefine index");
let mut info_res = db
.client
.query("INFO FOR TABLE knowledge_entity_embedding;")
.await
.with_context(|| "info query failed".to_string())?;
let info: SurrealValue = info_res
.take(0)
.with_context(|| "failed to take info result".to_string())?;
let info_json: serde_json::Value = serde_json::to_value(info)
.with_context(|| "failed to convert info to json".to_string())?;
let idx_sql = info_json
.get("Object")
.and_then(|v| v.get("indexes"))
.and_then(|v| v.get("Object"))
.and_then(|v| v.get("idx_embedding_knowledge_entity_embedding"))
.and_then(|v| v.get("Strand"))
.and_then(|v| v.as_str())
.expect("info query failed");
let info: SurrealValue = info_res.take(0).expect("failed to take info result");
let info_json: serde_json::Value =
serde_json::to_value(info).expect("failed to convert info to json");
let idx_sql = info_json["Object"]["indexes"]["Object"]
["idx_embedding_knowledge_entity_embedding"]["Strand"]
.as_str()
.unwrap_or_default();
assert!(
idx_sql.contains("DIMENSION 16"),
"expected index definition to contain new dimension, got: {idx_sql}"
);
assert!(
idx_sql.contains("DIST COSINE"),
"expected index definition to use cosine distance, got: {idx_sql}"
);
Ok(())
}
#[tokio::test]
async fn test_fetch_entity_via_record_id() -> anyhow::Result<()> {
#[derive(Deserialize)]
struct Row {
entity_id: KnowledgeEntity,
}
let db = prepare_knowledge_entity_test_db(3).await?;
async fn test_fetch_entity_via_record_id() {
let db = setup_test_db().await;
KnowledgeEntityEmbedding::redefine_hnsw_index(&db, 3)
.await
.expect("set test index dimension");
let user_id = "user_ke";
let entity_key = "entity-fetch";
let source_id = "source-fetch";
@@ -405,10 +359,15 @@ mod tests {
let entity = build_knowledge_entity_with_id(entity_key, source_id, user_id);
KnowledgeEntity::store_with_embedding(entity.clone(), vec![0.7_f32, 0.8, 0.9], &db)
.await
.with_context(|| "Failed to store entity with embedding".to_string())?;
.expect("Failed to store entity with embedding");
let entity_rid = RecordId::from_table_key(KnowledgeEntity::table_name(), &entity.id);
#[derive(Deserialize)]
struct Row {
entity_id: KnowledgeEntity,
}
let mut res = db
.client
.query(
@@ -416,63 +375,13 @@ mod tests {
)
.bind(("id", entity_rid.clone()))
.await
.with_context(|| "failed to fetch embedding with FETCH".to_string())?;
let rows: Vec<Row> = res
.take(0)
.with_context(|| "failed to deserialize fetch rows".to_string())?;
.expect("failed to fetch embedding with FETCH");
let rows: Vec<Row> = res.take(0).expect("failed to deserialize fetch rows");
assert_eq!(rows.len(), 1);
let fetched_entity = &rows
.first()
.context("Expected at least one result")?
.entity_id;
let fetched_entity = &rows[0].entity_id;
assert_eq!(fetched_entity.id, entity_key);
assert_eq!(fetched_entity.name, "Test entity");
assert_eq!(fetched_entity.user_id, user_id);
Ok(())
}
#[tokio::test]
async fn test_upsert_replaces_existing_embedding_row() -> anyhow::Result<()> {
let db = prepare_knowledge_entity_test_db(3).await?;
let user_id = "user-upsert";
let source_id = "source-upsert";
let entity = build_knowledge_entity_with_id("entity-upsert", source_id, user_id);
KnowledgeEntity::store_with_embedding(entity.clone(), vec![1.0_f32, 0.0, 0.0], &db)
.await
.with_context(|| "initial store".to_string())?;
let replacement = KnowledgeEntityEmbedding::new(
&entity.id,
source_id.to_owned(),
vec![0.0, 1.0, 0.0],
user_id.to_owned(),
);
db.upsert_item(replacement)
.await
.with_context(|| "upsert replacement embedding".to_string())?;
let entity_rid = RecordId::from_table_key(KnowledgeEntity::table_name(), &entity.id);
let rows: Vec<KnowledgeEntityEmbedding> = db
.client
.query(format!(
"SELECT * FROM {} WHERE entity_id = $entity_id",
KnowledgeEntityEmbedding::table_name()
))
.bind(("entity_id", entity_rid))
.await
.with_context(|| "count embeddings".to_string())?
.take(0)
.with_context(|| "take embeddings".to_string())?;
assert_eq!(rows.len(), 1);
let row = rows.first().expect("expected one embedding row");
assert_eq!(row.id, entity.id);
assert_eq!(row.embedding, vec![0.0, 1.0, 0.0]);
Ok(())
}
}
+186 -380
View File
@@ -1,5 +1,4 @@
use crate::storage::types::user::User;
use crate::utils::serde_helpers::deserialize_flexible_id;
use crate::storage::types::file_info::deserialize_flexible_id;
use crate::{error::AppError, storage::db::SurrealDbClient};
use serde::{Deserialize, Serialize};
use uuid::Uuid;
@@ -22,7 +21,6 @@ pub struct KnowledgeRelationship {
}
impl KnowledgeRelationship {
#[must_use]
pub fn new(
in_: String,
out: String,
@@ -41,69 +39,36 @@ impl KnowledgeRelationship {
},
}
}
pub async fn store_relationship(&self, db_client: &SurrealDbClient) -> Result<(), AppError> {
let query = format!(
r#"DELETE relates_to:`{rel_id}`;
RELATE knowledge_entity:`{in_id}`->relates_to:`{rel_id}`->knowledge_entity:`{out_id}`
SET
metadata.user_id = '{user_id}',
metadata.source_id = '{source_id}',
metadata.relationship_type = '{relationship_type}'"#,
rel_id = self.id,
in_id = self.in_,
out_id = self.out,
user_id = self.metadata.user_id.as_str(),
source_id = self.metadata.source_id.as_str(),
relationship_type = self.metadata.relationship_type.as_str()
);
pub async fn store_relationship(self, db_client: &SurrealDbClient) -> Result<(), AppError> {
User::get_and_validate_knowledge_entity(&self.in_, &self.metadata.user_id, db_client)
.await?;
User::get_and_validate_knowledge_entity(&self.out, &self.metadata.user_id, db_client)
.await?;
let Self {
id,
in_,
out,
metadata:
RelationshipMetadata {
user_id,
source_id,
relationship_type,
},
} = self;
db_client
.client
.query(
r#"BEGIN TRANSACTION;
LET $in_entity = type::thing('knowledge_entity', $in_id);
LET $out_entity = type::thing('knowledge_entity', $out_id);
LET $relation = type::thing('relates_to', $rel_id);
DELETE type::thing('relates_to', $rel_id);
RELATE $in_entity->$relation->$out_entity SET
metadata.user_id = $user_id,
metadata.source_id = $source_id,
metadata.relationship_type = $relationship_type;
COMMIT TRANSACTION;"#,
)
.bind(("rel_id", id))
.bind(("in_id", in_))
.bind(("out_id", out))
.bind(("user_id", user_id))
.bind(("source_id", source_id))
.bind(("relationship_type", relationship_type))
.await
.map_err(AppError::from)?
.check()
.map_err(AppError::from)?;
db_client.query(query).await?.check()?;
Ok(())
}
pub async fn delete_relationships_by_source_id(
source_id: &str,
user_id: &str,
db_client: &SurrealDbClient,
) -> Result<(), AppError> {
db_client
.client
.query(
"DELETE FROM relates_to WHERE metadata.source_id = $source_id AND metadata.user_id = $user_id",
)
.bind(("source_id", source_id.to_owned()))
.bind(("user_id", user_id.to_owned()))
.await
.map_err(AppError::from)?
.check()
.map_err(AppError::from)?;
let query = format!(
"DELETE knowledge_entity -> relates_to WHERE metadata.source_id = '{source_id}'"
);
db_client.query(query).await?;
Ok(())
}
@@ -113,29 +78,18 @@ impl KnowledgeRelationship {
user_id: &str,
db_client: &SurrealDbClient,
) -> Result<(), AppError> {
let mut delete_result = db_client
.client
.query(
"DELETE type::thing('relates_to', $id) WHERE metadata.user_id = $user_id RETURN BEFORE;",
)
.bind(("id", id.to_owned()))
.bind(("user_id", user_id.to_owned()))
.await
.map_err(AppError::from)?;
let deleted: Vec<KnowledgeRelationship> = delete_result.take(0).map_err(AppError::from)?;
if !deleted.is_empty() {
return Ok(());
}
let mut authorized_result = db_client
.query(format!(
"SELECT * FROM relates_to WHERE id = relates_to:`{id}` AND metadata.user_id = '{user_id}'"
))
.await?;
let authorized: Vec<KnowledgeRelationship> = authorized_result.take(0).unwrap_or_default();
if authorized.is_empty() {
let mut exists_result = db_client
.client
.query("SELECT * FROM type::thing('relates_to', $id)")
.bind(("id", id.to_owned()))
.await
.map_err(AppError::from)?;
let existing: Option<KnowledgeRelationship> =
exists_result.take(0).map_err(AppError::from)?;
.query(format!("SELECT * FROM relates_to:`{id}`"))
.await?;
let existing: Option<KnowledgeRelationship> = exists_result.take(0)?;
if existing.is_some() {
Err(AppError::Auth(
@@ -144,40 +98,24 @@ impl KnowledgeRelationship {
} else {
Err(AppError::NotFound(format!("Relationship {id} not found")))
}
} else {
db_client.query(format!("DELETE relates_to:`{id}`")).await?;
Ok(())
}
}
}
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use super::*;
use crate::storage::types::knowledge_entity::{KnowledgeEntity, KnowledgeEntityType};
use anyhow::{self, Context};
use crate::test_utils::setup_test_db;
async fn get_relationship_by_id(
relationship_id: &str,
db_client: &SurrealDbClient,
) -> Option<KnowledgeRelationship> {
let mut result = db_client
.client
.query("SELECT * FROM type::thing('relates_to', $id)")
.bind(("id", relationship_id.to_owned()))
.await
.expect("relationship query by id failed");
result.take(0).expect("failed to take relationship by id")
}
async fn create_test_entity(
name: &str,
user_id: &str,
db_client: &SurrealDbClient,
) -> anyhow::Result<String> {
// Helper function to create a test knowledge entity for the relationship tests
async fn create_test_entity(name: &str, db_client: &SurrealDbClient) -> String {
let source_id = "source123".to_string();
let description = format!("Description for {name}");
let description = format!("Description for {}", name);
let entity_type = KnowledgeEntityType::Document;
let user_id = "user123".to_string();
let entity = KnowledgeEntity::new(
source_id,
@@ -185,20 +123,18 @@ mod tests {
description,
entity_type,
None,
user_id.to_string(),
user_id,
);
let stored: Option<KnowledgeEntity> = db_client
.store_item(entity)
.await
.with_context(|| "Failed to store entity".to_string())?;
stored
.ok_or_else(|| anyhow::anyhow!("Expected stored entity to return Some"))
.map(|e| e.id)
.expect("Failed to store entity");
stored.unwrap().id
}
#[tokio::test]
async fn test_relationship_creation() -> anyhow::Result<()> {
async fn test_relationship_creation() {
let in_id = "entity1".to_string();
let out_id = "entity2".to_string();
let user_id = "user123".to_string();
@@ -213,157 +149,107 @@ mod tests {
relationship_type.clone(),
);
// Verify fields are correctly set
assert_eq!(relationship.in_, in_id);
assert_eq!(relationship.out, out_id);
assert_eq!(relationship.metadata.user_id, user_id);
assert_eq!(relationship.metadata.source_id, source_id);
assert_eq!(relationship.metadata.relationship_type, relationship_type);
assert!(!relationship.id.is_empty());
Ok(())
}
#[tokio::test]
async fn test_store_and_verify_by_source_id() -> anyhow::Result<()> {
let db = setup_test_db().await?;
let user_id = "user123";
async fn test_store_and_verify_by_source_id() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("Failed to start in-memory surrealdb");
let entity1_id = create_test_entity("Entity 1", user_id, &db).await?;
let entity2_id = create_test_entity("Entity 2", user_id, &db).await?;
db.apply_migrations()
.await
.expect("Failed to apply migrations");
// Create two entities to relate
let entity1_id = create_test_entity("Entity 1", &db).await;
let entity2_id = create_test_entity("Entity 2", &db).await;
// Create relationship
let user_id = "user123".to_string();
let source_id = "source123".to_string();
let relationship_type = "references".to_string();
let relationship = KnowledgeRelationship::new(
entity1_id.clone(),
entity2_id.clone(),
user_id.to_string(),
user_id.clone(),
source_id.clone(),
relationship_type,
);
let relationship_id = relationship.id.clone();
// Store the relationship
relationship
.store_relationship(&db)
.await
.with_context(|| "Failed to store relationship".to_string())?;
.expect("Failed to store relationship");
let persisted = get_relationship_by_id(&relationship_id, &db)
.await
.expect("Relationship should be retrievable by id");
assert_eq!(persisted.in_, entity1_id);
assert_eq!(persisted.out, entity2_id);
assert_eq!(persisted.metadata.user_id, user_id);
assert_eq!(persisted.metadata.source_id, source_id);
let mut check_result = db
.query("SELECT * FROM relates_to WHERE metadata.source_id = $source_id")
.bind(("source_id", source_id.clone()))
.await
.expect("Check query failed");
// Query to verify the relationship exists by checking for relationships with our source_id
// This approach is more reliable than trying to look up by ID
let check_query = format!(
"SELECT * FROM relates_to WHERE metadata.source_id = '{}'",
source_id
);
let mut check_result = db.query(check_query).await.expect("Check query failed");
let check_results: Vec<KnowledgeRelationship> = check_result.take(0).unwrap_or_default();
assert_eq!(
check_results.len(),
1,
"Expected one relationship for source_id"
// Just verify that a relationship was created
assert!(
!check_results.is_empty(),
"Relationship should exist in the database"
);
Ok(())
}
#[tokio::test]
async fn test_store_relationship_rejects_foreign_entity() -> anyhow::Result<()> {
let db = setup_test_db().await?;
let owner_entity = create_test_entity("Owner entity", "owner-user", &db).await?;
let other_entity = create_test_entity("Other entity", "other-user", &db).await?;
let relationship = KnowledgeRelationship::new(
owner_entity,
other_entity,
"owner-user".to_string(),
"source123".to_string(),
"references".to_string(),
);
let result = relationship.store_relationship(&db).await;
assert!(matches!(result, Err(AppError::Auth(_))));
Ok(())
}
#[tokio::test]
async fn test_store_relationship_resists_query_injection() -> anyhow::Result<()> {
let db = setup_test_db().await?;
let user_id = "user123";
let entity1_id = create_test_entity("Entity 1", user_id, &db).await?;
let entity2_id = create_test_entity("Entity 2", user_id, &db).await?;
let relationship = KnowledgeRelationship::new(
entity1_id,
entity2_id,
user_id.to_string(),
"source123'; DELETE FROM relates_to; --".to_string(),
"references'; UPDATE user SET admin = true; --".to_string(),
);
let relationship_id = relationship.id.clone();
relationship
.store_relationship(&db)
async fn test_store_and_delete_relationship() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("store relationship should safely handle quote-containing values");
.expect("Failed to start in-memory surrealdb");
let mut res = db
.client
.query("SELECT * FROM relates_to WHERE id = type::thing('relates_to', $id)")
.bind(("id", relationship_id))
.await
.expect("query relationship by id failed");
let rows: Vec<KnowledgeRelationship> = res.take(0).expect("take rows");
assert_eq!(rows.len(), 1);
let row = rows.first().expect("expected 1 row");
assert_eq!(
row.metadata.source_id,
"source123'; DELETE FROM relates_to; --"
);
Ok(())
}
#[tokio::test]
async fn test_store_and_delete_relationship() -> anyhow::Result<()> {
let db = setup_test_db().await?;
let user_id = "user123";
let entity1_id = create_test_entity("Entity 1", user_id, &db).await?;
let entity2_id = create_test_entity("Entity 2", user_id, &db).await?;
// Create two entities to relate
let entity1_id = create_test_entity("Entity 1", &db).await;
let entity2_id = create_test_entity("Entity 2", &db).await;
// Create relationship
let user_id = "user123".to_string();
let source_id = "source123".to_string();
let relationship_type = "references".to_string();
let relationship = KnowledgeRelationship::new(
entity1_id.clone(),
entity2_id.clone(),
user_id.to_string(),
user_id.clone(),
source_id.clone(),
relationship_type,
);
let relationship_id = relationship.id.clone();
// Store relationship
relationship
.store_relationship(&db)
.await
.with_context(|| "Failed to store relationship".to_string())?;
.expect("Failed to store relationship");
// Ensure relationship exists before deletion attempt
let mut existing_before_delete = db
.query("SELECT * FROM relates_to WHERE metadata.user_id = $user_id AND metadata.source_id = $source_id")
.bind(("user_id", user_id.to_string()))
.bind(("source_id", source_id.clone()))
.query(format!(
"SELECT * FROM relates_to WHERE metadata.user_id = '{}' AND metadata.source_id = '{}'",
user_id, source_id
))
.await
.with_context(|| "Query failed".to_string())?;
.expect("Query failed");
let before_results: Vec<KnowledgeRelationship> =
existing_before_delete.take(0).unwrap_or_default();
assert!(
@@ -371,52 +257,59 @@ mod tests {
"Relationship should exist before deletion"
);
KnowledgeRelationship::delete_relationship_by_id(&relationship_id, user_id, &db)
// Delete relationship by ID
KnowledgeRelationship::delete_relationship_by_id(&relationship.id, &user_id, &db)
.await
.with_context(|| "Failed to delete relationship by ID".to_string())?;
.expect("Failed to delete relationship by ID");
// Query to verify relationship was deleted
let mut result = db
.query("SELECT * FROM relates_to WHERE metadata.user_id = $user_id AND metadata.source_id = $source_id")
.bind(("user_id", user_id.to_string()))
.bind(("source_id", source_id))
.query(format!(
"SELECT * FROM relates_to WHERE metadata.user_id = '{}' AND metadata.source_id = '{}'",
user_id, source_id
))
.await
.with_context(|| "Query failed".to_string())?;
.expect("Query failed");
let results: Vec<KnowledgeRelationship> = result.take(0).unwrap_or_default();
// Verify relationship no longer exists
assert!(results.is_empty(), "Relationship should be deleted");
Ok(())
}
#[tokio::test]
async fn test_delete_relationship_by_id_unauthorized() -> anyhow::Result<()> {
let db = setup_test_db().await?;
let owner_user_id = "owner-user";
async fn test_delete_relationship_by_id_unauthorized() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("Failed to start in-memory surrealdb");
let entity1_id = create_test_entity("Entity 1", owner_user_id, &db).await?;
let entity2_id = create_test_entity("Entity 2", owner_user_id, &db).await?;
let entity1_id = create_test_entity("Entity 1", &db).await;
let entity2_id = create_test_entity("Entity 2", &db).await;
let owner_user_id = "owner-user".to_string();
let source_id = "source123".to_string();
let relationship = KnowledgeRelationship::new(
entity1_id.clone(),
entity2_id.clone(),
owner_user_id.to_string(),
owner_user_id.clone(),
source_id,
"references".to_string(),
);
let relationship_id = relationship.id.clone();
relationship
.store_relationship(&db)
.await
.with_context(|| "Failed to store relationship".to_string())?;
.expect("Failed to store relationship");
let mut before_attempt = db
.query("SELECT * FROM relates_to WHERE metadata.user_id = $user_id")
.bind(("user_id", owner_user_id.to_string()))
.query(format!(
"SELECT * FROM relates_to WHERE metadata.user_id = '{}'",
owner_user_id
))
.await
.with_context(|| "Query failed".to_string())?;
.expect("Query failed");
let before_results: Vec<KnowledgeRelationship> = before_attempt.take(0).unwrap_or_default();
assert!(
!before_results.is_empty(),
@@ -424,7 +317,7 @@ mod tests {
);
let result = KnowledgeRelationship::delete_relationship_by_id(
&relationship_id,
&relationship.id,
"different-user",
&db,
)
@@ -432,42 +325,48 @@ mod tests {
match result {
Err(AppError::Auth(_)) => {}
_ => anyhow::bail!(
"Expected authorization error when deleting someone else's relationship"
),
_ => panic!("Expected authorization error when deleting someone else's relationship"),
}
let mut after_attempt = db
.query("SELECT * FROM relates_to WHERE metadata.user_id = $user_id")
.bind(("user_id", owner_user_id.to_string()))
.query(format!(
"SELECT * FROM relates_to WHERE metadata.user_id = '{}'",
owner_user_id
))
.await
.with_context(|| "Query failed".to_string())?;
.expect("Query failed");
let results: Vec<KnowledgeRelationship> = after_attempt.take(0).unwrap_or_default();
assert!(
!results.is_empty(),
"Relationship should still exist after unauthorized delete attempt"
);
Ok(())
}
#[tokio::test]
async fn test_store_relationship_exists() -> anyhow::Result<()> {
let db = setup_test_db().await?;
let user_id = "user123";
async fn test_store_relationship_exists() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.expect("Failed to start in-memory surrealdb");
let entity1_id = create_test_entity("Entity 1", user_id, &db).await?;
let entity2_id = create_test_entity("Entity 2", user_id, &db).await?;
let entity3_id = create_test_entity("Entity 3", user_id, &db).await?;
// Create entities to relate
let entity1_id = create_test_entity("Entity 1", &db).await;
let entity2_id = create_test_entity("Entity 2", &db).await;
let entity3_id = create_test_entity("Entity 3", &db).await;
// Create relationships with the same source_id
let user_id = "user123".to_string();
let source_id = "source123".to_string();
let different_source_id = "different_source".to_string();
// Create two relationships with the same source_id
let relationship1 = KnowledgeRelationship::new(
entity1_id.clone(),
entity2_id.clone(),
user_id.to_string(),
user_id.clone(),
source_id.clone(),
"references".to_string(),
);
@@ -475,170 +374,77 @@ mod tests {
let relationship2 = KnowledgeRelationship::new(
entity2_id.clone(),
entity3_id.clone(),
user_id.to_string(),
user_id.clone(),
source_id.clone(),
"contains".to_string(),
);
// Create a relationship with a different source_id
let different_relationship = KnowledgeRelationship::new(
entity1_id.clone(),
entity3_id.clone(),
user_id.to_string(),
user_id.clone(),
different_source_id.clone(),
"mentions".to_string(),
);
let relationship1_id = relationship1.id.clone();
let relationship2_id = relationship2.id.clone();
let different_relationship_id = different_relationship.id.clone();
// Store all relationships
relationship1
.store_relationship(&db)
.await
.with_context(|| "Failed to store relationship 1".to_string())?;
.expect("Failed to store relationship 1");
relationship2
.store_relationship(&db)
.await
.with_context(|| "Failed to store relationship 2".to_string())?;
.expect("Failed to store relationship 2");
different_relationship
.store_relationship(&db)
.await
.with_context(|| "Failed to store different relationship".to_string())?;
.expect("Failed to store different relationship");
let mut before_delete = db
.query("SELECT * FROM relates_to WHERE metadata.source_id = $source_id")
.bind(("source_id", source_id.clone()))
// Delete relationships by source_id
KnowledgeRelationship::delete_relationships_by_source_id(&source_id, &db)
.await
.expect("before delete query failed");
let before_delete_rows: Vec<KnowledgeRelationship> =
before_delete.take(0).unwrap_or_default();
assert_eq!(before_delete_rows.len(), 2);
.expect("Failed to delete relationships by source_id");
let mut before_delete_different = db
.query("SELECT * FROM relates_to WHERE metadata.source_id = $source_id")
.bind(("source_id", different_source_id.clone()))
.await
.expect("before delete different query failed");
let before_delete_different_rows: Vec<KnowledgeRelationship> =
before_delete_different.take(0).unwrap_or_default();
assert_eq!(before_delete_different_rows.len(), 1);
KnowledgeRelationship::delete_relationships_by_source_id(&source_id, user_id, &db)
.await
.with_context(|| "Failed to delete relationships by source_id".to_string())?;
let result1 = get_relationship_by_id(&relationship1_id, &db).await;
let result2 = get_relationship_by_id(&relationship2_id, &db).await;
let different_result = get_relationship_by_id(&different_relationship_id, &db).await;
assert!(result1.is_none(), "Relationship 1 should be deleted");
assert!(result2.is_none(), "Relationship 2 should be deleted");
let remaining =
different_result.expect("Relationship with different source_id should remain");
assert_eq!(remaining.metadata.source_id, different_source_id);
Ok(())
}
#[tokio::test]
async fn test_delete_relationships_by_source_id_scoped_to_user() -> anyhow::Result<()> {
let db = setup_test_db().await?;
let user_a = "user-a";
let user_b = "user-b";
let shared_source = "shared-source";
let a1 = create_test_entity("A1", user_a, &db).await?;
let a2 = create_test_entity("A2", user_a, &db).await?;
let b1 = create_test_entity("B1", user_b, &db).await?;
let b2 = create_test_entity("B2", user_b, &db).await?;
let rel_a = KnowledgeRelationship::new(
a1,
a2,
user_a.to_string(),
shared_source.to_string(),
"references".to_string(),
);
let rel_b = KnowledgeRelationship::new(
b1,
b2,
user_b.to_string(),
shared_source.to_string(),
"references".to_string(),
);
let owner_relationship_id = rel_a.id.clone();
let other_relationship_id = rel_b.id.clone();
rel_a.store_relationship(&db).await?;
rel_b.store_relationship(&db).await?;
KnowledgeRelationship::delete_relationships_by_source_id(shared_source, user_a, &db)
.await?;
assert!(get_relationship_by_id(&owner_relationship_id, &db)
.await
.is_none());
assert!(get_relationship_by_id(&other_relationship_id, &db)
.await
.is_some());
Ok(())
}
#[tokio::test]
async fn test_delete_relationships_by_source_id_resists_query_injection() -> anyhow::Result<()>
{
let db = setup_test_db().await?;
let user_id = "user123";
let entity1_id = create_test_entity("Entity 1", user_id, &db).await?;
let entity2_id = create_test_entity("Entity 2", user_id, &db).await?;
let entity3_id = create_test_entity("Entity 3", user_id, &db).await?;
let safe_relationship = KnowledgeRelationship::new(
entity1_id.clone(),
entity2_id.clone(),
user_id.to_string(),
"safe_source".to_string(),
"references".to_string(),
// Query to verify the relationships with source_id were deleted
let query1 = format!("SELECT * FROM relates_to WHERE id = '{}'", relationship1.id);
let query2 = format!("SELECT * FROM relates_to WHERE id = '{}'", relationship2.id);
let different_query = format!(
"SELECT * FROM relates_to WHERE id = '{}'",
different_relationship.id
);
let other_relationship = KnowledgeRelationship::new(
entity2_id,
entity3_id,
user_id.to_string(),
"other_source".to_string(),
"contains".to_string(),
let mut result1 = db.query(query1).await.expect("Query 1 failed");
let results1: Vec<KnowledgeRelationship> = result1.take(0).unwrap_or_default();
let mut result2 = db.query(query2).await.expect("Query 2 failed");
let results2: Vec<KnowledgeRelationship> = result2.take(0).unwrap_or_default();
let mut different_result = db
.query(different_query)
.await
.expect("Different query failed");
let _different_results: Vec<KnowledgeRelationship> =
different_result.take(0).unwrap_or_default();
// Verify relationships with the source_id are deleted
assert!(results1.is_empty(), "Relationship 1 should be deleted");
assert!(results2.is_empty(), "Relationship 2 should be deleted");
// For the relationship with different source ID, we need to check differently
// Let's just verify we have a relationship where the source_id matches different_source_id
let check_query = format!(
"SELECT * FROM relates_to WHERE metadata.source_id = '{}'",
different_source_id
);
let safe_relationship_id = safe_relationship.id.clone();
let other_relationship_id = other_relationship.id.clone();
let mut check_result = db.query(check_query).await.expect("Check query failed");
let check_results: Vec<KnowledgeRelationship> = check_result.take(0).unwrap_or_default();
safe_relationship
.store_relationship(&db)
.await
.expect("store safe relationship");
other_relationship
.store_relationship(&db)
.await
.expect("store other relationship");
KnowledgeRelationship::delete_relationships_by_source_id(
"safe_source' OR 1=1 --",
user_id,
&db,
)
.await
.expect("delete call should succeed");
let remaining_safe = get_relationship_by_id(&safe_relationship_id, &db).await;
let remaining_other = get_relationship_by_id(&other_relationship_id, &db).await;
assert!(remaining_safe.is_some(), "Safe relationship should remain");
// Verify the relationship with a different source_id still exists
assert!(
remaining_other.is_some(),
"Other relationship should remain"
!check_results.is_empty(),
"Relationship with different source_id should still exist"
);
Ok(())
}
}
+29 -40
View File
@@ -1,12 +1,9 @@
#![allow(clippy::module_name_repetitions)]
use uuid::Uuid;
use std::fmt;
use std::fmt::Write;
use crate::stored_object;
#[derive(Deserialize, Debug, Clone, Copy, Serialize, PartialEq)]
#[derive(Deserialize, Debug, Clone, Serialize, PartialEq)]
pub enum MessageRole {
User,
AI,
@@ -21,7 +18,6 @@ stored_object!(Message, "message", {
});
impl Message {
#[must_use]
pub fn new(
conversation_id: String,
role: MessageRole,
@@ -58,31 +54,22 @@ impl fmt::Display for Message {
}
// helper function to format a vector of messages
#[must_use]
pub fn format_history(history: &[Message]) -> String {
let estimated: usize = history
history
.iter()
.map(|m| m.content.len().saturating_add(10))
.sum();
let mut out = String::with_capacity(estimated);
for (i, msg) in history.iter().enumerate() {
if i > 0 {
out.push('\n');
}
let _ = write!(out, "{msg}");
}
out
.map(|msg| format!("{msg}"))
.collect::<Vec<String>>()
.join("\n")
}
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use super::*;
use crate::storage::db::SurrealDbClient;
use anyhow::{self, Context};
#[tokio::test]
async fn test_message_creation() -> anyhow::Result<()> {
async fn test_message_creation() {
// Test basic message creation
let conversation_id = "test_conversation";
let content = "This is a test message";
let role = MessageRole::User;
@@ -90,28 +77,29 @@ mod tests {
let message = Message::new(
conversation_id.to_string(),
role,
role.clone(),
content.to_string(),
references.clone(),
);
// Verify message properties
assert_eq!(message.conversation_id, conversation_id);
assert_eq!(message.content, content);
assert_eq!(message.role, role);
assert_eq!(message.references, references);
assert!(!message.id.is_empty());
Ok(())
}
#[tokio::test]
async fn test_message_persistence() -> anyhow::Result<()> {
async fn test_message_persistence() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &uuid::Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
// Create and store a message
let conversation_id = "test_conversation";
let message = Message::new(
conversation_id.to_string(),
@@ -121,37 +109,39 @@ mod tests {
);
let message_id = message.id.clone();
// Store the message
db.store_item(message.clone())
.await
.with_context(|| "Failed to store message".to_string())?;
.expect("Failed to store message");
// Retrieve the message
let retrieved: Option<Message> = db
.get_item(&message_id)
.await
.with_context(|| "Failed to retrieve message".to_string())?;
.expect("Failed to retrieve message");
let retrieved = retrieved.ok_or_else(|| anyhow::anyhow!("Expected message to exist"))?;
assert!(retrieved.is_some());
let retrieved = retrieved.unwrap();
// Verify retrieved properties match original
assert_eq!(retrieved.id, message.id);
assert_eq!(retrieved.conversation_id, message.conversation_id);
assert_eq!(retrieved.role, message.role);
assert_eq!(retrieved.content, message.content);
assert_eq!(retrieved.references, message.references);
Ok(())
}
#[tokio::test]
async fn test_message_role_display() -> anyhow::Result<()> {
async fn test_message_role_display() {
// Test the Display implementation for MessageRole
assert_eq!(format!("{}", MessageRole::User), "User");
assert_eq!(format!("{}", MessageRole::AI), "AI");
assert_eq!(format!("{}", MessageRole::System), "System");
Ok(())
}
#[tokio::test]
async fn test_message_display() -> anyhow::Result<()> {
async fn test_message_display() {
// Test the Display implementation for Message
let message = Message {
id: "test_id".to_string(),
created_at: Utc::now(),
@@ -162,13 +152,12 @@ mod tests {
references: None,
};
assert_eq!(format!("{message}"), "User: Hello world");
Ok(())
assert_eq!(format!("{}", message), "User: Hello world");
}
#[tokio::test]
async fn test_format_history() -> anyhow::Result<()> {
async fn test_format_history() {
// Create a vector of messages
let messages = vec![
Message {
id: "1".to_string(),
@@ -190,10 +179,10 @@ mod tests {
},
];
// Format the history
let formatted = format_history(&messages);
// Verify the formatting
assert_eq!(formatted, "User: Hello\nAI: Hi there!");
Ok(())
}
}
+90 -8
View File
@@ -19,21 +19,103 @@ pub mod user;
pub trait StoredObject: Serialize + for<'de> Deserialize<'de> {
fn table_name() -> &'static str;
fn id(&self) -> &str;
fn get_id(&self) -> &str;
}
#[macro_export]
macro_rules! stored_object {
($(#[$struct_attr:meta])* $name:ident, $table:expr, {$($(#[$field_attr:meta])* $field:ident: $ty:ty),*}) => {
use serde::{Deserialize, Serialize};
use serde::{Deserialize, Deserializer, Serialize};
use surrealdb::sql::Thing;
use $crate::storage::types::StoredObject;
#[allow(unused_imports)]
use $crate::utils::serde_helpers::{
deserialize_flexible_id, serialize_datetime, deserialize_datetime,
serialize_option_datetime, deserialize_option_datetime,
};
use serde::de::{self, Visitor};
use std::fmt;
use chrono::{DateTime, Utc };
struct FlexibleIdVisitor;
impl<'de> Visitor<'de> for FlexibleIdVisitor {
type Value = String;
fn expecting(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
formatter.write_str("a string or a Thing")
}
fn visit_str<E>(self, value: &str) -> Result<Self::Value, E>
where
E: de::Error,
{
Ok(value.to_string())
}
fn visit_string<E>(self, value: String) -> Result<Self::Value, E>
where
E: de::Error,
{
Ok(value)
}
fn visit_map<A>(self, map: A) -> Result<Self::Value, A::Error>
where
A: de::MapAccess<'de>,
{
// Try to deserialize as Thing
let thing = Thing::deserialize(de::value::MapAccessDeserializer::new(map))?;
Ok(thing.id.to_raw())
}
}
pub fn deserialize_flexible_id<'de, D>(deserializer: D) -> Result<String, D::Error>
where
D: Deserializer<'de>,
{
deserializer.deserialize_any(FlexibleIdVisitor)
}
fn serialize_datetime<S>(date: &DateTime<Utc>, serializer: S) -> Result<S::Ok, S::Error>
where
S: serde::Serializer,
{
Into::<surrealdb::sql::Datetime>::into(*date).serialize(serializer)
}
fn deserialize_datetime<'de, D>(deserializer: D) -> Result<DateTime<Utc>, D::Error>
where
D: serde::Deserializer<'de>,
{
let dt = surrealdb::sql::Datetime::deserialize(deserializer)?;
Ok(DateTime::<Utc>::from(dt))
}
#[allow(dead_code)]
#[allow(clippy::ref_option)]
fn serialize_option_datetime<S>(
date: &Option<DateTime<Utc>>,
serializer: S,
) -> Result<S::Ok, S::Error>
where
S: serde::Serializer,
{
match date {
Some(dt) => serializer
.serialize_some(&Into::<surrealdb::sql::Datetime>::into(*dt)),
None => serializer.serialize_none(),
}
}
#[allow(dead_code)]
#[allow(clippy::ref_option)]
fn deserialize_option_datetime<'de, D>(
deserializer: D,
) -> Result<Option<DateTime<Utc>>, D::Error>
where
D: serde::Deserializer<'de>,
{
let value = Option::<surrealdb::sql::Datetime>::deserialize(deserializer)?;
Ok(value.map(DateTime::<Utc>::from))
}
$(#[$struct_attr])*
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct $name {
@@ -51,7 +133,7 @@ macro_rules! stored_object {
$table
}
fn id(&self) -> &str {
fn get_id(&self) -> &str {
&self.id
}
}
+55 -100
View File
@@ -28,7 +28,6 @@ stored_object!(Scratchpad, "scratchpad", {
});
impl Scratchpad {
#[must_use]
pub fn new(user_id: String, title: String) -> Self {
let now = ChronoUtc::now();
Self {
@@ -79,7 +78,7 @@ impl Scratchpad {
let scratchpad: Option<Scratchpad> = db.get_item(id).await?;
let scratchpad =
scratchpad.ok_or_else(|| AppError::NotFound("scratchpad not found".to_string()))?;
scratchpad.ok_or_else(|| AppError::NotFound("Scratchpad not found".to_string()))?;
if scratchpad.user_id != user_id {
return Err(AppError::Auth(
@@ -217,23 +216,20 @@ impl Scratchpad {
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use anyhow::{self, Context};
use super::*;
#[tokio::test]
async fn test_create_scratchpad() -> anyhow::Result<()> {
async fn test_create_scratchpad() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
.expect("Failed to apply migrations");
// Create a new scratchpad
let user_id = "test_user";
@@ -258,28 +254,29 @@ mod tests {
let retrieved: Option<Scratchpad> = db
.get_item(&scratchpad.id)
.await
.with_context(|| "Failed to retrieve scratchpad".to_string())?;
let retrieved = retrieved.with_context(|| "expected scratchpad to exist".to_string())?;
.expect("Failed to retrieve scratchpad");
assert!(retrieved.is_some());
let retrieved = retrieved.unwrap();
assert_eq!(retrieved.id, scratchpad.id);
assert_eq!(retrieved.user_id, user_id);
assert_eq!(retrieved.title, title);
assert!(!retrieved.is_archived);
assert!(retrieved.archived_at.is_none());
assert!(retrieved.ingested_at.is_none());
Ok(())
}
#[tokio::test]
async fn test_get_by_user() -> anyhow::Result<()> {
async fn test_get_by_user() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
.expect("Failed to apply migrations");
let user_id = "test_user";
@@ -291,30 +288,19 @@ mod tests {
// Store them
let scratchpad1_id = scratchpad1.id.clone();
let scratchpad2_id = scratchpad2.id.clone();
db.store_item(scratchpad1)
.await
.with_context(|| "store scratchpad1".to_string())?;
db.store_item(scratchpad2)
.await
.with_context(|| "store scratchpad2".to_string())?;
db.store_item(scratchpad3)
.await
.with_context(|| "store scratchpad3".to_string())?;
db.store_item(scratchpad1).await.unwrap();
db.store_item(scratchpad2).await.unwrap();
db.store_item(scratchpad3).await.unwrap();
// Archive one of the user's scratchpads
Scratchpad::archive(&scratchpad2_id, user_id, &db, false)
.await
.with_context(|| "archive".to_string())?;
.unwrap();
// Get scratchpads for user_id
let user_scratchpads = Scratchpad::get_by_user(user_id, &db)
.await
.with_context(|| "get_by_user".to_string())?;
let user_scratchpads = Scratchpad::get_by_user(user_id, &db).await.unwrap();
assert_eq!(user_scratchpads.len(), 1);
assert_eq!(
user_scratchpads.first().map(|s| &s.id),
Some(&scratchpad1_id)
);
assert_eq!(user_scratchpads[0].id, scratchpad1_id);
// Verify they belong to the user
for scratchpad in &user_scratchpads {
@@ -323,201 +309,177 @@ mod tests {
let archived = Scratchpad::get_archived_by_user(user_id, &db)
.await
.with_context(|| "get_archived_by_user".to_string())?;
.unwrap();
assert_eq!(archived.len(), 1);
assert_eq!(archived.first().map(|s| &s.id), Some(&scratchpad2_id));
assert!(archived.first().is_some_and(|s| s.is_archived));
assert!(archived.first().is_some_and(|s| s.ingested_at.is_none()));
Ok(())
assert_eq!(archived[0].id, scratchpad2_id);
assert!(archived[0].is_archived);
assert!(archived[0].ingested_at.is_none());
}
#[tokio::test]
async fn test_archive_and_restore() -> anyhow::Result<()> {
async fn test_archive_and_restore() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
.expect("Failed to apply migrations");
let user_id = "test_user";
let scratchpad = Scratchpad::new(user_id.to_string(), "Test".to_string());
let scratchpad_id = scratchpad.id.clone();
db.store_item(scratchpad)
.await
.with_context(|| "store scratchpad".to_string())?;
db.store_item(scratchpad).await.unwrap();
let archived = Scratchpad::archive(&scratchpad_id, user_id, &db, true)
.await
.with_context(|| "Failed to archive".to_string())?;
.expect("Failed to archive");
assert!(archived.is_archived);
assert!(archived.archived_at.is_some());
assert!(archived.ingested_at.is_some());
let restored = Scratchpad::restore(&scratchpad_id, user_id, &db)
.await
.with_context(|| "Failed to restore".to_string())?;
.expect("Failed to restore");
assert!(!restored.is_archived);
assert!(restored.archived_at.is_none());
assert!(restored.ingested_at.is_none());
Ok(())
}
#[tokio::test]
async fn test_update_content() -> anyhow::Result<()> {
async fn test_update_content() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
.expect("Failed to apply migrations");
let user_id = "test_user";
let scratchpad = Scratchpad::new(user_id.to_string(), "Test".to_string());
let scratchpad_id = scratchpad.id.clone();
db.store_item(scratchpad)
.await
.with_context(|| "store scratchpad".to_string())?;
db.store_item(scratchpad).await.unwrap();
let new_content = "Updated content";
let updated = Scratchpad::update_content(&scratchpad_id, user_id, new_content, &db)
.await
.with_context(|| "update_content".to_string())?;
.unwrap();
assert_eq!(updated.content, new_content);
assert!(!updated.is_dirty);
Ok(())
}
#[tokio::test]
async fn test_update_content_unauthorized() -> anyhow::Result<()> {
async fn test_update_content_unauthorized() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
.expect("Failed to apply migrations");
let owner_id = "owner";
let other_user = "other_user";
let scratchpad = Scratchpad::new(owner_id.to_string(), "Test".to_string());
let scratchpad_id = scratchpad.id.clone();
db.store_item(scratchpad)
.await
.with_context(|| "store scratchpad".to_string())?;
db.store_item(scratchpad).await.unwrap();
let result = Scratchpad::update_content(&scratchpad_id, other_user, "Hacked", &db).await;
assert!(result.is_err());
match result {
Err(AppError::Auth(_)) => {}
_ => anyhow::bail!("Expected Auth error"),
_ => panic!("Expected Auth error"),
}
Ok(())
}
#[tokio::test]
async fn test_delete_scratchpad() -> anyhow::Result<()> {
async fn test_delete_scratchpad() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
.expect("Failed to apply migrations");
let user_id = "test_user";
let scratchpad = Scratchpad::new(user_id.to_string(), "Test".to_string());
let scratchpad_id = scratchpad.id.clone();
db.store_item(scratchpad)
.await
.with_context(|| "store scratchpad".to_string())?;
db.store_item(scratchpad).await.unwrap();
// Delete should succeed
let result = Scratchpad::delete(&scratchpad_id, user_id, &db).await;
assert!(result.is_ok());
// Verify it's gone
let retrieved: Option<Scratchpad> = db
.get_item(&scratchpad_id)
.await
.with_context(|| "get_item".to_string())?;
let retrieved: Option<Scratchpad> = db.get_item(&scratchpad_id).await.unwrap();
assert!(retrieved.is_none());
Ok(())
}
#[tokio::test]
async fn test_delete_unauthorized() -> anyhow::Result<()> {
async fn test_delete_unauthorized() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
.expect("Failed to apply migrations");
let owner_id = "owner";
let other_user = "other_user";
let scratchpad = Scratchpad::new(owner_id.to_string(), "Test".to_string());
let scratchpad_id = scratchpad.id.clone();
db.store_item(scratchpad)
.await
.with_context(|| "store scratchpad".to_string())?;
db.store_item(scratchpad).await.unwrap();
let result = Scratchpad::delete(&scratchpad_id, other_user, &db).await;
assert!(result.is_err());
match result {
Err(AppError::Auth(_)) => {}
_ => anyhow::bail!("Expected Auth error"),
_ => panic!("Expected Auth error"),
}
// Verify it still exists
let retrieved: Option<Scratchpad> = db
.get_item(&scratchpad_id)
.await
.with_context(|| "get_item".to_string())?;
let retrieved: Option<Scratchpad> = db.get_item(&scratchpad_id).await.unwrap();
assert!(retrieved.is_some());
Ok(())
}
#[tokio::test]
async fn test_timezone_aware_scratchpad_conversion() -> anyhow::Result<()> {
async fn test_timezone_aware_scratchpad_conversion() {
let db = SurrealDbClient::memory("test_ns", &Uuid::new_v4().to_string())
.await
.with_context(|| "Failed to create test database".to_string())?;
.expect("Failed to create test database");
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
.expect("Failed to apply migrations");
let user_id = "test_user_123";
let scratchpad =
Scratchpad::new(user_id.to_string(), "Test Timezone Scratchpad".to_string());
let scratchpad_id = scratchpad.id.clone();
db.store_item(scratchpad)
.await
.with_context(|| "store scratchpad".to_string())?;
db.store_item(scratchpad).await.unwrap();
let retrieved = Scratchpad::get_by_id(&scratchpad_id, user_id, &db)
.await
.with_context(|| "get_by_id".to_string())?;
.unwrap();
// Test that datetime fields are preserved and can be used for timezone formatting
assert!(retrieved.created_at.timestamp() > 0);
@@ -531,17 +493,10 @@ mod tests {
// Archive the scratchpad to test optional datetime handling
let archived = Scratchpad::archive(&scratchpad_id, user_id, &db, false)
.await
.with_context(|| "archive".to_string())?;
.unwrap();
assert!(archived.archived_at.is_some());
assert!(
archived
.archived_at
.with_context(|| "expected archived_at".to_string())?
.timestamp()
> 0
);
assert!(archived.archived_at.unwrap().timestamp() > 0);
assert!(archived.ingested_at.is_none());
Ok(())
}
}
+4 -4
View File
@@ -1,4 +1,4 @@
pub const DEFAULT_QUERY_SYSTEM_PROMPT: &str = r#"You are a knowledgeable assistant with access to a specialized knowledge base. You will be provided with relevant knowledge entities from the database as context. Each knowledge entity contains a name, description, and type, representing different concepts, ideas, and information.
pub static DEFAULT_QUERY_SYSTEM_PROMPT: &str = r#"You are a knowledgeable assistant with access to a specialized knowledge base. You will be provided with relevant knowledge entities from the database as context. Each knowledge entity contains a name, description, and type, representing different concepts, ideas, and information.
Your task is to:
1. Carefully analyze the provided knowledge entities in the context
@@ -20,7 +20,7 @@ Example response formats:
"I found relevant information in multiple entries: [explanation...]"
"I apologize, but the provided context doesn't contain information about [topic]""#;
pub const DEFAULT_INGRESS_ANALYSIS_SYSTEM_PROMPT: &str = r#"You are an AI assistant. You will receive a text content, along with user context and a category. Your task is to provide a structured JSON object representing the content in a graph format suitable for a graph database. You will also be presented with some existing knowledge_entities from the database, do not replicate these! Your task is to create meaningful knowledge entities from the submitted content. Try and infer as much as possible from the users context and category when creating these. If the user submits a large content, create more general entities. If the user submits a narrow and precise content, try and create precise knowledge entities.
pub static DEFAULT_INGRESS_ANALYSIS_SYSTEM_PROMPT: &str = r#"You are an AI assistant. You will receive a text content, along with user context and a category. Your task is to provide a structured JSON object representing the content in a graph format suitable for a graph database. You will also be presented with some existing knowledge_entities from the database, do not replicate these! Your task is to create meaningful knowledge entities from the submitted content. Try and infer as much as possible from the users context and category when creating these. If the user submits a large content, create more general entities. If the user submits a narrow and precise content, try and create precise knowledge entities.
The JSON should have the following structure:
@@ -49,13 +49,13 @@ Guidelines:
2. Each KnowledgeEntity should have a unique `key`, a meaningful `name`, and a descriptive `description`.
3. Define the type of each KnowledgeEntity using the following categories: Idea, Project, Document, Page, TextSnippet.
4. Establish relationships between entities using types like RelatedTo, RelevantTo, SimilarTo.
5. Use the `source` key to indicate the originating entity and the `target` key to indicate the related entity.
5. Use the `source` key to indicate the originating entity and the `target` key to indicate the related entity"
6. You will be presented with a few existing KnowledgeEntities that are similar to the current ones. They will have an existing UUID. When creating relationships to these entities, use their UUID.
7. Only create relationships between existing KnowledgeEntities.
8. Entities that exist already in the database should NOT be created again. If there is only a minor overlap, skip creating a new entity.
9. A new relationship MUST include a newly created KnowledgeEntity."#;
pub const DEFAULT_IMAGE_PROCESSING_PROMPT: &str = r#"Analyze this image and respond based on its primary content:
pub static DEFAULT_IMAGE_PROCESSING_PROMPT: &str = r#"Analyze this image and respond based on its primary content:
- If the image is mainly text (document, screenshot, sign), transcribe the text verbatim.
- If the image is mainly visual (photograph, art, landscape), provide a concise description of the scene.
- For hybrid images (diagrams, ads), briefly describe the visual, then transcribe the text under a "Text:" heading.
+123 -474
View File
@@ -1,5 +1,4 @@
use crate::utils::config::EmbeddingBackend;
use crate::utils::serde_helpers::deserialize_flexible_id;
use crate::storage::types::file_info::deserialize_flexible_id;
use serde::{Deserialize, Serialize};
use crate::{error::AppError, storage::db::SurrealDbClient, storage::types::StoredObject};
@@ -14,9 +13,9 @@ pub struct SystemSettings {
pub processing_model: String,
pub embedding_model: String,
pub embedding_dimensions: u32,
/// Active embedding backend. Read-only for admin updates; synced from config at startup.
/// Active embedding backend ("openai", "fastembed", "hashed"). Read-only, synced from config.
#[serde(default)]
pub embedding_backend: Option<EmbeddingBackend>,
pub embedding_backend: Option<String>,
pub query_system_prompt: String,
pub ingestion_system_prompt: String,
pub image_processing_model: String,
@@ -24,151 +23,33 @@ pub struct SystemSettings {
pub voice_processing_model: String,
}
/// Partial update for singleton system settings without cloning unchanged fields.
#[derive(Debug, Default, Clone)]
#[allow(clippy::module_name_repetitions)]
pub struct SystemSettingsPatch {
pub registrations_enabled: Option<bool>,
pub require_email_verification: Option<bool>,
pub query_model: Option<String>,
pub processing_model: Option<String>,
pub embedding_model: Option<String>,
pub embedding_dimensions: Option<u32>,
pub query_system_prompt: Option<String>,
pub ingestion_system_prompt: Option<String>,
pub image_processing_model: Option<String>,
pub image_processing_prompt: Option<String>,
pub voice_processing_model: Option<String>,
}
enum UpdateMode {
User,
EmbeddingSync,
}
impl StoredObject for SystemSettings {
fn table_name() -> &'static str {
"system_settings"
}
fn id(&self) -> &str {
fn get_id(&self) -> &str {
&self.id
}
}
impl SystemSettingsPatch {
pub fn apply_to(self, settings: &mut SystemSettings) {
if let Some(value) = self.registrations_enabled {
settings.registrations_enabled = value;
}
if let Some(value) = self.require_email_verification {
settings.require_email_verification = value;
}
if let Some(value) = self.query_model {
settings.query_model = value;
}
if let Some(value) = self.processing_model {
settings.processing_model = value;
}
if let Some(value) = self.embedding_model {
settings.embedding_model = value;
}
if let Some(value) = self.embedding_dimensions {
settings.embedding_dimensions = value;
}
if let Some(value) = self.query_system_prompt {
settings.query_system_prompt = value;
}
if let Some(value) = self.ingestion_system_prompt {
settings.ingestion_system_prompt = value;
}
if let Some(value) = self.image_processing_model {
settings.image_processing_model = value;
}
if let Some(value) = self.image_processing_prompt {
settings.image_processing_prompt = value;
}
if let Some(value) = self.voice_processing_model {
settings.voice_processing_model = value;
}
}
pub async fn apply(self, db: &SurrealDbClient) -> Result<SystemSettings, AppError> {
let mut current = SystemSettings::get_current(db).await?;
self.apply_to(&mut current);
SystemSettings::update(db, current).await
}
}
impl SystemSettings {
pub const RECORD_ID: &'static str = "current";
#[allow(clippy::result_large_err)]
fn validate(&self) -> Result<(), AppError> {
if self.embedding_dimensions == 0 {
return Err(AppError::Validation(
"embedding_dimensions must be greater than 0".into(),
));
}
let model_fields = [
("query_model", &self.query_model),
("processing_model", &self.processing_model),
("embedding_model", &self.embedding_model),
("image_processing_model", &self.image_processing_model),
("voice_processing_model", &self.voice_processing_model),
];
for (name, value) in model_fields {
if value.trim().is_empty() {
return Err(AppError::Validation(format!("{name} must not be empty")));
}
}
let prompt_fields = [
("query_system_prompt", &self.query_system_prompt),
("ingestion_system_prompt", &self.ingestion_system_prompt),
("image_processing_prompt", &self.image_processing_prompt),
];
for (name, value) in prompt_fields {
if value.trim().is_empty() {
return Err(AppError::Validation(format!("{name} must not be empty")));
}
}
Ok(())
}
pub async fn get_current(db: &SurrealDbClient) -> Result<Self, AppError> {
let settings: Option<Self> = db.get_item(Self::RECORD_ID).await?;
settings.ok_or(AppError::NotFound("system settings not found".into()))
let settings: Option<Self> = db.get_item("current").await?;
settings.ok_or(AppError::NotFound("System settings not found".into()))
}
pub async fn update(db: &SurrealDbClient, changes: Self) -> Result<Self, AppError> {
Self::update_with_mode(db, changes, UpdateMode::User).await
}
async fn update_with_mode(
db: &SurrealDbClient,
mut changes: Self,
mode: UpdateMode,
) -> Result<Self, AppError> {
let current = Self::get_current(db).await?;
if matches!(mode, UpdateMode::User) {
changes.embedding_backend = current.embedding_backend;
}
changes.id = Self::RECORD_ID.to_string();
changes.validate()?;
// We need to use a direct query for the update with MERGE
let updated: Option<Self> = db
.client
.query("UPDATE type::thing('system_settings', $id) MERGE $changes RETURN AFTER")
.bind(("id", Self::RECORD_ID))
.query("UPDATE type::thing('system_settings', 'current') MERGE $changes RETURN AFTER")
.bind(("changes", changes))
.await?
.take(0)?;
updated.ok_or(AppError::NotFound(
"system settings record missing after update".into(),
updated.ok_or(AppError::Validation(
"Something went wrong updating the settings".into(),
))
}
@@ -182,23 +63,17 @@ impl SystemSettings {
let mut settings = Self::get_current(db).await?;
let mut needs_update = false;
let provider_backend = provider
.backend_label()
.parse::<EmbeddingBackend>()
.map_err(|e| AppError::Validation(e.to_string()))?;
let provider_dimensions = u32::try_from(provider.dimension()).map_err(|_| {
AppError::Validation(format!(
"embedding provider dimension {} exceeds u32::MAX",
provider.dimension()
))
})?;
let backend_label = provider.backend_label().to_string();
let provider_dimensions = provider.dimension() as u32;
let provider_model = provider.model_code();
if settings.embedding_backend != Some(provider_backend) {
settings.embedding_backend = Some(provider_backend);
// Sync backend label
if settings.embedding_backend.as_deref() != Some(&backend_label) {
settings.embedding_backend = Some(backend_label);
needs_update = true;
}
// Sync dimensions
if settings.embedding_dimensions != provider_dimensions {
tracing::info!(
old_dimensions = settings.embedding_dimensions,
@@ -209,6 +84,7 @@ impl SystemSettings {
needs_update = true;
}
// Sync model if provider has one
if let Some(model) = provider_model {
if settings.embedding_model != model {
tracing::info!(
@@ -222,7 +98,7 @@ impl SystemSettings {
}
if needs_update {
settings = Self::update_with_mode(db, settings, UpdateMode::EmbeddingSync).await?;
settings = Self::update(db, settings).await?;
}
Ok((settings, needs_update))
@@ -231,10 +107,9 @@ impl SystemSettings {
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use crate::storage::indexes::ensure_runtime;
use crate::storage::indexes::ensure_runtime_indexes;
use crate::storage::types::{knowledge_entity::KnowledgeEntity, text_chunk::TextChunk};
use anyhow::{self, Context};
use async_openai::Client;
use super::*;
use uuid::Uuid;
@@ -243,115 +118,88 @@ mod tests {
db: &SurrealDbClient,
table_name: &str,
index_name: &str,
) -> anyhow::Result<u32> {
) -> u32 {
let query = format!("INFO FOR TABLE {table_name};");
let mut response = db
.client
.query(query)
.await
.with_context(|| "Failed to fetch table info".to_string())?;
.expect("Failed to fetch table info");
let info: surrealdb::Value = response
.take(0)
.with_context(|| "Failed to extract table info response".to_string())?;
.expect("Failed to extract table info response");
let info_json: serde_json::Value = serde_json::to_value(info)
.with_context(|| "Failed to convert info to json".to_string())?;
let info_json: serde_json::Value =
serde_json::to_value(info).expect("Failed to convert info to json");
let indexes = info_json
.get("Object")
.and_then(|v| v.get("indexes"))
.and_then(|v| v.get("Object"))
.and_then(|v| v.as_object())
.with_context(|| format!("Indexes collection missing in table info: {info_json:#?}"))?;
let indexes = info_json["Object"]["indexes"]["Object"]
.as_object()
.unwrap_or_else(|| panic!("Indexes collection missing in table info: {info_json:#?}"));
let definition = indexes
.get(index_name)
.and_then(|definition| definition.get("Strand"))
.and_then(|v| v.as_str())
.with_context(|| format!("Index definition not found in table info: {info_json:#?}"))?;
.unwrap_or_else(|| panic!("Index definition not found in table info: {info_json:#?}"));
let dimension_part = definition
.split("DIMENSION")
.nth(1)
.with_context(|| "Index definition missing DIMENSION clause".to_string())?;
.expect("Index definition missing DIMENSION clause");
let dimension_token = dimension_part
.split_whitespace()
.next()
.with_context(|| "Dimension value missing in definition".to_string())?
.expect("Dimension value missing in definition")
.trim_end_matches(';');
dimension_token
.parse::<u32>()
.with_context(|| "Dimension value is not a valid number".to_string())
}
async fn simulate_reembedding(
db: &SurrealDbClient,
target_dimension: usize,
initial_chunk: TextChunk,
) -> anyhow::Result<()> {
db.query(
"REMOVE INDEX IF EXISTS idx_embedding_text_chunk_embedding ON TABLE text_chunk_embedding;",
)
.await
.with_context(|| "remove index".to_string())?;
let define_index_query = format!(
"DEFINE INDEX idx_embedding_text_chunk_embedding ON TABLE text_chunk_embedding FIELDS embedding HNSW DIMENSION {target_dimension};"
);
db.query(define_index_query)
.await
.with_context(|| "Re-defining index should succeed".to_string())?;
let new_embedding = vec![0.5; target_dimension];
let sql = "UPSERT type::thing('text_chunk_embedding', $id) SET chunk_id = type::thing('text_chunk', $id), embedding = $embedding, user_id = $user_id;";
db.client
.query(sql)
.bind(("id", initial_chunk.id.clone()))
.bind(("user_id", initial_chunk.user_id.clone()))
.bind(("embedding", new_embedding))
.await
.with_context(|| "upsert embedding".to_string())?;
Ok(())
.expect("Dimension value is not a valid number")
}
#[tokio::test]
async fn test_settings_initialization() -> anyhow::Result<()> {
async fn test_settings_initialization() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
// Test initialization of system settings
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
.expect("Failed to apply migrations");
let settings = SystemSettings::get_current(&db)
.await
.with_context(|| "Failed to get system settings".to_string())?;
.expect("Failed to get system settings");
// Verify initial state after initialization
assert_eq!(settings.id, "current");
assert!(settings.registrations_enabled);
assert!(!settings.require_email_verification);
assert_eq!(settings.registrations_enabled, true);
assert_eq!(settings.require_email_verification, false);
assert_eq!(settings.query_model, "gpt-4o-mini");
assert_eq!(settings.processing_model, "gpt-4o-mini");
assert_eq!(settings.image_processing_model, "gpt-4o-mini");
assert!(!settings.ingestion_system_prompt.contains("entity\"\n6."));
assert!(settings.ingestion_system_prompt.contains("related entity."));
// Dont test these for now, having a hard time getting the formatting exactly the same
// assert_eq!(
// settings.query_system_prompt,
// crate::storage::types::system_prompts::DEFAULT_QUERY_SYSTEM_PROMPT
// );
// assert_eq!(
// settings.ingestion_system_prompt,
// crate::storage::types::system_prompts::DEFAULT_INGRESS_ANALYSIS_SYSTEM_PROMPT
// );
// Test idempotency - ensure calling it again doesn't change anything
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
.expect("Failed to apply migrations");
let settings_again = SystemSettings::get_current(&db)
.await
.with_context(|| "Failed to get settings after initialization".to_string())?;
.expect("Failed to get settings after initialization");
assert_eq!(settings.id, settings_again.id);
assert_eq!(
@@ -362,52 +210,49 @@ mod tests {
settings.require_email_verification,
settings_again.require_email_verification
);
Ok(())
}
#[tokio::test]
async fn test_get_current_settings() -> anyhow::Result<()> {
async fn test_get_current_settings() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
// Initialize settings
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
.expect("Failed to apply migrations");
// Test get_current method
let settings = SystemSettings::get_current(&db)
.await
.with_context(|| "Failed to get current settings".to_string())?;
.expect("Failed to get current settings");
assert_eq!(settings.id, "current");
assert!(settings.registrations_enabled);
assert!(!settings.require_email_verification);
Ok(())
assert_eq!(settings.registrations_enabled, true);
assert_eq!(settings.require_email_verification, false);
}
#[tokio::test]
async fn test_update_settings() -> anyhow::Result<()> {
async fn test_update_settings() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
// Initialize settings
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
.expect("Failed to apply migrations");
// Create updated settings
let mut updated_settings = SystemSettings::get_current(&db)
.await
.with_context(|| "get_current".to_string())?;
let mut updated_settings = SystemSettings::get_current(&db).await.unwrap();
updated_settings.id = "current".to_string();
updated_settings.registrations_enabled = false;
updated_settings.require_email_verification = true;
updated_settings.query_model = "gpt-4".to_string();
@@ -415,32 +260,31 @@ mod tests {
// Test update method
let result = SystemSettings::update(&db, updated_settings)
.await
.with_context(|| "Failed to update settings".to_string())?;
.expect("Failed to update settings");
assert_eq!(result.id, "current");
assert!(!result.registrations_enabled);
assert!(result.require_email_verification);
assert_eq!(result.registrations_enabled, false);
assert_eq!(result.require_email_verification, true);
assert_eq!(result.query_model, "gpt-4");
// Verify changes persisted by getting current settings
let current = SystemSettings::get_current(&db)
.await
.with_context(|| "Failed to get current settings after update".to_string())?;
.expect("Failed to get current settings after update");
assert!(!current.registrations_enabled);
assert!(current.require_email_verification);
assert_eq!(current.registrations_enabled, false);
assert_eq!(current.require_email_verification, true);
assert_eq!(current.query_model, "gpt-4");
Ok(())
}
#[tokio::test]
async fn test_get_current_nonexistent() -> anyhow::Result<()> {
async fn test_get_current_nonexistent() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
// Don't initialize settings and try to get them
let result = SystemSettings::get_current(&db).await;
@@ -450,240 +294,21 @@ mod tests {
Err(AppError::NotFound(_)) => {
// Expected error
}
Err(e) => anyhow::bail!("Expected NotFound error, got: {e:?}"),
Ok(_) => anyhow::bail!("Expected error but got Ok"),
Err(e) => panic!("Expected NotFound error, got: {:?}", e),
Ok(_) => panic!("Expected error but got Ok"),
}
Ok(())
}
#[tokio::test]
async fn test_update_rejects_zero_embedding_dimensions() -> anyhow::Result<()> {
let db = SurrealDbClient::memory("test_ns", &Uuid::new_v4().to_string())
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
let mut invalid_settings = SystemSettings::get_current(&db)
.await
.with_context(|| "Failed to get system settings".to_string())?;
invalid_settings.embedding_dimensions = 0;
let result = SystemSettings::update(&db, invalid_settings).await;
assert!(matches!(result, Err(AppError::Validation(_))));
Ok(())
}
#[tokio::test]
async fn test_patch_updates_without_cloning_full_settings() -> anyhow::Result<()> {
let db = SurrealDbClient::memory("test_ns", &Uuid::new_v4().to_string())
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
let updated = SystemSettingsPatch {
registrations_enabled: Some(false),
..Default::default()
}
.apply(&db)
.await
.with_context(|| "Failed to patch settings".to_string())?;
assert!(!updated.registrations_enabled);
Ok(())
}
#[tokio::test]
async fn test_patch_leaves_unmentioned_fields_unchanged() -> anyhow::Result<()> {
let db = SurrealDbClient::memory("test_ns", &Uuid::new_v4().to_string())
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
let original = SystemSettings::get_current(&db)
.await
.with_context(|| "Failed to get system settings".to_string())?;
let sentinel = "custom-query-prompt-sentinel".to_string();
let patched = SystemSettingsPatch {
query_system_prompt: Some(sentinel.clone()),
..Default::default()
}
.apply(&db)
.await
.with_context(|| "Failed to patch query prompt".to_string())?;
assert_eq!(patched.query_system_prompt, sentinel);
assert_eq!(
patched.ingestion_system_prompt,
original.ingestion_system_prompt
);
assert_eq!(patched.query_model, original.query_model);
assert_eq!(
patched.registrations_enabled,
original.registrations_enabled
);
Ok(())
}
#[tokio::test]
async fn test_update_rejects_empty_model_name() -> anyhow::Result<()> {
let db = SurrealDbClient::memory("test_ns", &Uuid::new_v4().to_string())
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
let mut invalid_settings = SystemSettings::get_current(&db)
.await
.with_context(|| "Failed to get system settings".to_string())?;
invalid_settings.query_model = " ".to_string();
let result = SystemSettings::update(&db, invalid_settings).await;
assert!(matches!(result, Err(AppError::Validation(_))));
Ok(())
}
#[tokio::test]
async fn test_update_normalizes_record_id() -> anyhow::Result<()> {
let db = SurrealDbClient::memory("test_ns", &Uuid::new_v4().to_string())
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
let mut settings = SystemSettings::get_current(&db)
.await
.with_context(|| "Failed to get system settings".to_string())?;
settings.id = "wrong-id".to_string();
let updated = SystemSettings::update(&db, settings)
.await
.with_context(|| "Failed to update settings".to_string())?;
assert_eq!(updated.id, SystemSettings::RECORD_ID);
Ok(())
}
#[tokio::test]
async fn test_update_preserves_embedding_backend() -> anyhow::Result<()> {
use crate::utils::embedding::EmbeddingProvider;
let db = SurrealDbClient::memory("test_ns", &Uuid::new_v4().to_string())
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
let provider = EmbeddingProvider::new_hashed(384)
.with_context(|| "Failed to create hashed embedding provider".to_string())?;
SystemSettings::sync_from_embedding_provider(&db, &provider)
.await
.with_context(|| "Failed to sync embedding provider".to_string())?;
let synced = SystemSettings::get_current(&db)
.await
.with_context(|| "Failed to get synced settings".to_string())?;
assert_eq!(synced.embedding_backend, Some(EmbeddingBackend::Hashed));
let mut tampered = synced;
tampered.embedding_backend = Some(EmbeddingBackend::OpenAI);
let updated = SystemSettings::update(&db, tampered)
.await
.with_context(|| "Failed to update settings".to_string())?;
assert_eq!(updated.embedding_backend, Some(EmbeddingBackend::Hashed));
Ok(())
}
#[tokio::test]
async fn test_sync_from_embedding_provider_updates_mismatched_settings() -> anyhow::Result<()> {
use crate::utils::embedding::EmbeddingProvider;
let db = SurrealDbClient::memory("test_ns", &Uuid::new_v4().to_string())
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
let provider = EmbeddingProvider::new_hashed(384)
.with_context(|| "Failed to create hashed embedding provider".to_string())?;
let (settings, changed) = SystemSettings::sync_from_embedding_provider(&db, &provider)
.await
.with_context(|| "Failed to sync embedding provider".to_string())?;
assert!(changed);
assert_eq!(settings.embedding_backend, Some(EmbeddingBackend::Hashed));
assert_eq!(settings.embedding_dimensions, 384);
let persisted = SystemSettings::get_current(&db)
.await
.with_context(|| "Failed to reload synced settings".to_string())?;
assert_eq!(persisted.embedding_backend, Some(EmbeddingBackend::Hashed));
assert_eq!(persisted.embedding_dimensions, 384);
Ok(())
}
#[tokio::test]
async fn test_sync_from_embedding_provider_is_noop_when_already_synced() -> anyhow::Result<()> {
use crate::utils::embedding::EmbeddingProvider;
let db = SurrealDbClient::memory("test_ns", &Uuid::new_v4().to_string())
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
let provider = EmbeddingProvider::new_hashed(384)
.with_context(|| "Failed to create hashed embedding provider".to_string())?;
SystemSettings::sync_from_embedding_provider(&db, &provider)
.await
.with_context(|| "Failed to initial sync".to_string())?;
let (_, changed) = SystemSettings::sync_from_embedding_provider(&db, &provider)
.await
.with_context(|| "Failed to repeat sync".to_string())?;
assert!(!changed);
Ok(())
}
#[tokio::test]
async fn test_sync_rejects_provider_dimension_above_u32_max() -> anyhow::Result<()> {
use crate::utils::embedding::EmbeddingProvider;
let db = SurrealDbClient::memory("test_ns", &Uuid::new_v4().to_string())
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
db.apply_migrations()
.await
.with_context(|| "Failed to apply migrations".to_string())?;
let provider = EmbeddingProvider::new_hashed((u32::MAX as usize) + 1)
.with_context(|| "Failed to create oversized hashed provider".to_string())?;
let result = SystemSettings::sync_from_embedding_provider(&db, &provider).await;
assert!(matches!(result, Err(AppError::Validation(_))));
Ok(())
}
#[tokio::test]
async fn test_migration_after_changing_embedding_length() -> anyhow::Result<()> {
async fn test_migration_after_changing_embedding_length() {
let db = SurrealDbClient::memory("test", &Uuid::new_v4().to_string())
.await
.with_context(|| "Failed to start DB".to_string())?;
.expect("Failed to start DB");
// Apply initial migrations. This sets up the text_chunk index with DIMENSION 1536.
db.apply_migrations()
.await
.with_context(|| "Initial migration failed".to_string())?;
.expect("Initial migration failed");
let initial_chunk = TextChunk::new(
"source1".into(),
@@ -693,11 +318,43 @@ mod tests {
TextChunk::store_with_embedding(initial_chunk.clone(), vec![0.1; 1536], &db)
.await
.with_context(|| "Failed to store initial chunk with embedding".to_string())?;
.expect("Failed to store initial chunk with embedding");
async fn simulate_reembedding(
db: &SurrealDbClient,
target_dimension: usize,
initial_chunk: TextChunk,
) {
db.query(
"REMOVE INDEX IF EXISTS idx_embedding_text_chunk_embedding ON TABLE text_chunk_embedding;",
)
.await
.unwrap();
let define_index_query = format!(
"DEFINE INDEX idx_embedding_text_chunk_embedding ON TABLE text_chunk_embedding FIELDS embedding HNSW DIMENSION {};",
target_dimension
);
db.query(define_index_query)
.await
.expect("Re-defining index should succeed");
let new_embedding = vec![0.5; target_dimension];
let sql = "UPSERT type::thing('text_chunk_embedding', $id) SET chunk_id = type::thing('text_chunk', $id), embedding = $embedding, user_id = $user_id;";
let update_result = db
.client
.query(sql)
.bind(("id", initial_chunk.id.clone()))
.bind(("user_id", initial_chunk.user_id.clone()))
.bind(("embedding", new_embedding))
.await;
assert!(update_result.is_ok());
}
// Re-embed with the existing configured dimension to ensure migrations remain idempotent.
let target_dimension = 1536usize;
simulate_reembedding(&db, target_dimension, initial_chunk).await?;
simulate_reembedding(&db, target_dimension, initial_chunk).await;
let migration_result = db.apply_migrations().await;
@@ -706,38 +363,34 @@ mod tests {
"Migrations should not fail: {:?}",
migration_result.err()
);
Ok(())
}
#[tokio::test]
async fn test_should_change_embedding_length_on_indexes_when_switching_length(
) -> anyhow::Result<()> {
use crate::utils::embedding::EmbeddingProvider;
async fn test_should_change_embedding_length_on_indexes_when_switching_length() {
let db = SurrealDbClient::memory("test", &Uuid::new_v4().to_string())
.await
.with_context(|| "Failed to start DB".to_string())?;
.expect("Failed to start DB");
// Apply initial migrations. This sets up the text_chunk index with DIMENSION 1536.
db.apply_migrations()
.await
.with_context(|| "Initial migration failed".to_string())?;
.expect("Initial migration failed");
let mut current_settings = SystemSettings::get_current(&db)
.await
.with_context(|| "Failed to load current settings".to_string())?;
.expect("Failed to load current settings");
// Ensure runtime indexes exist with the current embedding dimension so INFO queries succeed.
ensure_runtime(&db, current_settings.embedding_dimensions as usize)
ensure_runtime_indexes(&db, current_settings.embedding_dimensions as usize)
.await
.with_context(|| "failed to build runtime indexes".to_string())?;
.expect("failed to build runtime indexes");
let initial_chunk_dimension = get_hnsw_index_dimension(
&db,
"text_chunk_embedding",
"idx_embedding_text_chunk_embedding",
)
.await?;
.await;
assert_eq!(
initial_chunk_dimension, current_settings.embedding_dimensions,
@@ -752,37 +405,34 @@ mod tests {
let updated_settings = SystemSettings::update(&db, current_settings)
.await
.with_context(|| "Failed to update settings".to_string())?;
.expect("Failed to update settings");
assert_eq!(
updated_settings.embedding_dimensions, new_dimension,
"Settings should reflect the new embedding dimension"
);
let provider = EmbeddingProvider::new_hashed(new_dimension as usize)
.map_err(|e| anyhow::anyhow!("{e}"))?;
let openai_client = Client::new();
TextChunk::update_all_embeddings(&db, &provider)
TextChunk::update_all_embeddings(&db, &openai_client, &new_model, new_dimension)
.await
.with_context(|| "TextChunk re-embedding should succeed on fresh DB".to_string())?;
KnowledgeEntity::update_all_embeddings(&db, &provider)
.expect("TextChunk re-embedding should succeed on fresh DB");
KnowledgeEntity::update_all_embeddings(&db, &openai_client, &new_model, new_dimension)
.await
.with_context(|| {
"KnowledgeEntity re-embedding should succeed on fresh DB".to_string()
})?;
.expect("KnowledgeEntity re-embedding should succeed on fresh DB");
let text_chunk_dimension = get_hnsw_index_dimension(
&db,
"text_chunk_embedding",
"idx_embedding_text_chunk_embedding",
)
.await?;
.await;
let knowledge_dimension = get_hnsw_index_dimension(
&db,
"knowledge_entity_embedding",
"idx_embedding_knowledge_entity_embedding",
)
.await?;
.await;
assert_eq!(
text_chunk_dimension, new_dimension,
@@ -795,11 +445,10 @@ mod tests {
let persisted_settings = SystemSettings::get_current(&db)
.await
.with_context(|| "Failed to reload updated settings".to_string())?;
.expect("Failed to reload updated settings");
assert_eq!(
persisted_settings.embedding_dimensions, new_dimension,
"Settings should persist new embedding dimension"
);
Ok(())
}
}
File diff suppressed because it is too large Load Diff
+200 -195
View File
@@ -1,11 +1,7 @@
use surrealdb::RecordId;
use crate::storage::types::text_chunk::TextChunk;
use crate::{
error::AppError,
storage::{db::SurrealDbClient, indexes::hnsw_index_redefine_transaction_sql},
stored_object,
};
use crate::{error::AppError, storage::db::SurrealDbClient, stored_object};
stored_object!(TextChunkEmbedding, "text_chunk_embedding", {
/// Record link to the owning text_chunk
@@ -27,42 +23,33 @@ impl TextChunkEmbedding {
db: &SurrealDbClient,
dimension: usize,
) -> Result<(), AppError> {
let query = hnsw_index_redefine_transaction_sql(
"idx_embedding_text_chunk_embedding",
Self::table_name(),
dimension,
let query = format!(
"BEGIN TRANSACTION;
REMOVE INDEX IF EXISTS idx_embedding_text_chunk_embedding ON TABLE {table};
DEFINE INDEX idx_embedding_text_chunk_embedding ON TABLE {table} FIELDS embedding HNSW DIMENSION {dimension};
COMMIT TRANSACTION;",
table = Self::table_name(),
);
let res = db.client.query(query).await.map_err(AppError::from)?;
res.check().map_err(AppError::from)?;
let res = db.client.query(query).await.map_err(AppError::Database)?;
res.check().map_err(AppError::Database)?;
Ok(())
}
/// Validates that an embedding vector matches the configured HNSW dimension.
#[allow(clippy::result_large_err)]
pub fn validate_dimension(embedding: &[f32], expected: usize) -> Result<(), AppError> {
if embedding.len() != expected {
return Err(AppError::Validation(format!(
"embedding dimension mismatch: got {}, expected {expected}",
embedding.len()
)));
}
Ok(())
}
/// Create a new text chunk embedding.
/// Create a new text chunk embedding
///
/// The embedding record id equals `chunk_id` so each chunk has at most one embedding row.
/// `chunk_id` is the **key** part of the text_chunk id (e.g. the UUID), not "text_chunk:uuid".
#[must_use]
/// `chunk_id` is the **key** part of the text_chunk id (e.g. the UUID),
/// not "text_chunk:uuid".
pub fn new(chunk_id: &str, source_id: String, embedding: Vec<f32>, user_id: String) -> Self {
let now = Utc::now();
Self {
id: chunk_id.to_owned(),
// NOTE: `stored_object!` macro defines `id` as `String`
id: uuid::Uuid::new_v4().to_string(),
created_at: now,
updated_at: now,
// Create a record<text_chunk> link: text_chunk:<chunk_id>
chunk_id: RecordId::from_table_key(TextChunk::table_name(), chunk_id),
source_id,
embedding,
@@ -85,9 +72,9 @@ impl TextChunkEmbedding {
.query(query)
.bind(("chunk_id", chunk_id.clone()))
.await
.map_err(AppError::from)?;
.map_err(AppError::Database)?;
let embeddings: Vec<Self> = result.take(0).map_err(AppError::from)?;
let embeddings: Vec<Self> = result.take(0).map_err(AppError::Database)?;
Ok(embeddings.into_iter().next())
}
@@ -106,32 +93,57 @@ impl TextChunkEmbedding {
.query(query)
.bind(("chunk_id", chunk_id.clone()))
.await
.map_err(AppError::from)?
.map_err(AppError::Database)?
.check()
.map_err(AppError::from)?;
.map_err(AppError::Database)?;
Ok(())
}
/// Delete all embeddings that belong to chunks with a given `source_id`
///
/// This uses the denormalized `source_id` on the embedding table.
/// This uses a subquery to the `text_chunk` table:
///
/// DELETE FROM text_chunk_embedding
/// WHERE chunk_id IN (SELECT id FROM text_chunk WHERE source_id = $source_id)
pub async fn delete_by_source_id(
source_id: &str,
db: &SurrealDbClient,
) -> Result<(), AppError> {
let query = format!(
"DELETE FROM {} WHERE source_id = $source_id",
Self::table_name()
#[allow(clippy::missing_docs_in_private_items)]
#[derive(Deserialize)]
struct IdRow {
id: RecordId,
}
let ids_query = format!(
"SELECT id FROM {} WHERE source_id = $source_id",
TextChunk::table_name()
);
db.client
.query(query)
let mut res = db
.client
.query(ids_query)
.bind(("source_id", source_id.to_owned()))
.await
.map_err(AppError::from)?
.map_err(AppError::Database)?;
let ids: Vec<IdRow> = res.take(0).map_err(AppError::Database)?;
if ids.is_empty() {
return Ok(());
}
let delete_query = format!(
"DELETE FROM {} WHERE chunk_id IN $chunk_ids",
Self::table_name()
);
db.client
.query(delete_query)
.bind((
"chunk_ids",
ids.into_iter().map(|row| row.id).collect::<Vec<_>>(),
))
.await
.map_err(AppError::Database)?
.check()
.map_err(AppError::from)?;
.map_err(AppError::Database)?;
Ok(())
}
@@ -139,20 +151,33 @@ impl TextChunkEmbedding {
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use anyhow::{self, Context};
use super::*;
use crate::storage::db::SurrealDbClient;
use crate::test_utils::{prepare_text_chunk_test_db, setup_test_db};
use surrealdb::Value as SurrealValue;
use uuid::Uuid;
/// Helper to create an in-memory DB and apply migrations
async fn setup_test_db() -> SurrealDbClient {
let namespace = "test_ns";
let database = Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, &database)
.await
.expect("Failed to start in-memory surrealdb");
db.apply_migrations()
.await
.expect("Failed to apply migrations");
db
}
/// Helper: create a text_chunk with a known key, return its RecordId
async fn create_text_chunk_with_id(
db: &SurrealDbClient,
key: &str,
source_id: &str,
user_id: &str,
) -> anyhow::Result<RecordId> {
) -> RecordId {
let chunk = TextChunk {
id: key.to_owned(),
created_at: Utc::now(),
@@ -164,62 +189,23 @@ mod tests {
db.store_item(chunk)
.await
.with_context(|| "Failed to create text_chunk".to_string())?;
.expect("Failed to create text_chunk");
Ok(RecordId::from_table_key(TextChunk::table_name(), key))
}
async fn get_idx_sql(db: &SurrealDbClient) -> anyhow::Result<String> {
let mut info_res = db
.client
.query("INFO FOR TABLE text_chunk_embedding;")
.await
.with_context(|| "info query failed".to_string())?;
let info: SurrealValue = info_res
.take(0)
.with_context(|| "failed to take info result".to_string())?;
let info_json: serde_json::Value = serde_json::to_value(info)
.with_context(|| "failed to convert info to json".to_string())?;
let idx_sql = info_json
.get("Object")
.and_then(|v| v.get("indexes"))
.and_then(|v| v.get("Object"))
.and_then(|v| v.get("idx_embedding_text_chunk_embedding"))
.and_then(|v| v.get("Strand"))
.and_then(|v| v.as_str())
.unwrap_or_default()
.to_string();
Ok(idx_sql)
}
#[test]
fn new_uses_chunk_id_as_record_id() {
let emb = TextChunkEmbedding::new(
"chunk-abc",
"source-1".to_owned(),
vec![0.1, 0.2],
"user-1".to_owned(),
);
assert_eq!(emb.id, "chunk-abc");
}
#[test]
fn validate_dimension_rejects_mismatch() {
let err = TextChunkEmbedding::validate_dimension(&[0.1, 0.2, 0.3], 2)
.expect_err("expected dimension mismatch");
assert!(matches!(err, AppError::Validation(_)));
RecordId::from_table_key(TextChunk::table_name(), key)
}
#[tokio::test]
async fn test_create_and_get_by_chunk_id() -> anyhow::Result<()> {
let db = prepare_text_chunk_test_db(3).await?;
async fn test_create_and_get_by_chunk_id() {
let db = setup_test_db().await;
let user_id = "user_a";
let chunk_key = "chunk-123";
let source_id = "source-1";
let chunk_rid = create_text_chunk_with_id(&db, chunk_key, source_id, user_id).await?;
// 1) Create a text_chunk with a known key
let chunk_rid = create_text_chunk_with_id(&db, chunk_key, source_id, user_id).await;
// 2) Create and store an embedding for that chunk
let embedding_vec = vec![0.1_f32, 0.2, 0.3];
let emb = TextChunkEmbedding::new(
chunk_key,
@@ -228,31 +214,41 @@ mod tests {
user_id.to_string(),
);
db.upsert_item(emb)
TextChunkEmbedding::redefine_hnsw_index(&db, emb.embedding.len())
.await
.with_context(|| "Failed to store embedding".to_string())?;
.expect("Failed to redefine index length");
let _: Option<TextChunkEmbedding> = db
.client
.create(TextChunkEmbedding::table_name())
.content(emb)
.await
.expect("Failed to store embedding")
.take()
.expect("Failed to deserialize stored embedding");
// 3) Fetch it via get_by_chunk_id
let fetched = TextChunkEmbedding::get_by_chunk_id(&chunk_rid, &db)
.await
.with_context(|| "Failed to get embedding by chunk_id".to_string())?
.with_context(|| "Expected an embedding to be found".to_string())?;
.expect("Failed to get embedding by chunk_id");
assert!(fetched.is_some(), "Expected an embedding to be found");
let fetched = fetched.unwrap();
assert_eq!(fetched.id, chunk_key);
assert_eq!(fetched.user_id, user_id);
assert_eq!(fetched.chunk_id, chunk_rid);
assert_eq!(fetched.embedding, embedding_vec);
Ok(())
}
#[tokio::test]
async fn test_delete_by_chunk_id() -> anyhow::Result<()> {
let db = prepare_text_chunk_test_db(3).await?;
async fn test_delete_by_chunk_id() {
let db = setup_test_db().await;
let user_id = "user_b";
let chunk_key = "chunk-delete";
let source_id = "source-del";
let chunk_rid = create_text_chunk_with_id(&db, chunk_key, source_id, user_id).await?;
let chunk_rid = create_text_chunk_with_id(&db, chunk_key, source_id, user_id).await;
let emb = TextChunkEmbedding::new(
chunk_key,
@@ -261,171 +257,180 @@ mod tests {
user_id.to_string(),
);
db.upsert_item(emb)
TextChunkEmbedding::redefine_hnsw_index(&db, emb.embedding.len())
.await
.with_context(|| "Failed to store embedding".to_string())?;
.expect("Failed to redefine index length");
let _: Option<TextChunkEmbedding> = db
.client
.create(TextChunkEmbedding::table_name())
.content(emb)
.await
.expect("Failed to store embedding")
.take()
.expect("Failed to deserialize stored embedding");
// Ensure it exists
let existing = TextChunkEmbedding::get_by_chunk_id(&chunk_rid, &db)
.await
.with_context(|| "Failed to get embedding before delete".to_string())?;
.expect("Failed to get embedding before delete");
assert!(existing.is_some(), "Embedding should exist before delete");
// Delete by chunk_id
TextChunkEmbedding::delete_by_chunk_id(&chunk_rid, &db)
.await
.with_context(|| "Failed to delete by chunk_id".to_string())?;
.expect("Failed to delete by chunk_id");
// Ensure it no longer exists
let after = TextChunkEmbedding::get_by_chunk_id(&chunk_rid, &db)
.await
.with_context(|| "Failed to get embedding after delete".to_string())?;
.expect("Failed to get embedding after delete");
assert!(after.is_none(), "Embedding should have been deleted");
Ok(())
}
#[tokio::test]
async fn test_delete_by_source_id() -> anyhow::Result<()> {
let db = prepare_text_chunk_test_db(1).await?;
async fn test_delete_by_source_id() {
let db = setup_test_db().await;
let user_id = "user_c";
let source_id = "shared-source";
let other_source = "other-source";
let chunk1_rid = create_text_chunk_with_id(&db, "chunk-s1", source_id, user_id).await?;
let chunk2_rid = create_text_chunk_with_id(&db, "chunk-s2", source_id, user_id).await?;
let chunk_other_rid =
create_text_chunk_with_id(&db, "chunk-other", other_source, user_id).await?;
// Two chunks with the same source_id
let chunk1_rid = create_text_chunk_with_id(&db, "chunk-s1", source_id, user_id).await;
let chunk2_rid = create_text_chunk_with_id(&db, "chunk-s2", source_id, user_id).await;
for (key, src, vec) in [
("chunk-s1", source_id, vec![0.1]),
("chunk-s2", source_id, vec![0.2]),
("chunk-other", other_source, vec![0.3]),
] {
let emb = TextChunkEmbedding::new(key, src.to_string(), vec, user_id.to_string());
db.upsert_item(emb)
// One chunk with a different source_id
let chunk_other_rid =
create_text_chunk_with_id(&db, "chunk-other", other_source, user_id).await;
// Create embeddings for all three
let emb1 = TextChunkEmbedding::new(
"chunk-s1",
source_id.to_string(),
vec![0.1],
user_id.to_string(),
);
let emb2 = TextChunkEmbedding::new(
"chunk-s2",
source_id.to_string(),
vec![0.2],
user_id.to_string(),
);
let emb3 = TextChunkEmbedding::new(
"chunk-other",
other_source.to_string(),
vec![0.3],
user_id.to_string(),
);
// Update length on index
TextChunkEmbedding::redefine_hnsw_index(&db, emb1.embedding.len())
.await
.with_context(|| format!("store embedding for {key}"))?;
.expect("Failed to redefine index length");
for emb in [emb1, emb2, emb3] {
let _: Option<TextChunkEmbedding> = db
.client
.create(TextChunkEmbedding::table_name())
.content(emb)
.await
.expect("Failed to store embedding")
.take()
.expect("Failed to deserialize stored embedding");
}
// Sanity check: they all exist
assert!(TextChunkEmbedding::get_by_chunk_id(&chunk1_rid, &db)
.await
.with_context(|| "get chunk1".to_string())?
.unwrap()
.is_some());
assert!(TextChunkEmbedding::get_by_chunk_id(&chunk2_rid, &db)
.await
.with_context(|| "get chunk2".to_string())?
.unwrap()
.is_some());
assert!(TextChunkEmbedding::get_by_chunk_id(&chunk_other_rid, &db)
.await
.with_context(|| "get chunk_other".to_string())?
.unwrap()
.is_some());
// Delete embeddings by source_id (shared-source)
TextChunkEmbedding::delete_by_source_id(source_id, &db)
.await
.with_context(|| "Failed to delete by source_id".to_string())?;
.expect("Failed to delete by source_id");
// Chunks from shared-source should have no embeddings
assert!(TextChunkEmbedding::get_by_chunk_id(&chunk1_rid, &db)
.await
.with_context(|| "check chunk1".to_string())?
.unwrap()
.is_none());
assert!(TextChunkEmbedding::get_by_chunk_id(&chunk2_rid, &db)
.await
.with_context(|| "check chunk2".to_string())?
.unwrap()
.is_none());
// The other chunk should still have its embedding
assert!(TextChunkEmbedding::get_by_chunk_id(&chunk_other_rid, &db)
.await
.with_context(|| "check chunk_other".to_string())?
.unwrap()
.is_some());
Ok(())
}
#[tokio::test]
async fn test_upsert_replaces_existing_embedding_row() -> anyhow::Result<()> {
let db = prepare_text_chunk_test_db(3).await?;
let user_id = "user-upsert";
let source_id = "source-upsert";
let chunk_key = "chunk-upsert";
create_text_chunk_with_id(&db, chunk_key, source_id, user_id).await?;
let initial = TextChunkEmbedding::new(
chunk_key,
source_id.to_owned(),
vec![1.0_f32, 0.0, 0.0],
user_id.to_owned(),
);
db.upsert_item(initial)
.await
.with_context(|| "initial upsert".to_string())?;
let replacement = TextChunkEmbedding::new(
chunk_key,
source_id.to_owned(),
vec![0.0, 1.0, 0.0],
user_id.to_owned(),
);
db.upsert_item(replacement)
.await
.with_context(|| "upsert replacement embedding".to_string())?;
let chunk_rid = RecordId::from_table_key(TextChunk::table_name(), chunk_key);
let rows: Vec<TextChunkEmbedding> = db
.client
.query(format!(
"SELECT * FROM {} WHERE chunk_id = $chunk_id",
TextChunkEmbedding::table_name()
))
.bind(("chunk_id", chunk_rid))
.await
.with_context(|| "count embeddings".to_string())?
.take(0)
.with_context(|| "take embeddings".to_string())?;
assert_eq!(rows.len(), 1);
let row = rows.first().expect("expected one embedding row");
assert_eq!(row.id, chunk_key);
assert_eq!(row.embedding, vec![0.0, 1.0, 0.0]);
Ok(())
}
#[tokio::test]
async fn test_redefine_hnsw_index_updates_dimension() -> anyhow::Result<()> {
let db = setup_test_db().await?;
async fn test_redefine_hnsw_index_updates_dimension() {
let db = setup_test_db().await;
// Change the index dimension from default (1536) to a smaller test value.
TextChunkEmbedding::redefine_hnsw_index(&db, 8)
.await
.with_context(|| "failed to redefine index".to_string())?;
.expect("failed to redefine index");
let idx_sql = get_idx_sql(&db).await?;
let mut info_res = db
.client
.query("INFO FOR TABLE text_chunk_embedding;")
.await
.expect("info query failed");
let info: SurrealValue = info_res.take(0).expect("failed to take info result");
let info_json: serde_json::Value =
serde_json::to_value(info).expect("failed to convert info to json");
let idx_sql = info_json["Object"]["indexes"]["Object"]
["idx_embedding_text_chunk_embedding"]["Strand"]
.as_str()
.unwrap_or_default();
assert!(
idx_sql.contains("DIMENSION 8"),
"expected index definition to contain new dimension, got: {idx_sql}"
);
assert!(
idx_sql.contains("DIST COSINE"),
"expected index definition to use cosine distance, got: {idx_sql}"
);
Ok(())
}
#[tokio::test]
async fn test_redefine_hnsw_index_is_idempotent() -> anyhow::Result<()> {
let db = setup_test_db().await?;
async fn test_redefine_hnsw_index_is_idempotent() {
let db = setup_test_db().await;
TextChunkEmbedding::redefine_hnsw_index(&db, 4)
.await
.with_context(|| "first redefine failed".to_string())?;
.expect("first redefine failed");
TextChunkEmbedding::redefine_hnsw_index(&db, 4)
.await
.with_context(|| "second redefine failed".to_string())?;
.expect("second redefine failed");
let idx_sql = get_idx_sql(&db).await?;
let mut info_res = db
.client
.query("INFO FOR TABLE text_chunk_embedding;")
.await
.expect("info query failed");
let info: SurrealValue = info_res.take(0).expect("failed to take info result");
let info_json: serde_json::Value =
serde_json::to_value(info).expect("failed to convert info to json");
let idx_sql = info_json["Object"]["indexes"]["Object"]
["idx_embedding_text_chunk_embedding"]["Strand"]
.as_str()
.unwrap_or_default();
assert!(
idx_sql.contains("DIMENSION 4"),
"expected index definition to retain dimension 4, got: {idx_sql}"
);
Ok(())
}
}
+32 -310
View File
@@ -1,8 +1,4 @@
use std::collections::{HashMap, HashSet};
use std::str::FromStr;
use surrealdb::opt::PatchOp;
use surrealdb::RecordId;
use uuid::Uuid;
use crate::{error::AppError, storage::db::SurrealDbClient, stored_object};
@@ -73,7 +69,6 @@ stored_object!(TextContent, "text_content", {
});
impl TextContent {
#[must_use]
pub fn new(
text: String,
context: Option<String>,
@@ -105,7 +100,7 @@ impl TextContent {
) -> Result<(), AppError> {
let now = Utc::now();
let updated: Option<Self> = db
let _res: Option<Self> = db
.update((Self::table_name(), id))
.patch(PatchOp::replace("/context", context))
.patch(PatchOp::replace("/category", category))
@@ -114,12 +109,7 @@ impl TextContent {
"/updated_at",
surrealdb::Datetime::from(now),
))
.await
.map_err(AppError::from)?;
if updated.is_none() {
return Err(AppError::NotFound(format!("text content {id} not found")));
}
.await?;
Ok(())
}
@@ -137,10 +127,9 @@ impl TextContent {
.bind(("table_name", TextContent::table_name()))
.bind(("file_id", file_id.to_owned()))
.bind(("exclude_id", exclude_id.to_owned()))
.await
.map_err(AppError::from)?;
.await?;
let existing: Option<surrealdb::sql::Thing> = response.take(0).map_err(AppError::from)?;
let existing: Option<surrealdb::sql::Thing> = response.take(0)?;
Ok(existing.is_some())
}
@@ -151,8 +140,7 @@ impl TextContent {
user_id: &str,
limit: usize,
) -> Result<Vec<TextContentSearchResult>, AppError> {
let sql = format!(
r#"
let sql = r#"
SELECT
*,
search::highlight('<b>', '</b>', 0) AS highlighted_text,
@@ -169,7 +157,7 @@ impl TextContent {
IF search::score(4) != NONE THEN search::score(4) ELSE 0 END +
IF search::score(5) != NONE THEN search::score(5) ELSE 0 END
) AS score
FROM {table}
FROM text_content
WHERE
(
text @0@ $terms OR
@@ -182,192 +170,25 @@ impl TextContent {
AND user_id = $user_id
ORDER BY score DESC
LIMIT $limit;
"#,
table = Self::table_name(),
);
"#;
db.client
Ok(db
.client
.query(sql)
.bind(("terms", search_terms.to_owned()))
.bind(("user_id", user_id.to_owned()))
.bind(("limit", limit))
.await
.map_err(AppError::from)?
.take(0)
.map_err(AppError::from)
.await?
.take(0)?)
}
/// Builds a fallback display label for a source id when no matching content row exists.
#[must_use]
pub fn fallback_source_label(source_id: &str) -> String {
format!("Text snippet: {}", source_id_suffix(source_id))
}
/// Resolves human-readable labels for the given source ids owned by `user_id`.
pub async fn resolve_source_labels(
db: &SurrealDbClient,
user_id: &str,
source_ids: impl IntoIterator<Item = impl AsRef<str>>,
) -> Result<HashMap<String, String>, AppError> {
let source_ids: HashSet<String> = source_ids
.into_iter()
.map(|id| id.as_ref().to_string())
.collect();
if source_ids.is_empty() {
return Ok(HashMap::new());
}
let record_ids: Vec<RecordId> = source_ids
.iter()
.filter_map(|id| {
if id.contains(':') {
RecordId::from_str(id).ok()
} else {
Some(RecordId::from_table_key(Self::table_name(), id))
}
})
.collect();
let mut response = db
.client
.query(
"SELECT id, url_info, file_info, context, category, text FROM type::table($table_name) WHERE user_id = $user_id AND id INSIDE $record_ids",
)
.bind(("table_name", Self::table_name()))
.bind(("user_id", user_id.to_owned()))
.bind(("record_ids", record_ids))
.await
.map_err(AppError::from)?;
let contents: Vec<SourceLabelRow> = response.take(0).map_err(AppError::from)?;
tracing::debug!(
source_id_count = source_ids.len(),
label_row_count = contents.len(),
"resolved source labels"
);
let mut labels = HashMap::new();
for content in contents {
let label = build_source_label(&content);
labels.insert(content.id.clone(), label.clone());
labels.insert(format!("{}:{}", Self::table_name(), content.id), label);
}
Ok(labels)
}
}
const SOURCE_LABEL_MAX_CHARS: usize = 80;
#[derive(Deserialize)]
struct SourceLabelRow {
#[serde(deserialize_with = "deserialize_flexible_id")]
id: String,
#[serde(default)]
url_info: Option<UrlInfo>,
#[serde(default)]
file_info: Option<FileInfo>,
#[serde(default)]
context: Option<String>,
#[serde(default)]
category: String,
#[serde(default)]
text: String,
}
fn source_id_suffix(source_id: &str) -> String {
let start = source_id.len().saturating_sub(8);
source_id[start..].to_string()
}
fn truncate_with_ellipsis(value: &str, max_chars: usize) -> String {
const ELLIPSIS: &str = "";
if max_chars == 0 {
return if value.is_empty() {
String::new()
} else {
ELLIPSIS.to_string()
};
}
let mut end_byte = value.len();
for (count, (idx, _)) in value.char_indices().enumerate() {
if count == max_chars {
end_byte = idx;
break;
}
}
if end_byte == value.len() {
return value.to_string();
}
format!("{}{}", &value[..end_byte], ELLIPSIS)
}
fn first_non_empty_line(text: &str, max_chars: usize) -> Option<String> {
text.lines().find_map(|line| {
let trimmed = line.trim();
if trimmed.is_empty() {
None
} else {
Some(truncate_with_ellipsis(trimmed, max_chars))
}
})
}
fn build_source_label(row: &SourceLabelRow) -> String {
if let Some(url_info) = row.url_info.as_ref() {
let title = url_info.title.trim();
if !title.is_empty() {
return title.to_string();
}
let url = url_info.url.trim();
if !url.is_empty() {
return url.to_string();
}
}
if let Some(file_info) = row.file_info.as_ref() {
let name = file_info.file_name.trim();
if !name.is_empty() {
return name.to_string();
}
}
if let Some(context) = row.context.as_ref() {
let trimmed = context.trim();
if !trimmed.is_empty() {
return truncate_with_ellipsis(trimmed, SOURCE_LABEL_MAX_CHARS);
}
}
if let Some(text_label) = first_non_empty_line(&row.text, SOURCE_LABEL_MAX_CHARS) {
return text_label;
}
let category = row.category.trim();
if !category.is_empty() {
return truncate_with_ellipsis(category, SOURCE_LABEL_MAX_CHARS);
}
TextContent::fallback_source_label(&row.id)
}
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use anyhow::{self, Context};
use super::*;
use crate::test_utils::setup_test_db_with_runtime_indexes;
#[tokio::test]
async fn test_text_content_creation() -> anyhow::Result<()> {
async fn test_text_content_creation() {
// Test basic object creation
let text = "Test content text".to_string();
let context = "Test context".to_string();
@@ -391,11 +212,10 @@ mod tests {
assert!(text_content.file_info.is_none());
assert!(text_content.url_info.is_none());
assert!(!text_content.id.is_empty());
Ok(())
}
#[tokio::test]
async fn test_text_content_with_url() -> anyhow::Result<()> {
async fn test_text_content_with_url() {
// Test creating with URL
let text = "Content with URL".to_string();
let context = "URL context".to_string();
@@ -412,27 +232,26 @@ mod tests {
});
let text_content = TextContent::new(
text,
Some(context),
category,
text.clone(),
Some(context.clone()),
category.clone(),
None,
url_info.clone(),
user_id,
user_id.clone(),
);
// Check URL field is set
assert_eq!(text_content.url_info, url_info);
Ok(())
}
#[tokio::test]
async fn test_text_content_patch() -> anyhow::Result<()> {
async fn test_text_content_patch() {
// Setup in-memory database for testing
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
// Create initial text content
let initial_text = "Initial text".to_string();
@@ -453,7 +272,7 @@ mod tests {
let stored: Option<TextContent> = db
.store_item(text_content.clone())
.await
.with_context(|| "Failed to store text content".to_string())?;
.expect("Failed to store text content");
assert!(stored.is_some());
// New values for patch
@@ -464,42 +283,31 @@ mod tests {
// Apply the patch
TextContent::patch(&text_content.id, new_context, new_category, new_text, &db)
.await
.with_context(|| "Failed to patch text content".to_string())?;
.expect("Failed to patch text content");
// Retrieve the updated content
let updated: Option<TextContent> = db
.get_item(&text_content.id)
.await
.with_context(|| "Failed to get updated text content".to_string())?;
let updated_content = updated.with_context(|| "expected updated content".to_string())?;
.expect("Failed to get updated text content");
assert!(updated.is_some());
let updated_content = updated.unwrap();
// Verify the updates
assert_eq!(updated_content.context, Some(new_context.to_string()));
assert_eq!(updated_content.category, new_category);
assert_eq!(updated_content.text, new_text);
assert!(updated_content.updated_at > text_content.updated_at);
Ok(())
}
#[tokio::test]
async fn test_text_content_patch_not_found() -> anyhow::Result<()> {
let db = setup_test_db_with_runtime_indexes().await?;
let err = TextContent::patch("missing-id", "ctx", "cat", "text", &db)
.await
.expect_err("expected not found");
assert!(matches!(err, AppError::NotFound(_)));
Ok(())
}
#[tokio::test]
async fn test_has_other_with_file_detects_shared_usage() -> anyhow::Result<()> {
async fn test_has_other_with_file_detects_shared_usage() {
let namespace = "test_ns";
let database = &Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, database)
.await
.with_context(|| "Failed to start in-memory surrealdb".to_string())?;
.expect("Failed to start in-memory surrealdb");
let user_id = "user123".to_string();
let file_info = FileInfo {
@@ -532,110 +340,24 @@ mod tests {
db.store_item(content_a.clone())
.await
.with_context(|| "Failed to store first content".to_string())?;
.expect("Failed to store first content");
db.store_item(content_b.clone())
.await
.with_context(|| "Failed to store second content".to_string())?;
.expect("Failed to store second content");
let has_other = TextContent::has_other_with_file(&file_info.id, &content_a.id, &db)
.await
.with_context(|| "Failed to check for shared file usage".to_string())?;
.expect("Failed to check for shared file usage");
assert!(has_other);
let _removed: Option<TextContent> = db
.delete_item(&content_b.id)
.await
.with_context(|| "Failed to delete second content".to_string())?;
.expect("Failed to delete second content");
let has_other_after = TextContent::has_other_with_file(&file_info.id, &content_a.id, &db)
.await
.with_context(|| "Failed to check shared usage after delete".to_string())?;
.expect("Failed to check shared usage after delete");
assert!(!has_other_after);
Ok(())
}
#[tokio::test]
async fn test_search_returns_empty_when_no_content() -> anyhow::Result<()> {
let db = setup_test_db_with_runtime_indexes().await?;
let results = TextContent::search(&db, "hello", "user", 5)
.await
.with_context(|| "search".to_string())?;
assert!(results.is_empty());
Ok(())
}
#[tokio::test]
async fn test_search_finds_matching_text_and_filters_user() -> anyhow::Result<()> {
let db = setup_test_db_with_runtime_indexes().await?;
let user_id = "search_user";
let matching = TextContent::new(
"rust programming language".to_string(),
Some("context".to_string()),
"notes".to_string(),
None,
None,
user_id.to_string(),
);
let other_user = TextContent::new(
"rust programming language".to_string(),
None,
"notes".to_string(),
None,
None,
"other_user".to_string(),
);
db.store_item(matching.clone())
.await
.with_context(|| "store matching".to_string())?;
db.store_item(other_user)
.await
.with_context(|| "store other user".to_string())?;
let results = TextContent::search(&db, "rust", user_id, 5)
.await
.with_context(|| "search".to_string())?;
assert_eq!(results.len(), 1);
let row = results.first().context("expected one result")?;
assert_eq!(row.id, matching.id);
assert_eq!(row.user_id, user_id);
assert!(row.score.is_finite());
Ok(())
}
#[tokio::test]
async fn test_resolve_source_labels_uses_url_title() -> anyhow::Result<()> {
let db = setup_test_db_with_runtime_indexes().await?;
let user_id = "label_user";
let content = TextContent::new(
"body".to_string(),
None,
"notes".to_string(),
None,
Some(UrlInfo {
url: "https://example.com/doc".to_string(),
title: "Example Document".to_string(),
image_id: String::new(),
}),
user_id.to_string(),
);
db.store_item(content.clone()).await?;
let labels = TextContent::resolve_source_labels(&db, user_id, [content.id.clone()]).await?;
assert_eq!(
labels.get(&content.id),
Some(&"Example Document".to_string())
);
assert_eq!(
labels.get(&format!("text_content:{}", content.id)),
Some(&"Example Document".to_string())
);
Ok(())
}
}
+102 -286
View File
@@ -25,58 +25,6 @@ pub struct CategoryResponse {
category: String,
}
use std::str::FromStr;
/// Supported UI themes.
#[derive(Serialize, Deserialize, Debug, Clone, Copy, PartialEq, Eq, Default)]
#[serde(rename_all = "kebab-case")]
pub enum Theme {
Light,
Dark,
WarmPaper,
ObsidianPrism,
#[default]
System,
}
impl FromStr for Theme {
type Err = ();
fn from_str(s: &str) -> Result<Self, Self::Err> {
match s {
"light" => Ok(Self::Light),
"dark" => Ok(Self::Dark),
"warm-paper" => Ok(Self::WarmPaper),
"obsidian-prism" => Ok(Self::ObsidianPrism),
"system" => Ok(Self::System),
_ => Err(()),
}
}
}
impl Theme {
#[must_use]
pub fn as_str(&self) -> &'static str {
match self {
Self::Light => "light",
Self::Dark => "dark",
Self::WarmPaper => "warm-paper",
Self::ObsidianPrism => "obsidian-prism",
Self::System => "system",
}
}
/// Returns the theme that should be initially applied.
/// For "system", defaults to "light".
#[must_use]
pub fn initial_theme(&self) -> &'static str {
match self {
Self::System => "light",
other => other.as_str(),
}
}
}
stored_object!(
#[allow(clippy::unsafe_derive_deserialize)]
User, "user", {
@@ -86,21 +34,9 @@ stored_object!(
api_key: Option<String>,
admin: bool,
#[serde(default)]
timezone: String,
#[serde(default, deserialize_with = "deserialize_theme_or_default")]
theme: Theme
timezone: String
});
fn deserialize_theme_or_default<'de, D>(deserializer: D) -> Result<Theme, D::Error>
where
D: serde::Deserializer<'de>,
{
let raw = Option::<String>::deserialize(deserializer)?;
Ok(raw
.and_then(|value| Theme::from_str(value.as_str()).ok())
.unwrap_or_default())
}
#[async_trait]
impl Authentication<User, String, Surreal<Any>> for User {
async fn load_user(userid: String, db: Option<&Surreal<Any>>) -> Result<User, anyhow::Error> {
@@ -134,11 +70,6 @@ fn validate_timezone(input: &str) -> String {
"UTC".to_owned()
}
/// Ensures a theme string is valid, defaulting to "system" when invalid.
fn validate_theme(input: &str) -> Theme {
Theme::from_str(input).unwrap_or_default()
}
#[derive(Serialize, Deserialize, Debug, Clone)]
pub struct DashboardStats {
pub total_documents: i64,
@@ -237,7 +168,6 @@ impl User {
password: String,
db: &SurrealDbClient,
timezone: String,
theme: String,
) -> Result<Self, AppError> {
// verify that the application allows new creations
let systemsettings = SystemSettings::get_current(db).await?;
@@ -246,7 +176,6 @@ impl User {
}
let validated_tz = validate_timezone(&timezone);
let validated_theme = validate_theme(&theme);
let now = Utc::now();
let id = Uuid::new_v4().to_string();
@@ -261,8 +190,7 @@ impl User {
anonymous = false,
created_at = $created_at,
updated_at = $updated_at,
timezone = $timezone,
theme = $theme",
timezone = $timezone",
)
.bind(("table", "user"))
.bind(("id", id))
@@ -271,7 +199,6 @@ impl User {
.bind(("created_at", surrealdb::Datetime::from(now)))
.bind(("updated_at", surrealdb::Datetime::from(now)))
.bind(("timezone", validated_tz))
.bind(("theme", validated_theme.as_str()))
.await?
.take(1)?;
@@ -373,7 +300,7 @@ impl User {
.client
.query(
"UPDATE type::thing('user', $id)
SET api_key = NONE
SET api_key = test_string_nullish
RETURN AFTER",
)
.bind(("id", id.to_owned()))
@@ -534,25 +461,13 @@ impl User {
db: &SurrealDbClient,
) -> Result<(), AppError> {
db.query("UPDATE type::thing('user', $user_id) SET timezone = $timezone")
.bind(("table_name", Self::table_name()))
.bind(("user_id", user_id.to_string()))
.bind(("timezone", timezone.to_string()))
.await?;
Ok(())
}
pub async fn update_theme(
user_id: &str,
theme: &str,
db: &SurrealDbClient,
) -> Result<(), AppError> {
let validated_theme = validate_theme(theme);
db.query("UPDATE type::thing('user', $user_id) SET theme = $theme")
.bind(("user_id", user_id.to_string()))
.bind(("theme", validated_theme.as_str()))
.await?;
Ok(())
}
pub async fn get_user_categories(
user_id: &str,
db: &SurrealDbClient,
@@ -580,7 +495,7 @@ impl User {
let entity: KnowledgeEntity = db
.get_item(id)
.await?
.ok_or_else(|| AppError::NotFound("entity not found".into()))?;
.ok_or_else(|| AppError::NotFound("Entity not found".into()))?;
if entity.user_id != user_id {
return Err(AppError::Auth("Access denied".into()));
@@ -597,7 +512,7 @@ impl User {
let text_content: TextContent = db
.get_item(id)
.await?
.ok_or_else(|| AppError::NotFound("content not found".into()))?;
.ok_or_else(|| AppError::NotFound("Content not found".into()))?;
if text_content.user_id != user_id {
return Err(AppError::Auth("Access denied".into()));
@@ -688,7 +603,7 @@ impl User {
db.delete_item::<IngestionTask>(id)
.await
.map_err(AppError::from)?;
.map_err(AppError::Database)?;
Ok(())
}
@@ -724,20 +639,30 @@ impl User {
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use anyhow::{self, Context};
use super::*;
use crate::storage::types::ingestion_payload::IngestionPayload;
use crate::storage::types::ingestion_task::{IngestionTask, TaskState, MAX_ATTEMPTS};
use std::collections::HashSet;
use crate::test_utils::setup_test_db;
// Helper function to set up a test database with SystemSettings
async fn setup_test_db() -> SurrealDbClient {
let namespace = "test_ns";
let database = Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(namespace, &database)
.await
.expect("Failed to start in-memory surrealdb");
db.apply_migrations()
.await
.expect("Failed to setup the migrations");
db
}
#[tokio::test]
async fn test_user_creation() -> anyhow::Result<()> {
async fn test_user_creation() {
// Setup test database
let db = setup_test_db().await?;
let db = setup_test_db().await;
// Create a user
let email = "test@example.com";
@@ -749,10 +674,9 @@ mod tests {
password.to_string(),
&db,
timezone.to_string(),
"system".to_string(),
)
.await
.with_context(|| "Failed to create user".to_string())?;
.expect("Failed to create user");
// Verify user properties
assert!(!user.id.is_empty());
@@ -765,17 +689,18 @@ mod tests {
let retrieved: Option<User> = db
.get_item(&user.id)
.await
.with_context(|| "Failed to retrieve user".to_string())?;
let retrieved = retrieved.with_context(|| "expected user to exist".to_string())?;
.expect("Failed to retrieve user");
assert!(retrieved.is_some());
let retrieved = retrieved.unwrap();
assert_eq!(retrieved.id, user.id);
assert_eq!(retrieved.email, email);
Ok(())
}
#[tokio::test]
async fn test_user_authentication() -> anyhow::Result<()> {
async fn test_user_authentication() {
// Setup test database
let db = setup_test_db().await?;
let db = setup_test_db().await;
// Create a user
let email = "auth_test@example.com";
@@ -786,10 +711,9 @@ mod tests {
password.to_string(),
&db,
"UTC".to_string(),
"system".to_string(),
)
.await
.with_context(|| "Failed to create user".to_string())?;
.expect("Failed to create user");
// Test successful authentication
let auth_result = User::authenticate(email, password, &db).await;
@@ -802,12 +726,11 @@ mod tests {
// Test failed authentication with non-existent user
let nonexistent = User::authenticate("nonexistent@example.com", password, &db).await;
assert!(nonexistent.is_err());
Ok(())
}
#[tokio::test]
async fn test_get_unfinished_ingestion_tasks_filters_correctly() -> anyhow::Result<()> {
let db = setup_test_db().await?;
async fn test_get_unfinished_ingestion_tasks_filters_correctly() {
let db = setup_test_db().await;
let user_id = "unfinished_user";
let other_user_id = "other_user";
@@ -821,14 +744,14 @@ mod tests {
let created_task = IngestionTask::new(payload.clone(), user_id.to_string());
db.store_item(created_task.clone())
.await
.with_context(|| "Failed to store created task".to_string())?;
.expect("Failed to store created task");
let mut processing_task = IngestionTask::new(payload.clone(), user_id.to_string());
processing_task.state = TaskState::Processing;
processing_task.attempts = 1;
db.store_item(processing_task.clone())
.await
.with_context(|| "Failed to store processing task".to_string())?;
.expect("Failed to store processing task");
let mut failed_retry_task = IngestionTask::new(payload.clone(), user_id.to_string());
failed_retry_task.state = TaskState::Failed;
@@ -836,7 +759,7 @@ mod tests {
failed_retry_task.scheduled_at = chrono::Utc::now() - chrono::Duration::minutes(5);
db.store_item(failed_retry_task.clone())
.await
.with_context(|| "Failed to store retryable failed task".to_string())?;
.expect("Failed to store retryable failed task");
let mut failed_blocked_task = IngestionTask::new(payload.clone(), user_id.to_string());
failed_blocked_task.state = TaskState::Failed;
@@ -844,13 +767,13 @@ mod tests {
failed_blocked_task.error_message = Some("Too many failures".into());
db.store_item(failed_blocked_task.clone())
.await
.with_context(|| "Failed to store blocked task".to_string())?;
.expect("Failed to store blocked task");
let mut completed_task = IngestionTask::new(payload.clone(), user_id.to_string());
completed_task.state = TaskState::Succeeded;
db.store_item(completed_task.clone())
.await
.with_context(|| "Failed to store completed task".to_string())?;
.expect("Failed to store completed task");
let other_payload = IngestionPayload::Text {
text: "Other".to_string(),
@@ -861,11 +784,11 @@ mod tests {
let other_task = IngestionTask::new(other_payload, other_user_id.to_string());
db.store_item(other_task)
.await
.with_context(|| "Failed to store other user task".to_string())?;
.expect("Failed to store other user task");
let unfinished = User::get_unfinished_ingestion_tasks(user_id, &db)
.await
.with_context(|| "Failed to fetch unfinished tasks".to_string())?;
.expect("Failed to fetch unfinished tasks");
let unfinished_ids: HashSet<String> =
unfinished.iter().map(|task| task.id.clone()).collect();
@@ -876,12 +799,11 @@ mod tests {
assert!(!unfinished_ids.contains(&failed_blocked_task.id));
assert!(!unfinished_ids.contains(&completed_task.id));
assert_eq!(unfinished_ids.len(), 3);
Ok(())
}
#[tokio::test]
async fn test_get_all_ingestion_tasks_returns_sorted() -> anyhow::Result<()> {
let db = setup_test_db().await?;
async fn test_get_all_ingestion_tasks_returns_sorted() {
let db = setup_test_db().await;
let user_id = "archive_user";
let other_user_id = "other_user";
@@ -894,19 +816,15 @@ mod tests {
// Oldest task
let mut first = IngestionTask::new(payload.clone(), user_id.to_string());
first.created_at -= chrono::Duration::minutes(1);
first.created_at = first.created_at - chrono::Duration::minutes(1);
first.updated_at = first.created_at;
first.state = TaskState::Succeeded;
db.store_item(first.clone())
.await
.with_context(|| "store first".to_string())?;
db.store_item(first.clone()).await.expect("store first");
// Latest task
let mut second = IngestionTask::new(payload.clone(), user_id.to_string());
second.state = TaskState::Processing;
db.store_item(second.clone())
.await
.with_context(|| "store second".to_string())?;
db.store_item(second.clone()).await.expect("store second");
let other_payload = IngestionPayload::Text {
text: "Other".to_string(),
@@ -915,24 +833,21 @@ mod tests {
user_id: other_user_id.to_string(),
};
let other_task = IngestionTask::new(other_payload, other_user_id.to_string());
db.store_item(other_task)
.await
.with_context(|| "store other".to_string())?;
db.store_item(other_task).await.expect("store other");
let tasks = User::get_all_ingestion_tasks(user_id, &db)
.await
.with_context(|| "fetch all tasks".to_string())?;
.expect("fetch all tasks");
assert_eq!(tasks.len(), 2);
assert_eq!(tasks.first().map(|t| &t.id), Some(&second.id)); // newest first
assert_eq!(tasks.get(1).map(|t| &t.id), Some(&first.id));
Ok(())
assert_eq!(tasks[0].id, second.id); // newest first
assert_eq!(tasks[1].id, first.id);
}
#[tokio::test]
async fn test_find_by_email() -> anyhow::Result<()> {
async fn test_find_by_email() {
// Setup test database
let db = setup_test_db().await?;
let db = setup_test_db().await;
// Create a user
let email = "find_test@example.com";
@@ -943,31 +858,30 @@ mod tests {
password.to_string(),
&db,
"UTC".to_string(),
"system".to_string(),
)
.await
.with_context(|| "Failed to create user".to_string())?;
.expect("Failed to create user");
// Test finding user by email
let found_user = User::find_by_email(email, &db)
.await
.with_context(|| "Error searching for user".to_string())?
.with_context(|| "expected user to exist".to_string())?;
.expect("Error searching for user");
assert!(found_user.is_some());
let found_user = found_user.unwrap();
assert_eq!(found_user.id, created_user.id);
assert_eq!(found_user.email, email);
// Test finding non-existent user
let not_found = User::find_by_email("nonexistent@example.com", &db)
.await
.with_context(|| "Error searching for user".to_string())?;
.expect("Error searching for user");
assert!(not_found.is_none());
Ok(())
}
#[tokio::test]
async fn test_api_key_management() -> anyhow::Result<()> {
async fn test_api_key_management() {
// Setup test database
let db = setup_test_db().await?;
let db = setup_test_db().await;
// Create a user
let email = "apikey_test@example.com";
@@ -978,10 +892,9 @@ mod tests {
password.to_string(),
&db,
"UTC".to_string(),
"system".to_string(),
)
.await
.with_context(|| "Failed to create user".to_string())?;
.expect("Failed to create user");
// Initially, user should have no API key
assert!(user.api_key.is_none());
@@ -989,7 +902,7 @@ mod tests {
// Generate API key
let api_key = User::set_api_key(&user.id, &db)
.await
.with_context(|| "Failed to set API key".to_string())?;
.expect("Failed to set API key");
assert!(!api_key.is_empty());
assert!(api_key.starts_with("sk_"));
@@ -997,78 +910,44 @@ mod tests {
let updated_user: Option<User> = db
.get_item(&user.id)
.await
.with_context(|| "Failed to retrieve user".to_string())?;
let updated_user = updated_user.with_context(|| "expected updated user".to_string())?;
.expect("Failed to retrieve user");
assert!(updated_user.is_some());
let updated_user = updated_user.unwrap();
assert_eq!(updated_user.api_key, Some(api_key.clone()));
// Test finding user by API key
let found_user = User::find_by_api_key(&api_key, &db)
.await
.with_context(|| "Error searching by API key".to_string())?
.with_context(|| "expected user found by api key".to_string())?;
.expect("Error searching by API key");
assert!(found_user.is_some());
let found_user = found_user.unwrap();
assert_eq!(found_user.id, user.id);
// Revoke API key
User::revoke_api_key(&user.id, &db)
.await
.with_context(|| "Failed to revoke API key".to_string())?;
.expect("Failed to revoke API key");
// Verify API key was revoked
let revoked_user: Option<User> = db
.get_item(&user.id)
.await
.with_context(|| "Failed to retrieve user".to_string())?;
let revoked_user = revoked_user.with_context(|| "expected revoked user".to_string())?;
.expect("Failed to retrieve user");
assert!(revoked_user.is_some());
let revoked_user = revoked_user.unwrap();
assert!(revoked_user.api_key.is_none());
// Test searching by revoked API key
let not_found = User::find_by_api_key(&api_key, &db)
.await
.with_context(|| "Error searching by API key".to_string())?;
.expect("Error searching by API key");
assert!(not_found.is_none());
Ok(())
}
#[tokio::test]
async fn test_set_api_key_with_none_theme() {
let db = setup_test_db().await.expect("Failed to setup test db");
let user = User::create_new(
"legacy_theme@example.com".to_string(),
"apikey_password".to_string(),
&db,
"UTC".to_string(),
"system".to_string(),
)
.await
.expect("Failed to create user");
db.client
.query("UPDATE type::thing('user', $id) SET theme = NONE")
.bind(("id", user.id.clone()))
.await
.expect("Failed to set user theme to NONE");
let api_key = User::set_api_key(&user.id, &db)
.await
.expect("set_api_key should tolerate NONE theme");
assert!(api_key.starts_with("sk_"));
let updated_user = db
.get_item::<User>(&user.id)
.await
.expect("Failed to retrieve user")
.expect("User should still exist");
assert_eq!(updated_user.theme, Theme::System);
assert_eq!(updated_user.api_key, Some(api_key));
}
#[tokio::test]
async fn test_password_update() -> anyhow::Result<()> {
async fn test_password_update() {
// Setup test database
let db = setup_test_db().await?;
let db = setup_test_db().await;
// Create a user
let email = "pwd_test@example.com";
@@ -1080,10 +959,9 @@ mod tests {
old_password.to_string(),
&db,
"UTC".to_string(),
"system".to_string(),
)
.await
.with_context(|| "Failed to create user".to_string())?;
.expect("Failed to create user");
// Authenticate with old password
let auth_result = User::authenticate(email, old_password, &db).await;
@@ -1092,7 +970,7 @@ mod tests {
// Update password
User::patch_password(email, new_password, &db)
.await
.with_context(|| "Failed to update password".to_string())?;
.expect("Failed to update password");
// Old password should no longer work
let old_auth = User::authenticate(email, old_password, &db).await;
@@ -1101,11 +979,10 @@ mod tests {
// New password should work
let new_auth = User::authenticate(email, new_password, &db).await;
assert!(new_auth.is_ok());
Ok(())
}
#[tokio::test]
async fn test_validate_timezone() -> anyhow::Result<()> {
async fn test_validate_timezone() {
// Valid timezones should be accepted as-is
assert_eq!(validate_timezone("America/New_York"), "America/New_York");
assert_eq!(validate_timezone("Europe/London"), "Europe/London");
@@ -1115,13 +992,12 @@ mod tests {
// Invalid timezones should be replaced with UTC
assert_eq!(validate_timezone("Invalid/Timezone"), "UTC");
assert_eq!(validate_timezone("Not_Real"), "UTC");
Ok(())
}
#[tokio::test]
async fn test_timezone_update() -> anyhow::Result<()> {
async fn test_timezone_update() {
// Setup test database
let db = setup_test_db().await?;
let db = setup_test_db().await;
// Create user with default timezone
let email = "timezone_test@example.com";
@@ -1130,10 +1006,9 @@ mod tests {
"password".to_string(),
&db,
"UTC".to_string(),
"system".to_string(),
)
.await
.with_context(|| "Failed to create user".to_string())?;
.expect("Failed to create user");
assert_eq!(user.timezone, "UTC");
@@ -1141,64 +1016,58 @@ mod tests {
let new_timezone = "Europe/Paris";
User::update_timezone(&user.id, new_timezone, &db)
.await
.with_context(|| "Failed to update timezone".to_string())?;
.expect("Failed to update timezone");
// Verify timezone was updated
let updated_user: Option<User> = db
.get_item(&user.id)
.await
.with_context(|| "Failed to retrieve user".to_string())?;
let updated_user = updated_user.with_context(|| "expected updated user".to_string())?;
.expect("Failed to retrieve user");
assert!(updated_user.is_some());
let updated_user = updated_user.unwrap();
assert_eq!(updated_user.timezone, new_timezone);
Ok(())
}
#[tokio::test]
async fn test_conversations_order() -> anyhow::Result<()> {
let db = setup_test_db().await?;
async fn test_conversations_order() {
let db = setup_test_db().await;
let user_id = "user_order_test";
// Create conversations with varying updated_at timestamps
let mut conversations = Vec::new();
for i in 0..5 {
let mut conv = Conversation::new(user_id.to_string(), format!("Conv {i}"));
let mut conv = Conversation::new(user_id.to_string(), format!("Conv {}", i));
// Fake updated_at i minutes apart
conv.created_at = chrono::Utc::now() - chrono::Duration::minutes(i);
db.store_item(conv.clone())
.await
.with_context(|| "Failed to store conversation".to_string())?;
.expect("Failed to store conversation");
conversations.push(conv);
}
// Retrieve via get_user_conversations - should be ordered by updated_at DESC
let retrieved = User::get_user_conversations(user_id, &db)
.await
.with_context(|| "Failed to get conversations".to_string())?;
.expect("Failed to get conversations");
assert_eq!(retrieved.len(), conversations.len());
for pair in retrieved.windows(2) {
let a = pair.first().context("expected first in pair")?;
let b = pair.get(1).context("expected second in pair")?;
for window in retrieved.windows(2) {
// Assert each earlier conversation has updated_at >= later conversation
assert!(
a.created_at >= b.created_at,
window[0].created_at >= window[1].created_at,
"Conversations not ordered descending by created_at"
);
}
// Check first conversation title matches the most recently updated
let most_recent = conversations
.iter()
.max_by_key(|c| c.created_at)
.context("expected most recent")?;
let r0 = retrieved.first().context("expected first result")?;
assert_eq!(r0.id, most_recent.id);
Ok(())
let most_recent = conversations.iter().max_by_key(|c| c.created_at).unwrap();
assert_eq!(retrieved[0].id, most_recent.id);
}
#[tokio::test]
async fn test_get_latest_text_contents_returns_last_five() -> anyhow::Result<()> {
let db = setup_test_db().await?;
async fn test_get_latest_text_contents_returns_last_five() {
let db = setup_test_db().await;
let user_id = "latest_text_user";
let mut inserted_ids = Vec::new();
@@ -1206,8 +1075,8 @@ mod tests {
for i in 0..12 {
let mut item = TextContent::new(
format!("Text {i}"),
Some(format!("Context {i}")),
format!("Text {}", i),
Some(format!("Context {}", i)),
"Category".to_string(),
None,
None,
@@ -1220,19 +1089,18 @@ mod tests {
db.store_item(item.clone())
.await
.with_context(|| "Failed to store text content".to_string())?;
.expect("Failed to store text content");
inserted_ids.push(item.id.clone());
}
let latest = User::get_latest_text_contents(user_id, &db)
.await
.with_context(|| "Failed to fetch latest text contents".to_string())?;
.expect("Failed to fetch latest text contents");
assert_eq!(latest.len(), 5, "Expected exactly five items");
let start = inserted_ids.len().saturating_sub(5);
let mut expected_ids = inserted_ids.get(start..).unwrap_or_default().to_vec();
let mut expected_ids = inserted_ids[inserted_ids.len() - 5..].to_vec();
expected_ids.reverse();
let returned_ids: Vec<String> = latest.iter().map(|item| item.id.clone()).collect();
@@ -1241,63 +1109,11 @@ mod tests {
"Latest items did not match expectation"
);
for pair in latest.windows(2) {
let a = pair.first().context("expected first in pair")?;
let b = pair.get(1).context("expected second in pair")?;
for window in latest.windows(2) {
assert!(
a.created_at >= b.created_at,
window[0].created_at >= window[1].created_at,
"Results are not ordered by created_at descending"
);
}
Ok(())
}
#[tokio::test]
async fn test_validate_theme() -> anyhow::Result<()> {
assert_eq!(validate_theme("light"), Theme::Light);
assert_eq!(validate_theme("dark"), Theme::Dark);
assert_eq!(validate_theme("system"), Theme::System);
assert_eq!(validate_theme("invalid"), Theme::System);
Ok(())
}
#[tokio::test]
async fn test_theme_update() -> anyhow::Result<()> {
let db = setup_test_db().await?;
let email = "theme_test@example.com";
let user = User::create_new(
email.to_string(),
"password".to_string(),
&db,
"UTC".to_string(),
"system".to_string(),
)
.await
.with_context(|| "Failed to create user".to_string())?;
assert_eq!(user.theme, Theme::System);
User::update_theme(&user.id, "dark", &db)
.await
.with_context(|| "update theme".to_string())?;
let updated = db
.get_item::<User>(&user.id)
.await
.with_context(|| "get user".to_string())?
.with_context(|| "expected user".to_string())?;
assert_eq!(updated.theme, Theme::Dark);
// Invalid theme should default to system (but update_theme calls validate_theme)
User::update_theme(&user.id, "invalid", &db)
.await
.with_context(|| "update theme invalid".to_string())?;
let updated2 = db
.get_item::<User>(&user.id)
.await
.with_context(|| "get user".to_string())?
.with_context(|| "expected user".to_string())?;
assert_eq!(updated2.theme, Theme::System);
Ok(())
}
}
-93
View File
@@ -1,93 +0,0 @@
//! Shared helpers for in-memory SurrealDB tests.
#![cfg(any(test, feature = "test-utils"))]
use anyhow::{Context, Result};
use uuid::Uuid;
use crate::storage::{
db::SurrealDbClient,
indexes::{ensure_runtime, rebuild},
types::{
knowledge_entity_embedding::KnowledgeEntityEmbedding, system_settings::SystemSettings,
text_chunk_embedding::TextChunkEmbedding,
},
};
const TEST_NAMESPACE: &str = "test_ns";
/// Starts an in-memory database, applies migrations, and returns a client.
///
/// # Errors
///
/// Returns an error if the database cannot be started or migrations fail.
pub async fn setup_test_db() -> Result<SurrealDbClient> {
let database = Uuid::new_v4().to_string();
let db = SurrealDbClient::memory(TEST_NAMESPACE, &database)
.await
.context("start in-memory surrealdb")?;
db.apply_migrations().await.context("apply migrations")?;
Ok(db)
}
/// Updates singleton [`SystemSettings`] embedding dimensions for tests.
///
/// # Errors
///
/// Returns an error if settings cannot be loaded or updated.
pub async fn configure_embedding_dimension(db: &SurrealDbClient, dimension: u32) -> Result<()> {
let mut settings = SystemSettings::get_current(db).await?;
settings.embedding_dimensions = dimension;
SystemSettings::update(db, settings).await?;
Ok(())
}
/// Starts a test database and sets the embedding dimension in system settings.
///
/// # Errors
///
/// Returns an error if setup or settings update fails.
pub async fn setup_test_db_with_embedding_dimension(dimension: u32) -> Result<SurrealDbClient> {
let db = setup_test_db().await?;
configure_embedding_dimension(&db, dimension).await?;
Ok(db)
}
/// Prepares a database for text-chunk embedding tests at the given dimension.
///
/// # Errors
///
/// Returns an error if setup, settings update, or index redefinition fails.
pub async fn prepare_text_chunk_test_db(dimension: u32) -> Result<SurrealDbClient> {
let db = setup_test_db_with_embedding_dimension(dimension).await?;
TextChunkEmbedding::redefine_hnsw_index(&db, dimension as usize)
.await
.with_context(|| format!("set text chunk index dimension to {dimension}"))?;
Ok(db)
}
/// Prepares a database for knowledge-entity embedding tests at the given dimension.
///
/// # Errors
///
/// Returns an error if setup, settings update, or index redefinition fails.
pub async fn prepare_knowledge_entity_test_db(dimension: u32) -> Result<SurrealDbClient> {
let db = setup_test_db_with_embedding_dimension(dimension).await?;
KnowledgeEntityEmbedding::redefine_hnsw_index(&db, dimension as usize)
.await
.with_context(|| format!("set knowledge entity index dimension to {dimension}"))?;
Ok(db)
}
/// Starts a test database and ensures runtime FTS/HNSW indexes are ready.
///
/// # Errors
///
/// Returns an error if setup, index creation, or rebuild fails.
pub async fn setup_test_db_with_runtime_indexes() -> Result<SurrealDbClient> {
let db = setup_test_db().await?;
ensure_runtime(&db, 1536).await?;
rebuild(&db).await?;
Ok(db)
}
+9 -128
View File
@@ -1,18 +1,9 @@
use config::{Config, ConfigError, Environment, File};
use serde::{Deserialize, Serialize};
use std::{env, str::FromStr, sync::Once};
use thiserror::Error;
/// Error returned when parsing an embedding backend name.
#[derive(Debug, Error, PartialEq, Eq)]
#[error("unknown embedding backend '{input}': expected 'openai', 'hashed', or 'fastembed'")]
pub struct ParseEmbeddingBackendError {
/// The unrecognized input string.
pub input: String,
}
use serde::Deserialize;
use std::env;
/// Selects the embedding backend for vector generation.
#[derive(Clone, Copy, Deserialize, Serialize, Debug, Default, PartialEq, Eq)]
#[derive(Clone, Deserialize, Debug, Default, PartialEq)]
#[serde(rename_all = "lowercase")]
pub enum EmbeddingBackend {
/// Use OpenAI-compatible API for embeddings.
@@ -24,38 +15,11 @@ pub enum EmbeddingBackend {
Hashed,
}
impl EmbeddingBackend {
#[must_use]
pub fn as_str(self) -> &'static str {
match self {
Self::OpenAI => "openai",
Self::FastEmbed => "fastembed",
Self::Hashed => "hashed",
}
}
}
impl FromStr for EmbeddingBackend {
type Err = ParseEmbeddingBackendError;
fn from_str(s: &str) -> Result<Self, Self::Err> {
match s.to_ascii_lowercase().as_str() {
"openai" => Ok(Self::OpenAI),
"hashed" => Ok(Self::Hashed),
"fastembed" | "fast-embed" | "fast" => Ok(Self::FastEmbed),
other => Err(ParseEmbeddingBackendError {
input: other.to_string(),
}),
}
}
}
#[derive(Clone, Copy, Deserialize, Debug, PartialEq)]
#[derive(Clone, Deserialize, Debug, PartialEq)]
#[serde(rename_all = "lowercase")]
pub enum StorageKind {
Local,
Memory,
S3,
}
/// Default storage backend when none is configured.
@@ -63,12 +27,8 @@ fn default_storage_kind() -> StorageKind {
StorageKind::Local
}
fn default_s3_region() -> String {
"us-east-1".to_string()
}
/// Selects the strategy used for PDF ingestion.
#[derive(Clone, Copy, Deserialize, Debug)]
#[derive(Clone, Deserialize, Debug)]
#[serde(rename_all = "kebab-case")]
pub enum PdfIngestMode {
/// Only rely on classic text extraction (no LLM fallbacks).
@@ -99,12 +59,6 @@ pub struct AppConfig {
pub openai_base_url: String,
#[serde(default = "default_storage_kind")]
pub storage: StorageKind,
#[serde(default)]
pub s3_bucket: Option<String>,
#[serde(default)]
pub s3_endpoint: Option<String>,
#[serde(default = "default_s3_region")]
pub s3_region: String,
#[serde(default = "default_pdf_ingest_mode")]
pub pdf_ingest_mode: PdfIngestMode,
#[serde(default = "default_reranking_enabled")]
@@ -117,24 +71,10 @@ pub struct AppConfig {
pub fastembed_show_download_progress: Option<bool>,
#[serde(default)]
pub fastembed_max_length: Option<usize>,
/// HuggingFace-style FastEmbed `model_code` (e.g. `Xenova/bge-small-en-v1.5`). Overrides
/// `system_settings.embedding_model` when `embedding_backend` is `fastembed`.
#[serde(default)]
pub fastembed_model: Option<String>,
pub retrieval_strategy: Option<String>,
#[serde(default)]
pub embedding_backend: EmbeddingBackend,
#[serde(default)]
pub embedding_pool_size: Option<usize>,
#[serde(default = "default_ingest_max_body_bytes")]
pub ingest_max_body_bytes: usize,
#[serde(default = "default_ingest_max_files")]
pub ingest_max_files: usize,
#[serde(default = "default_ingest_max_content_bytes")]
pub ingest_max_content_bytes: usize,
#[serde(default = "default_ingest_max_context_bytes")]
pub ingest_max_context_bytes: usize,
#[serde(default = "default_ingest_max_category_bytes")]
pub ingest_max_category_bytes: usize,
}
/// Default data directory for persisted assets.
@@ -152,37 +92,11 @@ fn default_reranking_enabled() -> bool {
false
}
fn default_ingest_max_body_bytes() -> usize {
20_000_000
}
fn default_ingest_max_files() -> usize {
5
}
fn default_ingest_max_content_bytes() -> usize {
262_144
}
fn default_ingest_max_context_bytes() -> usize {
16_384
}
fn default_ingest_max_category_bytes() -> usize {
128
}
static ORT_PATH_INIT: Once = Once::new();
/// Sets `ORT_DYLIB_PATH` once per process when a bundled ONNX runtime library is found.
pub fn ensure_ort_path() {
ORT_PATH_INIT.call_once(|| {
if env::var_os("ORT_DYLIB_PATH").is_some() {
return;
}
let Ok(mut exe) = env::current_exe() else {
return;
};
if let Ok(mut exe) = env::current_exe() {
exe.pop();
if cfg!(target_os = "windows") {
@@ -205,7 +119,7 @@ pub fn ensure_ort_path() {
if p.exists() {
env::set_var("ORT_DYLIB_PATH", p);
}
});
}
}
impl Default for AppConfig {
@@ -221,23 +135,14 @@ impl Default for AppConfig {
http_port: 0,
openai_base_url: default_base_url(),
storage: default_storage_kind(),
s3_bucket: None,
s3_endpoint: None,
s3_region: default_s3_region(),
pdf_ingest_mode: default_pdf_ingest_mode(),
reranking_enabled: default_reranking_enabled(),
reranking_pool_size: None,
fastembed_cache_dir: None,
fastembed_show_download_progress: None,
fastembed_max_length: None,
fastembed_model: None,
retrieval_strategy: None,
embedding_backend: EmbeddingBackend::default(),
embedding_pool_size: None,
ingest_max_body_bytes: default_ingest_max_body_bytes(),
ingest_max_files: default_ingest_max_files(),
ingest_max_content_bytes: default_ingest_max_content_bytes(),
ingest_max_context_bytes: default_ingest_max_context_bytes(),
ingest_max_category_bytes: default_ingest_max_category_bytes(),
}
}
}
@@ -254,27 +159,3 @@ pub fn get_config() -> Result<AppConfig, ConfigError> {
config.try_deserialize()
}
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used)]
use super::EmbeddingBackend;
#[test]
fn embedding_backend_defaults_to_fastembed() {
assert_eq!(EmbeddingBackend::default(), EmbeddingBackend::FastEmbed);
}
#[test]
fn embedding_backend_parses_aliases() {
assert_eq!(
"openai".parse::<EmbeddingBackend>().expect("openai"),
EmbeddingBackend::OpenAI
);
assert_eq!(
"fast".parse::<EmbeddingBackend>().expect("fast"),
EmbeddingBackend::FastEmbed
);
}
}
+171 -505
View File
@@ -2,25 +2,44 @@ use std::{
collections::hash_map::DefaultHasher,
hash::{Hash, Hasher},
str::FromStr,
sync::{Arc, Mutex},
thread::available_parallelism,
sync::Arc,
};
use serde::Serialize;
use tracing::warn;
use anyhow::{anyhow, Context, Result};
use async_openai::{types::CreateEmbeddingRequestArgs, Client};
use fastembed::{EmbeddingModel, ModelTrait, TextEmbedding, TextInitOptions};
use tokio::sync::{OwnedSemaphorePermit, Semaphore};
use tokio::sync::Mutex;
use tracing::debug;
use crate::{
error::{AppError, EmbeddingError},
error::AppError,
storage::{db::SurrealDbClient, types::system_settings::SystemSettings},
utils::config::AppConfig,
};
/// Supported embedding backends.
#[allow(clippy::module_name_repetitions)]
pub use crate::utils::config::{EmbeddingBackend, ParseEmbeddingBackendError};
#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)]
pub enum EmbeddingBackend {
#[default]
OpenAI,
FastEmbed,
Hashed,
}
impl std::str::FromStr for EmbeddingBackend {
type Err = anyhow::Error;
fn from_str(s: &str) -> Result<Self, Self::Err> {
match s.to_ascii_lowercase().as_str() {
"openai" => Ok(Self::OpenAI),
"hashed" => Ok(Self::Hashed),
"fastembed" | "fast-embed" | "fast" => Ok(Self::FastEmbed),
other => Err(anyhow!(
"unknown embedding backend '{other}'. Expected 'openai', 'hashed', or 'fastembed'."
)),
}
}
}
/// Wrapper around the chosen embedding backend.
#[allow(clippy::module_name_repetitions)]
@@ -38,7 +57,7 @@ enum EmbeddingInner {
/// Client used to issue embedding requests.
client: Arc<Client<async_openai::config::OpenAIConfig>>,
/// Model identifier for the API.
model: Arc<str>,
model: String,
/// Expected output dimensions.
dimensions: u32,
},
@@ -49,8 +68,8 @@ enum EmbeddingInner {
},
/// Uses `FastEmbed` running locally.
FastEmbed {
/// Pool of `FastEmbed` engines providing bounded-concurrency local embedding.
pool: Arc<FastEmbedPool>,
/// Shared `FastEmbed` model.
model: Arc<Mutex<TextEmbedding>>,
/// Model metadata used for info logging.
model_name: EmbeddingModel,
/// Output vector length.
@@ -58,250 +77,7 @@ enum EmbeddingInner {
},
}
/// Batch size used when re-embedding stored data in bulk. Bounds peak memory and preserves
/// progress logging while still amortising per-call lock/dispatch overhead.
pub const RE_EMBED_BATCH_SIZE: usize = 128;
/// Default FastEmbed model (`BGESmallENV15`) when config and DB do not specify a valid code.
pub const DEFAULT_FASTEMBED_MODEL_CODE: &str = "Xenova/bge-small-en-v1.5";
/// A supported FastEmbed model for admin UI and documentation.
#[derive(Clone, Debug, Serialize)]
pub struct FastEmbedModelOption {
/// HuggingFace-style `model_code` accepted by [`EmbeddingModel::from_str`].
pub model_code: String,
/// Fixed output dimension for this model.
pub dimension: u32,
/// Short human-readable description from fastembed metadata.
pub description: String,
}
/// Lists supported FastEmbed text embedding models (sorted by `model_code`).
#[must_use]
pub fn list_fastembed_embedding_models() -> Vec<FastEmbedModelOption> {
let mut list: Vec<FastEmbedModelOption> = TextEmbedding::list_supported_models()
.into_iter()
.filter_map(|info| {
let dimension = u32::try_from(info.dim).ok()?;
Some(FastEmbedModelOption {
model_code: info.model_code,
dimension,
description: info.description,
})
})
.collect();
list.sort_by(|left, right| left.model_code.cmp(&right.model_code));
list
}
/// Returns true when `code` is a supported FastEmbed `model_code` (HuggingFace-style id).
#[must_use]
pub fn is_valid_fastembed_model_code(code: &str) -> bool {
!code.trim().is_empty() && EmbeddingModel::from_str(code.trim()).is_ok()
}
/// Vector dimension for a supported FastEmbed `model_code`.
///
/// # Errors
///
/// Returns [`EmbeddingError::UnknownModel`] when the code is not recognized.
pub fn fastembed_model_dimension(code: &str) -> Result<u32, EmbeddingError> {
let model = EmbeddingModel::from_str(code.trim())
.map_err(|_| EmbeddingError::UnknownModel(unknown_fastembed_model_message(code)))?;
let dim = EmbeddingModel::get_model_info(&model)
.ok_or_else(|| {
EmbeddingError::Config(format!("fastembed model metadata missing for {code}"))
})?
.dim;
u32::try_from(dim).map_err(|_| {
EmbeddingError::Config(format!("fastembed model dimension {dim} exceeds u32::MAX"))
})
}
/// Resolves the FastEmbed model code to load: config override, then DB, then default.
///
/// When `config.fastembed_model` is set it must be valid. When only the DB value is used and it
/// is not a FastEmbed code (e.g. legacy `text-embedding-3-small`), returns the default model.
///
/// # Errors
///
/// Returns [`EmbeddingError::UnknownModel`] if `config.fastembed_model` is set but invalid.
pub fn resolve_fastembed_model_code(
config: &AppConfig,
settings_model: &str,
) -> Result<String, EmbeddingError> {
if let Some(code) = config.fastembed_model.as_deref() {
let trimmed = code.trim();
if trimmed.is_empty() {
return Err(EmbeddingError::Config(
"fastembed_model must not be empty when set".into(),
));
}
EmbeddingModel::from_str(trimmed)
.map_err(|_| EmbeddingError::UnknownModel(unknown_fastembed_model_message(trimmed)))?;
return Ok(trimmed.to_owned());
}
let trimmed = settings_model.trim();
if is_valid_fastembed_model_code(trimmed) {
return Ok(trimmed.to_owned());
}
if !trimmed.is_empty() {
warn!(
stored_model = trimmed,
default_model = DEFAULT_FASTEMBED_MODEL_CODE,
"system_settings.embedding_model is not a FastEmbed model code; using default"
);
}
Ok(DEFAULT_FASTEMBED_MODEL_CODE.to_owned())
}
/// Persists a FastEmbed-compatible `embedding_model` and `embedding_dimensions` before startup
/// when the active backend is FastEmbed and stored settings still carry OpenAI defaults.
///
/// # Errors
///
/// Returns [`AppError`] if settings cannot be loaded, resolved, or updated.
pub async fn align_fastembed_system_settings(
db: &SurrealDbClient,
config: &AppConfig,
) -> Result<SystemSettings, AppError> {
if config.embedding_backend != EmbeddingBackend::FastEmbed {
return SystemSettings::get_current(db).await;
}
let mut settings = SystemSettings::get_current(db).await?;
let resolved = resolve_fastembed_model_code(config, &settings.embedding_model)?;
let dimension = fastembed_model_dimension(&resolved)?;
if settings.embedding_model == resolved && settings.embedding_dimensions == dimension {
return Ok(settings);
}
tracing::info!(
old_model = %settings.embedding_model,
new_model = %resolved,
old_dimensions = settings.embedding_dimensions,
new_dimensions = dimension,
"Aligning system settings with FastEmbed model"
);
settings.embedding_model = resolved;
settings.embedding_dimensions = dimension;
SystemSettings::update(db, settings).await
}
fn unknown_fastembed_model_message(code: &str) -> String {
let mut codes: Vec<String> = TextEmbedding::list_supported_models()
.into_iter()
.map(|info| info.model_code)
.collect();
codes.sort();
let examples: Vec<&str> = codes.iter().take(6).map(String::as_str).collect();
format!(
"unknown FastEmbed model '{code}' (expected a HuggingFace model_code such as {}). \
Set fastembed_model in config.yaml or update system_settings; \
see docs/configuration.md ({count} models supported)",
examples.join(", "),
count = codes.len()
)
}
/// Default FastEmbed pool size.
///
/// Kept small on purpose: the ONNX runtime already uses intra-op threads per inference, so
/// running many engines concurrently oversubscribes the CPU and each engine duplicates the
/// model weights in memory. Mirrors the reranker pool default.
#[must_use]
pub fn default_embedding_pool_size() -> usize {
available_parallelism()
.map_or(2, |value| value.get().min(2))
.max(1)
}
/// Pool of `FastEmbed` engines enabling bounded-concurrency local embedding.
///
/// A single [`TextEmbedding`] embeds one batch at a time (`&mut self`), so the pool keeps
/// several instances and hands out a distinct idle engine per checkout. The semaphore bounds
/// total in-flight embeds (backpressure); the free list guarantees each active lease holds a
/// different engine — unlike a round-robin index, which can hand the same engine to two callers.
struct FastEmbedPool {
/// Idle engines; one is popped on checkout and returned on lease drop.
engines: Mutex<Vec<Arc<Mutex<TextEmbedding>>>>,
/// Sized to the engine count; gates concurrent checkouts.
semaphore: Arc<Semaphore>,
}
impl FastEmbedPool {
fn new(engines: Vec<Arc<Mutex<TextEmbedding>>>) -> Self {
let permits = engines.len().max(1);
Self {
engines: Mutex::new(engines),
semaphore: Arc::new(Semaphore::new(permits)),
}
}
/// Acquire a permit and borrow a distinct idle engine. The permit guarantees an engine is
/// available, so the pop always succeeds for a correctly sized pool.
async fn checkout(self: &Arc<Self>) -> Result<FastEmbedLease, EmbeddingError> {
let permit = Arc::clone(&self.semaphore)
.acquire_owned()
.await
.map_err(|_| EmbeddingError::Config("embedding pool is closed".into()))?;
let engine = self
.engines
.lock()
.map_err(EmbeddingError::mutex_poisoned)?
.pop()
.ok_or_else(|| EmbeddingError::Config("embedding pool unexpectedly empty".into()))?;
Ok(FastEmbedLease {
pool: Arc::clone(self),
engine,
_permit: permit,
})
}
}
/// Active borrow of a single `FastEmbed` engine; returns it to the pool on drop.
struct FastEmbedLease {
pool: Arc<FastEmbedPool>,
engine: Arc<Mutex<TextEmbedding>>,
/// Released after the engine is returned, unblocking the next checkout.
_permit: OwnedSemaphorePermit,
}
impl FastEmbedLease {
async fn embed(&self, texts: &[String]) -> Result<Vec<Vec<f32>>, EmbeddingError> {
let engine = Arc::clone(&self.engine);
let texts = texts.to_vec();
tokio::task::spawn_blocking(move || -> Result<Vec<Vec<f32>>, EmbeddingError> {
let mut guard = engine.lock().map_err(EmbeddingError::mutex_poisoned)?;
guard.embed(texts, None).map_err(EmbeddingError::fastembed)
})
.await
.map_err(EmbeddingError::from)?
}
}
impl Drop for FastEmbedLease {
fn drop(&mut self) {
if let Ok(mut free) = self.pool.engines.lock() {
free.push(Arc::clone(&self.engine));
}
}
}
async fn run_fastembed(
pool: &Arc<FastEmbedPool>,
texts: &[String],
) -> Result<Vec<Vec<f32>>, EmbeddingError> {
let lease = pool.checkout().await?;
lease.embed(texts).await
}
impl EmbeddingProvider {
#[must_use]
pub fn backend_label(&self) -> &'static str {
match self.inner {
EmbeddingInner::Hashed { .. } => "hashed",
@@ -310,7 +86,6 @@ impl EmbeddingProvider {
}
}
#[must_use]
pub fn dimension(&self) -> usize {
match &self.inner {
EmbeddingInner::Hashed { dimension } | EmbeddingInner::FastEmbed { dimension, .. } => {
@@ -320,28 +95,26 @@ impl EmbeddingProvider {
}
}
#[must_use]
pub fn model_code(&self) -> Option<String> {
match &self.inner {
EmbeddingInner::FastEmbed { model_name, .. } => Some(model_name.to_string()),
EmbeddingInner::OpenAI { model, .. } => Some(model.as_ref().to_owned()),
EmbeddingInner::OpenAI { model, .. } => Some(model.clone()),
EmbeddingInner::Hashed { .. } => None,
}
}
/// Generate an embedding vector for the given text.
///
/// # Errors
///
/// Returns [`EmbeddingError`] if the backend API call fails, FastEmbed initialisation fails,
/// or the backend returns no embedding data.
pub async fn embed(&self, text: &str) -> Result<Vec<f32>, EmbeddingError> {
pub async fn embed(&self, text: &str) -> Result<Vec<f32>> {
match &self.inner {
EmbeddingInner::Hashed { dimension } => Ok(hashed_embedding(text, *dimension)),
EmbeddingInner::FastEmbed { pool, .. } => {
let text = text.to_owned();
let embeddings = run_fastembed(pool, std::slice::from_ref(&text)).await?;
embeddings.into_iter().next().ok_or(EmbeddingError::NoData)
EmbeddingInner::FastEmbed { model, .. } => {
let mut guard = model.lock().await;
let embeddings = guard
.embed(vec![text.to_owned()], None)
.context("generating fastembed vector")?;
embeddings
.into_iter()
.next()
.ok_or_else(|| anyhow!("fastembed returned no embedding for input"))
}
EmbeddingInner::OpenAI {
client,
@@ -349,7 +122,7 @@ impl EmbeddingProvider {
dimensions,
} => {
let request = CreateEmbeddingRequestArgs::default()
.model(model.as_ref())
.model(model.clone())
.input([text])
.dimensions(*dimensions)
.build()?;
@@ -359,7 +132,7 @@ impl EmbeddingProvider {
let embedding = response
.data
.first()
.ok_or(EmbeddingError::NoData)?
.ok_or_else(|| anyhow!("No embedding data received from OpenAI API"))?
.embedding
.clone();
@@ -368,23 +141,20 @@ impl EmbeddingProvider {
}
}
/// Generate embedding vectors for a batch of texts.
///
/// # Errors
///
/// Returns [`EmbeddingError`] if the backend API call fails or returns no embedding data.
/// Returns an empty `Vec` when `texts` is empty.
pub async fn embed_batch(&self, texts: &[String]) -> Result<Vec<Vec<f32>>, EmbeddingError> {
pub async fn embed_batch(&self, texts: Vec<String>) -> Result<Vec<Vec<f32>>> {
match &self.inner {
EmbeddingInner::Hashed { dimension } => Ok(texts
.iter()
.map(|text| hashed_embedding(text, *dimension))
.into_iter()
.map(|text| hashed_embedding(&text, *dimension))
.collect()),
EmbeddingInner::FastEmbed { pool, .. } => {
EmbeddingInner::FastEmbed { model, .. } => {
if texts.is_empty() {
return Ok(Vec::new());
}
run_fastembed(pool, texts).await
let mut guard = model.lock().await;
guard
.embed(texts, None)
.context("generating fastembed batch embeddings")
}
EmbeddingInner::OpenAI {
client,
@@ -396,8 +166,8 @@ impl EmbeddingProvider {
}
let request = CreateEmbeddingRequestArgs::default()
.model(model.as_ref())
.input(texts.to_vec())
.model(model.clone())
.input(texts)
.dimensions(*dimensions)
.build()?;
@@ -414,84 +184,51 @@ impl EmbeddingProvider {
}
}
/// # Errors
///
/// Currently infallible; reserved for future validation.
pub fn new_openai(
client: Arc<Client<async_openai::config::OpenAIConfig>>,
model: impl AsRef<str>,
model: String,
dimensions: u32,
) -> Result<Self, EmbeddingError> {
) -> Result<Self> {
Ok(Self {
inner: EmbeddingInner::OpenAI {
client,
model: Arc::from(model.as_ref()),
model,
dimensions,
},
})
}
/// Initialise a local FastEmbed provider backed by a pool of `pool_size` engines.
///
/// `pool_size` is clamped to at least 1. Larger pools allow concurrent embeds at the cost of
/// `pool_size`× model memory; see [`default_embedding_pool_size`] for guidance.
///
/// # Errors
///
/// Returns [`EmbeddingError`] if the model name is unknown or FastEmbed initialisation fails.
pub async fn new_fastembed(
model_override: Option<String>,
pool_size: usize,
) -> Result<Self, EmbeddingError> {
let pool_size = pool_size.max(1);
pub async fn new_fastembed(model_override: Option<String>) -> Result<Self> {
let model_name = if let Some(code) = model_override {
EmbeddingModel::from_str(code.trim())
.map_err(|_| EmbeddingError::UnknownModel(unknown_fastembed_model_message(&code)))?
EmbeddingModel::from_str(&code).map_err(|err| anyhow!(err))?
} else {
EmbeddingModel::default()
};
let options = TextInitOptions::new(model_name.clone()).with_show_download_progress(true);
let model_name_for_task = model_name.clone();
let model_name_code = model_name.to_string();
let (engines, dimension) =
match tokio::task::spawn_blocking(move || -> Result<_, EmbeddingError> {
let info =
EmbeddingModel::get_model_info(&model_name_for_task).ok_or_else(|| {
EmbeddingError::Config(format!(
"fastembed model metadata missing for {model_name_code}"
))
})?;
let mut engines = Vec::with_capacity(pool_size);
for index in 0..pool_size {
let options = TextInitOptions::new(model_name_for_task.clone())
// Only the first engine reports download progress; the rest reuse the cache.
.with_show_download_progress(index == 0);
let (model, dimension) = tokio::task::spawn_blocking(move || -> Result<_> {
let model =
TextEmbedding::try_new(options).map_err(EmbeddingError::fastembed)?;
engines.push(Arc::new(Mutex::new(model)));
}
Ok((engines, info.dim))
TextEmbedding::try_new(options).context("initialising FastEmbed text model")?;
let info = EmbeddingModel::get_model_info(&model_name_for_task)
.ok_or_else(|| anyhow!("FastEmbed model metadata missing for {model_name_code}"))?;
Ok((model, info.dim))
})
.await
{
Ok(result) => result?,
Err(join_error) => return Err(EmbeddingError::from(join_error)),
};
.context("joining FastEmbed initialisation task")??;
Ok(EmbeddingProvider {
inner: EmbeddingInner::FastEmbed {
pool: Arc::new(FastEmbedPool::new(engines)),
model: Arc::new(Mutex::new(model)),
model_name,
dimension,
},
})
}
/// # Errors
///
/// Currently infallible; reserved for future validation.
pub fn new_hashed(dimension: usize) -> Result<Self, EmbeddingError> {
pub fn new_hashed(dimension: usize) -> Result<Self> {
Ok(EmbeddingProvider {
inner: EmbeddingInner::Hashed {
dimension: dimension.max(1),
@@ -499,44 +236,30 @@ impl EmbeddingProvider {
})
}
/// Creates an embedding provider from persisted settings and bootstrap config.
/// Creates an embedding provider based on application configuration.
///
/// OpenAI/hashed model settings come from [`SystemSettings`]. FastEmbed uses
/// [`resolve_fastembed_model_code`] (config `fastembed_model` overrides DB). The active
/// backend is taken from `config.embedding_backend`; [`SystemSettings::sync_from_embedding_provider`]
/// persists the resolved backend to the database after startup.
///
/// # Errors
///
/// Returns [`EmbeddingError`] if the selected backend cannot be initialised.
pub async fn from_system_settings(
settings: &SystemSettings,
config: &AppConfig,
/// Dispatches to the appropriate constructor based on `config.embedding_backend`:
/// - `OpenAI`: Requires a valid OpenAI client
/// - `FastEmbed`: Uses local embedding model
/// - `Hashed`: Uses deterministic hashed embeddings (for testing)
pub async fn from_config(
config: &crate::utils::config::AppConfig,
openai_client: Option<Arc<Client<async_openai::config::OpenAIConfig>>>,
) -> Result<Self, EmbeddingError> {
let dimensions = settings.embedding_dimensions;
) -> Result<Self> {
use crate::utils::config::EmbeddingBackend;
match config.embedding_backend {
EmbeddingBackend::OpenAI => {
let client = openai_client.ok_or_else(|| {
EmbeddingError::Config(
"openai embedding backend requires an openai client".into(),
)
})?;
Self::new_openai(client, settings.embedding_model.as_str(), dimensions)
let client = openai_client
.ok_or_else(|| anyhow!("OpenAI embedding backend requires an OpenAI client"))?;
// Use defaults that match SystemSettings initial values
Self::new_openai(client, "text-embedding-3-small".to_string(), 1536)
}
EmbeddingBackend::FastEmbed => {
let pool_size = config
.embedding_pool_size
.unwrap_or_else(default_embedding_pool_size);
let model_code = resolve_fastembed_model_code(config, &settings.embedding_model)?;
Self::new_fastembed(Some(model_code), pool_size).await
}
EmbeddingBackend::Hashed => {
let dimension = usize::try_from(dimensions).map_err(|_| {
EmbeddingError::Config("embedding_dimensions exceeds usize::MAX".into())
})?;
Self::new_hashed(dimension)
// Use nomic-embed-text-v1.5 as the default FastEmbed model
Self::new_fastembed(Some("nomic-ai/nomic-embed-text-v1.5".to_string())).await
}
EmbeddingBackend::Hashed => Self::new_hashed(384),
}
}
}
@@ -583,151 +306,94 @@ fn bucket(token: &str, dimension: usize) -> usize {
usize::try_from(hasher.finish()).unwrap_or_default() % safe_dimension
}
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used)]
use super::{
align_fastembed_system_settings, fastembed_model_dimension,
list_fastembed_embedding_models, resolve_fastembed_model_code, EmbeddingError,
DEFAULT_FASTEMBED_MODEL_CODE,
};
use crate::storage::types::system_settings::SystemSettings;
use crate::utils::config::{AppConfig, EmbeddingBackend, ParseEmbeddingBackendError};
use serde_json::json;
#[test]
fn embedding_backend_defaults_to_fastembed() {
assert_eq!(EmbeddingBackend::default(), EmbeddingBackend::FastEmbed);
// Backward compatibility function
pub async fn generate_embedding_with_provider(
provider: &EmbeddingProvider,
input: &str,
) -> Result<Vec<f32>, AppError> {
provider.embed(input).await.map_err(AppError::from)
}
#[test]
fn embedding_backend_as_str_matches_serde_names() {
assert_eq!(EmbeddingBackend::OpenAI.as_str(), "openai");
assert_eq!(EmbeddingBackend::FastEmbed.as_str(), "fastembed");
assert_eq!(EmbeddingBackend::Hashed.as_str(), "hashed");
/// Generates an embedding vector for the given input text using `OpenAI`'s embedding model.
///
/// This function takes a text input and converts it into a numerical vector representation (embedding)
/// using `OpenAI`'s text-embedding-3-small model. These embeddings can be used for semantic similarity
/// comparisons, vector search, and other natural language processing tasks.
///
/// # Arguments
///
/// * `client`: The `OpenAI` client instance used to make API requests.
/// * `input`: The text string to generate embeddings for.
///
/// # Returns
///
/// Returns a `Result` containing either:
/// * `Ok(Vec<f32>)`: A vector of 32-bit floating point numbers representing the text embedding
/// * `Err(ProcessingError)`: An error if the embedding generation fails
///
/// # Errors
///
/// This function can return a `AppError` in the following cases:
/// * If the `OpenAI` API request fails
/// * If the request building fails
/// * If no embedding data is received in the response
#[allow(clippy::module_name_repetitions)]
pub async fn generate_embedding(
client: &async_openai::Client<async_openai::config::OpenAIConfig>,
input: &str,
db: &SurrealDbClient,
) -> Result<Vec<f32>, AppError> {
let model = SystemSettings::get_current(db).await?;
assert_eq!(
serde_json::to_string(&EmbeddingBackend::FastEmbed).expect("serialize"),
"\"fastembed\""
let request = CreateEmbeddingRequestArgs::default()
.model(model.embedding_model)
.dimensions(model.embedding_dimensions)
.input([input])
.build()?;
// Send the request to OpenAI
let response = client.embeddings().create(request).await?;
// Extract the embedding vector
let embedding: Vec<f32> = response
.data
.first()
.ok_or_else(|| AppError::LLMParsing("No embedding data received".into()))?
.embedding
.clone();
Ok(embedding)
}
/// Generates an embedding vector using a specific model and dimension.
///
/// This is used for the re-embedding process where the model and dimensions
/// are known ahead of time and shouldn't be repeatedly fetched from settings.
pub async fn generate_embedding_with_params(
client: &async_openai::Client<async_openai::config::OpenAIConfig>,
input: &str,
model: &str,
dimensions: u32,
) -> Result<Vec<f32>, AppError> {
let request = CreateEmbeddingRequestArgs::default()
.model(model)
.input([input])
.dimensions(dimensions)
.build()?;
let response = client.embeddings().create(request).await?;
let embedding = response
.data
.first()
.ok_or_else(|| AppError::LLMParsing("No embedding data received from API".into()))?
.embedding
.clone();
debug!(
"Embedding was created with {:?} dimensions",
embedding.len()
);
}
#[test]
fn embedding_backend_deserializes_lowercase_values() {
let openai: EmbeddingBackend = serde_json::from_str("\"openai\"").expect("openai");
let fastembed: EmbeddingBackend = serde_json::from_str("\"fastembed\"").expect("fastembed");
let hashed: EmbeddingBackend = serde_json::from_str("\"hashed\"").expect("hashed");
assert_eq!(openai, EmbeddingBackend::OpenAI);
assert_eq!(fastembed, EmbeddingBackend::FastEmbed);
assert_eq!(hashed, EmbeddingBackend::Hashed);
}
#[test]
fn embedding_backend_from_str_accepts_aliases() {
assert_eq!(
"fast-embed"
.parse::<EmbeddingBackend>()
.expect("fast-embed"),
EmbeddingBackend::FastEmbed
);
assert_eq!(
"FASTEMBED".parse::<EmbeddingBackend>().expect("FASTEMBED"),
EmbeddingBackend::FastEmbed
);
assert!(matches!(
"unknown-backend".parse::<EmbeddingBackend>(),
Err(ParseEmbeddingBackendError { .. })
));
}
#[test]
fn list_fastembed_embedding_models_includes_default() {
let models = list_fastembed_embedding_models();
assert!(
models
.iter()
.any(|m| m.model_code == DEFAULT_FASTEMBED_MODEL_CODE),
"catalog should include the default FastEmbed model"
);
}
#[test]
fn resolve_fastembed_model_prefers_config_over_db() {
let config = AppConfig {
fastembed_model: Some("Xenova/bge-base-en-v1.5".into()),
..AppConfig::default()
};
let resolved =
resolve_fastembed_model_code(&config, "text-embedding-3-small").expect("config model");
assert_eq!(resolved, "Xenova/bge-base-en-v1.5");
}
#[test]
fn resolve_fastembed_model_falls_back_from_openai_default() {
let config = AppConfig::default();
let resolved =
resolve_fastembed_model_code(&config, "text-embedding-3-small").expect("default model");
assert_eq!(resolved, DEFAULT_FASTEMBED_MODEL_CODE);
}
#[test]
fn resolve_fastembed_model_rejects_invalid_config_override() {
let config = AppConfig {
fastembed_model: Some("not-a-real-model".into()),
..AppConfig::default()
};
let err = resolve_fastembed_model_code(&config, "Xenova/bge-small-en-v1.5")
.expect_err("invalid config model");
assert!(matches!(err, EmbeddingError::UnknownModel(_)));
}
#[test]
fn fastembed_model_dimension_matches_model_metadata() {
let dim = fastembed_model_dimension(DEFAULT_FASTEMBED_MODEL_CODE).expect("dim");
assert_eq!(dim, 384);
}
#[tokio::test]
async fn align_fastembed_system_settings_replaces_openai_default() -> anyhow::Result<()> {
use crate::storage::db::SurrealDbClient;
use uuid::Uuid;
let db = SurrealDbClient::memory("align_fe", &Uuid::new_v4().to_string()).await?;
db.apply_migrations().await?;
let config = AppConfig {
embedding_backend: EmbeddingBackend::FastEmbed,
..AppConfig::default()
};
let settings = align_fastembed_system_settings(&db, &config).await?;
assert_eq!(settings.embedding_model, DEFAULT_FASTEMBED_MODEL_CODE);
assert_eq!(settings.embedding_dimensions, 384);
Ok(())
}
#[test]
fn system_settings_deserializes_embedding_backend_field() {
let value = json!({
"id": "current",
"registrations_enabled": true,
"require_email_verification": false,
"query_model": "gpt-4o-mini",
"processing_model": "gpt-4o-mini",
"embedding_model": "text-embedding-3-small",
"embedding_dimensions": 1536,
"embedding_backend": "hashed",
"query_system_prompt": "query",
"ingestion_system_prompt": "ingestion",
"image_processing_model": "gpt-4o-mini",
"image_processing_prompt": "image",
"voice_processing_model": "whisper-1",
});
let settings: SystemSettings =
serde_json::from_value(value).expect("deserialize system settings");
assert_eq!(settings.embedding_backend, Some(EmbeddingBackend::Hashed));
}
Ok(embedding)
}
-155
View File
@@ -1,155 +0,0 @@
use thiserror::Error;
use super::config::AppConfig;
/// Errors raised when validating ingestion payloads against configured limits.
#[derive(Error, Debug, Clone, PartialEq, Eq)]
pub enum IngestValidationError {
/// The payload exceeds a configured size limit (content, context, or category).
#[error("payload too large: {0}")]
PayloadTooLarge(String),
/// The request violates a non-size constraint (e.g., too many files).
#[error("bad request: {0}")]
BadRequest(String),
}
/// Validates ingestion input against configured limits.
///
/// Checks file count, content size, context size, and category length.
///
/// # Errors
///
/// Returns `IngestValidationError::BadRequest` if the file count exceeds the maximum.
/// Returns `IngestValidationError::PayloadTooLarge` if content, context, or
/// category exceed their configured byte limits.
pub fn validate_ingest_input(
config: &AppConfig,
content: Option<&str>,
ctx: &str,
category: &str,
file_count: usize,
) -> Result<(), IngestValidationError> {
let content_bytes = content.map_or(0, str::len);
let text_field_bytes = content_bytes
.saturating_add(ctx.len())
.saturating_add(category.len());
if text_field_bytes > config.ingest_max_body_bytes {
return Err(IngestValidationError::PayloadTooLarge(format!(
"request text fields exceed maximum allowed body size of {} bytes",
config.ingest_max_body_bytes
)));
}
if file_count > config.ingest_max_files {
return Err(IngestValidationError::BadRequest(format!(
"too many files: maximum allowed is {}",
config.ingest_max_files
)));
}
if let Some(content) = content {
if content.len() > config.ingest_max_content_bytes {
return Err(IngestValidationError::PayloadTooLarge(format!(
"content is too large: maximum allowed is {} bytes",
config.ingest_max_content_bytes
)));
}
}
if ctx.len() > config.ingest_max_context_bytes {
return Err(IngestValidationError::PayloadTooLarge(format!(
"context is too large: maximum allowed is {} bytes",
config.ingest_max_context_bytes
)));
}
if category.len() > config.ingest_max_category_bytes {
return Err(IngestValidationError::PayloadTooLarge(format!(
"category is too large: maximum allowed is {} bytes",
config.ingest_max_category_bytes
)));
}
Ok(())
}
#[cfg(test)]
mod tests {
#![allow(clippy::expect_used, clippy::must_use_candidate)]
use super::*;
#[test]
fn validate_ingest_input_rejects_too_many_files() {
let config = AppConfig {
ingest_max_files: 1,
..Default::default()
};
let result = validate_ingest_input(&config, Some("ok"), "ctx", "cat", 2);
assert!(matches!(result, Err(IngestValidationError::BadRequest(_))));
}
#[test]
fn validate_ingest_input_rejects_oversized_content() {
let config = AppConfig {
ingest_max_content_bytes: 4,
..Default::default()
};
let result = validate_ingest_input(&config, Some("12345"), "ctx", "cat", 0);
assert!(matches!(
result,
Err(IngestValidationError::PayloadTooLarge(_))
));
}
#[test]
fn validate_ingest_input_rejects_oversized_context() {
let config = AppConfig {
ingest_max_context_bytes: 2,
..Default::default()
};
let result = validate_ingest_input(&config, None, "long", "cat", 0);
assert!(matches!(
result,
Err(IngestValidationError::PayloadTooLarge(_))
));
}
#[test]
fn validate_ingest_input_rejects_oversized_category() {
let config = AppConfig {
ingest_max_category_bytes: 2,
..Default::default()
};
let result = validate_ingest_input(&config, None, "ok", "long", 0);
assert!(matches!(
result,
Err(IngestValidationError::PayloadTooLarge(_))
));
}
#[test]
fn validate_ingest_input_accepts_valid_payload() {
let config = AppConfig::default();
let result = validate_ingest_input(&config, Some("ok"), "ctx", "cat", 1);
assert!(result.is_ok());
}
#[test]
fn validate_ingest_input_rejects_oversized_text_fields() {
let config = AppConfig {
ingest_max_body_bytes: 10,
..Default::default()
};
let result = validate_ingest_input(&config, Some("123456"), "ctx", "cat", 0);
assert!(matches!(
result,
Err(IngestValidationError::PayloadTooLarge(_))
));
}
}
-2
View File
@@ -1,5 +1,3 @@
pub mod config;
pub mod embedding;
pub mod ingest_limits;
pub mod serde_helpers;
pub mod template_engine;
-82
View File
@@ -1,82 +0,0 @@
use chrono::{DateTime, Utc};
use serde::de::{self, Visitor};
use serde::{Deserialize, Deserializer, Serialize, Serializer};
use std::fmt;
use surrealdb::sql::Thing;
struct FlexibleIdVisitor;
impl<'de> Visitor<'de> for FlexibleIdVisitor {
type Value = String;
fn expecting(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
formatter.write_str("a string or a Thing")
}
fn visit_str<E>(self, value: &str) -> Result<Self::Value, E>
where
E: de::Error,
{
Ok(value.to_string())
}
fn visit_string<E>(self, value: String) -> Result<Self::Value, E>
where
E: de::Error,
{
Ok(value)
}
fn visit_map<A>(self, map: A) -> Result<Self::Value, A::Error>
where
A: de::MapAccess<'de>,
{
let thing = Thing::deserialize(de::value::MapAccessDeserializer::new(map))?;
Ok(thing.id.to_raw())
}
}
pub fn deserialize_flexible_id<'de, D>(deserializer: D) -> Result<String, D::Error>
where
D: Deserializer<'de>,
{
deserializer.deserialize_any(FlexibleIdVisitor)
}
pub fn serialize_datetime<S>(date: &DateTime<Utc>, serializer: S) -> Result<S::Ok, S::Error>
where
S: Serializer,
{
Into::<surrealdb::sql::Datetime>::into(*date).serialize(serializer)
}
pub fn deserialize_datetime<'de, D>(deserializer: D) -> Result<DateTime<Utc>, D::Error>
where
D: Deserializer<'de>,
{
let dt = surrealdb::sql::Datetime::deserialize(deserializer)?;
Ok(DateTime::<Utc>::from(dt))
}
pub fn serialize_option_datetime<S>(
date: &Option<DateTime<Utc>>,
serializer: S,
) -> Result<S::Ok, S::Error>
where
S: Serializer,
{
match date {
Some(dt) => serializer.serialize_some(&Into::<surrealdb::sql::Datetime>::into(*dt)),
None => serializer.serialize_none(),
}
}
pub fn deserialize_option_datetime<'de, D>(
deserializer: D,
) -> Result<Option<DateTime<Utc>>, D::Error>
where
D: Deserializer<'de>,
{
let value = Option::<surrealdb::sql::Datetime>::deserialize(deserializer)?;
Ok(value.map(DateTime::<Utc>::from))
}
+3 -34
View File
@@ -19,52 +19,21 @@ pub enum TemplateEngine {
Embedded(Arc<Environment<'static>>),
}
#[allow(clippy::module_name_repetitions)]
#[macro_export]
macro_rules! create_template_engine {
// Single path argument
($relative_path:expr) => {
$crate::create_template_engine!($relative_path, Option::<&str>::None)
};
// Path + Fallback argument
($relative_path:expr, $fallback_path:expr) => {{
// Macro takes the relative path to the templates dir as input
($relative_path:expr) => {{
// Code for debug builds (AutoReload)
#[cfg(debug_assertions)]
{
// These lines execute in the CALLING crate's context
let crate_dir = std::path::PathBuf::from(env!("CARGO_MANIFEST_DIR"));
let template_path = crate_dir.join($relative_path);
let fallback_path = $fallback_path.map(|p| crate_dir.join(p));
let reloader = $crate::utils::template_engine::AutoReloader::new(move |notifier| {
let mut env = $crate::utils::template_engine::Environment::new();
let loader_primary = $crate::utils::template_engine::path_loader(&template_path);
// Clone fallback_path for the closure
let fallback = fallback_path.clone();
env.set_loader(move |name| match loader_primary(name) {
Ok(Some(tmpl)) => Ok(Some(tmpl)),
Ok(None) => {
if let Some(ref fb_path) = fallback {
let loader_fallback =
$crate::utils::template_engine::path_loader(fb_path);
loader_fallback(name)
} else {
Ok(None)
}
}
Err(e) => Err(e),
});
env.set_loader($crate::utils::template_engine::path_loader(&template_path));
notifier.set_fast_reload(true);
notifier.watch_path(&template_path, true);
if let Some(ref fb) = fallback_path {
notifier.watch_path(fb, true);
}
// Add contrib filters/functions
$crate::utils::template_engine::minijinja_contrib::add_to_environment(&mut env);
Ok(env)
+20 -20
View File
@@ -3,10 +3,10 @@
"devenv": {
"locked": {
"dir": "src/modules",
"lastModified": 1771066302,
"lastModified": 1761839147,
"owner": "cachix",
"repo": "devenv",
"rev": "1b355dec9bddbaddbe4966d6fc30d7aa3af8575b",
"rev": "bb7849648b68035f6b910120252c22b28195cf54",
"type": "github"
},
"original": {
@@ -22,10 +22,10 @@
"rust-analyzer-src": "rust-analyzer-src"
},
"locked": {
"lastModified": 1771052630,
"lastModified": 1761893049,
"owner": "nix-community",
"repo": "fenix",
"rev": "d0555da98576b8611c25df0c208e51e9a182d95f",
"rev": "c2ac9a5c0d6d16630c3b225b874bd14528d1abe6",
"type": "github"
},
"original": {
@@ -37,14 +37,14 @@
"flake-compat": {
"flake": false,
"locked": {
"lastModified": 1767039857,
"owner": "NixOS",
"lastModified": 1761588595,
"owner": "edolstra",
"repo": "flake-compat",
"rev": "5edf11c44bc78a0d334f6334cdaf7d60d732daab",
"rev": "f387cd2afec9419c8ee37694406ca490c3f34ee5",
"type": "github"
},
"original": {
"owner": "NixOS",
"owner": "edolstra",
"repo": "flake-compat",
"type": "github"
}
@@ -58,10 +58,10 @@
]
},
"locked": {
"lastModified": 1770726378,
"lastModified": 1760663237,
"owner": "cachix",
"repo": "git-hooks.nix",
"rev": "5eaaedde414f6eb1aea8b8525c466dc37bba95ae",
"rev": "ca5b894d3e3e151ffc1db040b6ce4dcc75d31c37",
"type": "github"
},
"original": {
@@ -78,10 +78,10 @@
]
},
"locked": {
"lastModified": 1762808025,
"lastModified": 1709087332,
"owner": "hercules-ci",
"repo": "gitignore.nix",
"rev": "cb5e3fdca1de58ccbc3ef53de65bd372b48f567c",
"rev": "637db329424fd7e46cf4185293b9cc8c88c95394",
"type": "github"
},
"original": {
@@ -92,10 +92,10 @@
},
"nixpkgs": {
"locked": {
"lastModified": 1771008912,
"lastModified": 1761672384,
"owner": "nixos",
"repo": "nixpkgs",
"rev": "a82ccc39b39b621151d6732718e3e250109076fa",
"rev": "08dacfca559e1d7da38f3cf05f1f45ee9bfd213c",
"type": "github"
},
"original": {
@@ -107,10 +107,10 @@
},
"nixpkgs_2": {
"locked": {
"lastModified": 1770843696,
"lastModified": 1761880412,
"owner": "nixos",
"repo": "nixpkgs",
"rev": "2343bbb58f99267223bc2aac4fc9ea301a155a16",
"rev": "a7fc11be66bdfb5cdde611ee5ce381c183da8386",
"type": "github"
},
"original": {
@@ -135,10 +135,10 @@
"rust-analyzer-src": {
"flake": false,
"locked": {
"lastModified": 1771007332,
"lastModified": 1761849405,
"owner": "rust-lang",
"repo": "rust-analyzer",
"rev": "bbc84d335fbbd9b3099d3e40c7469ee57dbd1873",
"rev": "f7de8ae045a5fe80f1203c5a1c3015b05f7c3550",
"type": "github"
},
"original": {
@@ -155,10 +155,10 @@
]
},
"locked": {
"lastModified": 1771038269,
"lastModified": 1761878277,
"owner": "oxalica",
"repo": "rust-overlay",
"rev": "d7a86c8a4df49002446737603a3e0d7ef91a9637",
"rev": "6604534e44090c917db714faa58d47861657690c",
"type": "github"
},
"original": {
+3 -30
View File
@@ -4,59 +4,32 @@
config,
inputs,
...
}:
let
ortVersion = lib.removeSuffix "\n" (builtins.readFile "${toString ./.}/ort-version");
_ortVersionCheck =
if pkgs.onnxruntime.version == ortVersion
then null
else
throw "pkgs.onnxruntime.version (${pkgs.onnxruntime.version}) must match ort-version (${ortVersion})";
in {
}: {
cachix.enable = false;
packages = [
pkgs.openssl
pkgs.nodejs
pkgs.watchman
pkgs.vscode-langservers-extracted
pkgs.cargo-dist
pkgs.cargo-xwin
pkgs.clang
pkgs.onnxruntime
pkgs.cargo-watch
pkgs.tailwindcss_4
];
languages.rust = {
enable = true;
channel = "stable";
version = "1.91.1";
components = ["rustc" "clippy" "rustfmt" "cargo" "rust-analyzer"];
channel = "nightly";
targets = ["x86_64-unknown-linux-gnu" "x86_64-pc-windows-msvc"];
mold.enable = true;
};
env = {
ORT_DYLIB_PATH = "${pkgs.onnxruntime}/lib/libonnxruntime.so";
S3_ENDPOINT = "http://127.0.0.1:19000";
S3_BUCKET = "minne-tests";
MINNE_TEST_S3_ENDPOINT = "http://127.0.0.1:19000";
MINNE_TEST_S3_BUCKET = "minne-tests";
};
services.minio = {
enable = true;
listenAddress = "127.0.0.1:19000";
consoleAddress = "127.0.0.1:19001";
buckets = ["minne-tests"];
accessKey = "minioadmin";
secretKey = "minioadmin";
region = "us-east-1";
};
processes = {
surreal_db.exec = "docker run --rm --pull always -p 8000:8000 --net=host --user $(id -u) -v $(pwd)/database:/database surrealdb/surrealdb:v2.6.5-dev start rocksdb:/database/database.db --user root_user --pass root_password";
tailwind.exec = "tailwindcss --cwd html-router -i app.css -o assets/style.css --watch=always";
surreal_db.exec = "docker run --rm --pull always -p 8000:8000 --net=host --user $(id -u) -v $(pwd)/database:/database surrealdb/surrealdb:latest-dev start rocksdb:/database/database.db --user root_user --pass root_password";
};
}
+2 -4
View File
@@ -4,7 +4,7 @@ members = ["cargo:."]
# Config for 'dist'
[dist]
# The preferred dist version to use in CI (Cargo.toml SemVer syntax)
cargo-dist-version = "0.30.3"
cargo-dist-version = "0.30.0"
# CI backends to support
ci = "github"
# Extra static files to include in each App (path relative to this Cargo.toml's dir)
@@ -12,13 +12,11 @@ include = ["lib"]
# The installers to generate for each app
installers = []
# Target platforms to build apps for (Rust target-triple syntax)
targets = ["aarch64-apple-darwin", "x86_64-apple-darwin", "x86_64-unknown-linux-gnu", "x86_64-pc-windows-msvc"]
targets = ["aarch64-apple-darwin", "x86_64-unknown-linux-gnu", "x86_64-pc-windows-msvc"]
# Skip checking whether the specified configuration files are up to date
allow-dirty = ["ci"]
[dist.github-custom-runners]
aarch64-apple-darwin = "macos-latest"
x86_64-apple-darwin = "macos-15-intel"
x86_64-unknown-linux-gnu = "ubuntu-22.04"
x86_64-unknown-linux-musl = "ubuntu-22.04"
x86_64-pc-windows-msvc = "windows-latest"
+1 -1
View File
@@ -22,7 +22,7 @@ services:
command: ["sh", "-c", "echo 'Waiting for SurrealDB to start...' && sleep 10 && echo 'Starting application...' && /usr/local/bin/main"]
surrealdb:
image: surrealdb/surrealdb:v2.6.5
image: surrealdb/surrealdb:latest
container_name: minne_surrealdb
ports:
- "8000:8000"

Some files were not shown because too many files have changed in this diff Show More