starred/minne
1
0
Fork 0
mirror of https://github.com/perstarkse/minne.git synced 2026-03-12 05:22:01 +01:00
Code Issues Packages Projects Releases 10 Wiki Activity

feat: system prompt customisable

Browse Source
This commit is contained in:
Per Stark
2025-03-23 22:20:57 +01:00
parent 334a0e3ea6
commit 6bb6b8c69a
19 changed files with 493 additions and 141 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
crates/common/src/storage/types/mod.rs
View File

@@ -8,6 +8,7 @@ pub mod ingestion_task;
pub mod knowledge_entity;
pub mod knowledge_relationship;
pub mod message;
pub mod system_prompts;
pub mod system_settings;
pub mod text_chunk;
pub mod text_content;

56
crates/common/src/storage/types/system_prompts.rs Normal file
View File

@@ -0,0 +1,56 @@
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
2. Answer user questions based on this information
3. Provide clear, concise, and accurate responses
4. When referencing information, briefly mention which knowledge entity it came from
5. If the provided context doesn't contain enough information to answer the question confidently, clearly state this
6. If only partial information is available, explain what you can answer and what information is missing
7. Avoid making assumptions or providing information not supported by the context
8. Output the references to the documents. Use the UUIDs and make sure they are correct!
Remember:
- Be direct and honest about the limitations of your knowledge
- Cite the relevant knowledge entities when providing information, but only provide the UUIDs in the reference array
- If you need to combine information from multiple entities, explain how they connect
- Don't speculate beyond what's provided in the context
Example response formats:
"Based on [Entity Name], [answer...]"
"I found relevant information in multiple entries: [explanation...]"
"I apologize, but the provided context doesn't contain information about [topic]""#;
pub static DEFAULT_INGRESS_ANALYSIS_SYSTEM_PROMPT: &str = r#"You are an AI assistant. You will receive a text content, along with user instructions 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 instructions 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:
{
"knowledge_entities": [
{
"key": "unique-key-1",
"name": "Entity Name",
"description": "A detailed description of the entity.",
"entity_type": "TypeOfEntity"
},
// More entities...
],
"relationships": [
{
"type": "RelationshipType",
"source": "unique-key-1 or UUID from existing database",
"target": "unique-key-1 or UUID from existing database"
},
// More relationships...
]
}
Guidelines:
1. Do NOT generate any IDs or UUIDs. Use a unique `key` for each knowledge entity.
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"
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."#;

21
crates/common/src/storage/types/system_settings.rs
View File

@@ -1,5 +1,6 @@
use crate::storage::types::file_info::deserialize_flexible_id;
use serde::{Deserialize, Serialize};
use uuid::Uuid;
use crate::{error::AppError, storage::db::SurrealDbClient};
@@ -9,6 +10,10 @@ pub struct SystemSettings {
pub id: String,
pub registrations_enabled: bool,
pub require_email_verification: bool,
pub query_model: String,
pub processing_model: String,
pub query_system_prompt: String,
pub ingestion_system_prompt: String,
}
impl SystemSettings {
@@ -22,6 +27,10 @@ impl SystemSettings {
id: "current".to_string(),
registrations_enabled: true,
require_email_verification: false,
query_model: "gpt-4o-mini".to_string(),
processing_model: "gpt-4o-mini".to_string(),
query_system_prompt: crate::storage::types::system_prompts::DEFAULT_QUERY_SYSTEM_PROMPT.to_string(),
ingestion_system_prompt: crate::storage::types::system_prompts::DEFAULT_INGRESS_ANALYSIS_SYSTEM_PROMPT.to_string(),
})
.await?;
@@ -53,4 +62,16 @@ impl SystemSettings {
"Something went wrong updating the settings".into(),
))
}
pub fn new() -> Self {
Self {
id: Uuid::new_v4().to_string(),
query_system_prompt: crate::storage::types::system_prompts::DEFAULT_QUERY_SYSTEM_PROMPT.to_string(),
ingestion_system_prompt: crate::storage::types::system_prompts::DEFAULT_INGRESS_ANALYSIS_SYSTEM_PROMPT.to_string(),
query_model: "gpt-4o-mini".to_string(),
processing_model: "gpt-4o-mini".to_string(),
registrations_enabled: true,
require_email_verification: false,
}
}
}

9
crates/composite-retrieval/src/answer_retrieval.rs
View File

@@ -13,6 +13,7 @@ use common::{
types::{
knowledge_entity::KnowledgeEntity,
message::{format_history, Message},
system_settings::SystemSettings,
},
},
};
@@ -63,11 +64,12 @@ pub async fn get_answer_with_references(
user_id: &str,
) -> Result<Answer, AppError> {
let entities = retrieve_entities(surreal_db_client, openai_client, query, user_id).await?;
let settings = SystemSettings::get_current(surreal_db_client).await?;
let entities_json = format_entities_json(&entities);
let user_message = create_user_message(&entities_json, query);
let request = create_chat_request(user_message)?;
let request = create_chat_request(user_message, &settings)?;
let response = openai_client.chat().create(request).await?;
let llm_response = process_llm_response(response).await?;
@@ -139,6 +141,7 @@ pub fn create_user_message_with_history(
pub fn create_chat_request(
user_message: String,
settings: &SystemSettings,
) -> Result<CreateChatCompletionRequest, OpenAIError> {
let response_format = ResponseFormat::JsonSchema {
json_schema: ResponseFormatJsonSchema {
@@ -150,11 +153,11 @@ pub fn create_chat_request(
};
CreateChatCompletionRequestArgs::default()
.model("gpt-4o-mini")
.model(&settings.query_model)
.temperature(0.2)
.max_tokens(3048u32)
.messages([
ChatCompletionRequestSystemMessage::from(QUERY_SYSTEM_PROMPT).into(),
ChatCompletionRequestSystemMessage::from(settings.query_system_prompt.clone()).into(),
ChatCompletionRequestUserMessage::from(user_message).into(),
])
.response_format(response_format)

26
crates/composite-retrieval/src/answer_retrieval_helper.rs
View File

@@ -1,29 +1,7 @@
use common::storage::types::system_prompts::DEFAULT_QUERY_SYSTEM_PROMPT;
use serde_json::{json, Value};
pub static 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
2. Answer user questions based on this information
3. Provide clear, concise, and accurate responses
4. When referencing information, briefly mention which knowledge entity it came from
5. If the provided context doesn't contain enough information to answer the question confidently, clearly state this
6. If only partial information is available, explain what you can answer and what information is missing
7. Avoid making assumptions or providing information not supported by the context
8. Output the references to the documents. Use the UUIDs and make sure they are correct!
Remember:
- Be direct and honest about the limitations of your knowledge
- Cite the relevant knowledge entities when providing information, but only provide the UUIDs in the reference array
- If you need to combine information from multiple entities, explain how they connect
- Don't speculate beyond what's provided in the context
Example response formats:
"Based on [Entity Name], [answer...]"
"I found relevant information in multiple entries: [explanation...]"
"I apologize, but the provided context doesn't contain information about [topic]"
"#;
pub static QUERY_SYSTEM_PROMPT: &str = DEFAULT_QUERY_SYSTEM_PROMPT;
pub fn get_query_response_schema() -> Value {
json!({

3
crates/composite-retrieval/src/lib.rs
View File

@@ -13,6 +13,7 @@ use common::{
use futures::future::{try_join, try_join_all};
use graph::{find_entities_by_relationship_by_id, find_entities_by_source_ids};
use std::collections::HashMap;
use tracing::info;
use vector::find_items_by_vector_similarity;
/// Performs a comprehensive knowledge entity retrieval using multiple search strategies
@@ -42,8 +43,6 @@ pub async fn retrieve_entities(
query: &str,
user_id: &str,
) -> Result<Vec<KnowledgeEntity>, AppError> {
// info!("Received input: {:?}", query);
let (items_from_knowledge_entity_similarity, closest_chunks) = try_join(
find_items_by_vector_similarity(
10,

1
crates/composite-retrieval/src/vector.rs
View File

@@ -1,6 +1,7 @@
use surrealdb::{engine::any::Any, Surreal};
use common::{error::AppError, utils::embedding::generate_embedding};
use tracing::info;
/// Compares vectors and retrieves a number of items from the specified table.
///

14
crates/html-router/src/lib.rs
View File

@@ -22,7 +22,10 @@ use routes::{
change_password, delete_account, set_api_key, show_account_page, show_change_password,
update_timezone,
},
admin_panel::{show_admin_panel, toggle_registration_status},
admin_panel::{
patch_ingestion_prompt, patch_query_prompt, show_admin_panel, show_edit_ingestion_prompt,
show_edit_system_prompt, toggle_registration_status, update_model_settings,
},
chat::{
message_response_stream::get_response_stream, new_chat_user_message, new_user_message,
references::show_reference_tooltip, show_chat_base, show_existing_chat,
@@ -107,9 +110,16 @@ where
"/knowledge-relationship/:id",
delete(delete_knowledge_relationship),
)
.route("/account", get(show_account_page))
// Admin page
.route("/admin", get(show_admin_panel))
.route("/toggle-registrations", patch(toggle_registration_status))
.route("/update-model-settings", patch(update_model_settings))
.route("/edit-query-prompt", get(show_edit_system_prompt))
.route("/update-query-prompt", patch(patch_query_prompt))
.route("/edit-ingestion-prompt", get(show_edit_ingestion_prompt))
.route("/update-ingestion-prompt", patch(patch_ingestion_prompt))
// User account page
.route("/account", get(show_account_page))
.route("/set-api-key", post(set_api_key))
.route("/update-timezone", patch(update_timezone))
.route(

172
crates/html-router/src/routes/admin_panel.rs
View File

@@ -5,7 +5,12 @@ use crate::{
middleware_auth::RequireUser,
template_response::{HtmlError, TemplateResponse},
};
use common::storage::types::{analytics::Analytics, system_settings::SystemSettings, user::User};
use common::storage::types::{
analytics::Analytics,
system_prompts::{DEFAULT_INGRESS_ANALYSIS_SYSTEM_PROMPT, DEFAULT_QUERY_SYSTEM_PROMPT},
system_settings::SystemSettings,
user::User,
};
use crate::html_state::HtmlState;
@@ -15,6 +20,7 @@ pub struct AdminPanelData {
settings: SystemSettings,
analytics: Analytics,
users: i64,
default_query_prompt: String,
}
pub async fn show_admin_panel(
@@ -32,6 +38,7 @@ pub async fn show_admin_panel(
settings,
analytics,
users: users_count,
default_query_prompt: DEFAULT_QUERY_SYSTEM_PROMPT.to_string(),
},
))
}
@@ -85,3 +92,166 @@ pub async fn toggle_registration_status(
},
))
}
#[derive(Deserialize)]
pub struct ModelSettingsInput {
query_model: String,
processing_model: String,
}
#[derive(Serialize)]
pub struct ModelSettingsData {
settings: SystemSettings,
}
pub async fn update_model_settings(
State(state): State<HtmlState>,
RequireUser(user): RequireUser,
Form(input): Form<ModelSettingsInput>,
) -> Result<impl IntoResponse, HtmlError> {
// Early return if the user is not admin
if !user.admin {
return Ok(TemplateResponse::redirect("/"));
};
let current_settings = SystemSettings::get_current(&state.db).await?;
let new_settings = SystemSettings {
query_model: input.query_model,
processing_model: input.processing_model,
..current_settings
};
SystemSettings::update(&state.db, new_settings.clone()).await?;
Ok(TemplateResponse::new_partial(
"auth/admin_panel.html",
"model_settings_form",
ModelSettingsData {
settings: new_settings,
},
))
}
#[derive(Serialize)]
pub struct SystemPromptEditData {
settings: SystemSettings,
default_query_prompt: String,
}
pub async fn show_edit_system_prompt(
State(state): State<HtmlState>,
RequireUser(user): RequireUser,
) -> Result<impl IntoResponse, HtmlError> {
// Early return if the user is not admin
if !user.admin {
return Ok(TemplateResponse::redirect("/"));
};
let settings = SystemSettings::get_current(&state.db).await?;
Ok(TemplateResponse::new_template(
"admin/edit_query_prompt_modal.html",
SystemPromptEditData {
settings,
default_query_prompt: DEFAULT_QUERY_SYSTEM_PROMPT.to_string(),
},
))
}
#[derive(Deserialize)]
pub struct SystemPromptUpdateInput {
query_system_prompt: String,
}
#[derive(Serialize)]
pub struct SystemPromptSectionData {
settings: SystemSettings,
}
pub async fn patch_query_prompt(
State(state): State<HtmlState>,
RequireUser(user): RequireUser,
Form(input): Form<SystemPromptUpdateInput>,
) -> Result<impl IntoResponse, HtmlError> {
// Early return if the user is not admin
if !user.admin {
return Ok(TemplateResponse::redirect("/"));
};
let current_settings = SystemSettings::get_current(&state.db).await?;
let new_settings = SystemSettings {
query_system_prompt: input.query_system_prompt,
..current_settings.clone()
};
SystemSettings::update(&state.db, new_settings.clone()).await?;
Ok(TemplateResponse::new_partial(
"auth/admin_panel.html",
"system_prompt_section",
SystemPromptSectionData {
settings: new_settings,
},
))
}
#[derive(Serialize)]
pub struct IngestionPromptEditData {
settings: SystemSettings,
default_ingestion_prompt: String,
}
pub async fn show_edit_ingestion_prompt(
State(state): State<HtmlState>,
RequireUser(user): RequireUser,
) -> Result<impl IntoResponse, HtmlError> {
// Early return if the user is not admin
if !user.admin {
return Ok(TemplateResponse::redirect("/"));
};
let settings = SystemSettings::get_current(&state.db).await?;
Ok(TemplateResponse::new_template(
"admin/edit_ingestion_prompt_modal.html",
IngestionPromptEditData {
settings,
default_ingestion_prompt: DEFAULT_INGRESS_ANALYSIS_SYSTEM_PROMPT.to_string(),
},
))
}
#[derive(Deserialize)]
pub struct IngestionPromptUpdateInput {
ingestion_system_prompt: String,
}
pub async fn patch_ingestion_prompt(
State(state): State<HtmlState>,
RequireUser(user): RequireUser,
Form(input): Form<IngestionPromptUpdateInput>,
) -> Result<impl IntoResponse, HtmlError> {
// Early return if the user is not admin
if !user.admin {
return Ok(TemplateResponse::redirect("/"));
};
let current_settings = SystemSettings::get_current(&state.db).await?;
let new_settings = SystemSettings {
ingestion_system_prompt: input.ingestion_system_prompt,
..current_settings.clone()
};
SystemSettings::update(&state.db, new_settings.clone()).await?;
Ok(TemplateResponse::new_partial(
"auth/admin_panel.html",
"system_prompt_section",
SystemPromptSectionData {
settings: new_settings,
},
))
}

9
crates/html-router/src/routes/chat/message_response_stream.rs
View File

@@ -34,6 +34,7 @@ use common::storage::{
conversation::Conversation,
message::{Message, MessageRole},
user::User,
system_settings::SystemSettings,
},
};
@@ -139,7 +140,13 @@ pub async fn get_response_stream(
let entities_json = format_entities_json(&entities);
let formatted_user_message =
create_user_message_with_history(&entities_json, &history, &user_message.content);
let request = match create_chat_request(formatted_user_message) {
let settings = match SystemSettings::get_current(&state.db).await {
Ok(s) => s,
Err(_) => {
return Sse::new(create_error_stream("Failed to retrieve system settings"));
}
};
let request = match create_chat_request(formatted_user_message, &settings) {
Ok(req) => req,
Err(..) => {
return Sse::new(create_error_stream("Failed to create chat request"));

51
crates/ingestion-pipeline/src/enricher.rs
View File

@@ -10,11 +10,17 @@ use async_openai::{
};
use common::{
error::AppError,
storage::{db::SurrealDbClient, types::knowledge_entity::KnowledgeEntity},
storage::{
db::SurrealDbClient,
types::{
knowledge_entity::KnowledgeEntity,
system_settings::SystemSettings,
},
},
};
use composite_retrieval::retrieve_entities;
use serde_json::json;
use tracing::debug;
use tracing::{debug, info};
use crate::{
types::llm_enrichment_result::LLMEnrichmentResult,
@@ -44,11 +50,13 @@ impl IngestionEnricher {
text: &str,
user_id: &str,
) -> Result<LLMEnrichmentResult, AppError> {
info!("getting similar entitities");
let similar_entities = self
.find_similar_entities(category, instructions, text, user_id)
.await?;
info!("got similar entitities");
let llm_request =
self.prepare_llm_request(category, instructions, text, &similar_entities)?;
self.prepare_llm_request(category, instructions, text, &similar_entities).await?;
self.perform_analysis(llm_request).await
}
@@ -67,13 +75,15 @@ impl IngestionEnricher {
retrieve_entities(&self.db_client, &self.openai_client, &input_text, user_id).await
}
fn prepare_llm_request(
async fn prepare_llm_request(
&self,
category: &str,
instructions: &str,
text: &str,
similar_entities: &[KnowledgeEntity],
) -> Result<CreateChatCompletionRequest, OpenAIError> {
) -> Result<CreateChatCompletionRequest, AppError> {
let settings = SystemSettings::get_current(&self.db_client).await?;
let entities_json = json!(similar_entities
.iter()
.map(|entity| {
@@ -103,8 +113,8 @@ impl IngestionEnricher {
},
};
CreateChatCompletionRequestArgs::default()
.model("gpt-4o-mini")
let request = CreateChatCompletionRequestArgs::default()
.model(&settings.processing_model)
.temperature(0.2)
.max_tokens(3048u32)
.messages([
@@ -112,30 +122,27 @@ impl IngestionEnricher {
ChatCompletionRequestUserMessage::from(user_message).into(),
])
.response_format(response_format)
.build()
.build()?;
Ok(request)
}
async fn perform_analysis(
&self,
request: CreateChatCompletionRequest,
) -> Result<LLMEnrichmentResult, AppError> {
let response = self.openai_client.chat().create(request).await?;
debug!("Received LLM response: {:?}", response);
response
let content = response
.choices
.first()
.and_then(|choice| choice.message.content.as_ref())
.ok_or(AppError::LLMParsing(
"No content found in LLM response".to_string(),
))
.and_then(|content| {
serde_json::from_str(content).map_err(|e| {
AppError::LLMParsing(format!(
"Failed to parse LLM response into analysis: {}",
e
))
})
})
"No content found in LLM response".into(),
))?;
serde_json::from_str::<LLMEnrichmentResult>(content).map_err(|e| {
AppError::LLMParsing(format!("Failed to parse LLM response into analysis: {}", e))
})
}
}

2
crates/ingestion-pipeline/src/pipeline.rs
View File

@@ -53,7 +53,7 @@ impl IngestionPipeline {
)
.await?;
let text_content = to_text_content(task.content, &self.openai_client).await?;
let text_content = to_text_content(task.content, &self.openai_client, &self.db).await?;
match self.process(&text_content).await {
Ok(_) => {

21
crates/ingestion-pipeline/src/types/mod.rs
View File

@@ -10,16 +10,19 @@ use common::{
error::AppError,
storage::types::{
file_info::FileInfo, ingestion_payload::IngestionPayload, text_content::TextContent,
system_settings::SystemSettings,
},
};
use reqwest;
use scraper::{Html, Selector};
use std::fmt::Write;
use tiktoken_rs::{o200k_base, CoreBPE};
use common::storage::db::SurrealDbClient;
pub async fn to_text_content(
ingestion_payload: IngestionPayload,
openai_client: &Arc<async_openai::Client<async_openai::config::OpenAIConfig>>,
db_client: &Arc<SurrealDbClient>,
) -> Result<TextContent, AppError> {
match ingestion_payload {
IngestionPayload::Url {
@@ -28,7 +31,7 @@ pub async fn to_text_content(
category,
user_id,
} => {
let text = fetch_text_from_url(&url, openai_client).await?;
let text = fetch_text_from_url(&url, openai_client, db_client).await?;
Ok(TextContent::new(
text,
instructions.into(),
@@ -74,6 +77,7 @@ pub async fn to_text_content(
async fn fetch_text_from_url(
url: &str,
openai_client: &Arc<async_openai::Client<async_openai::config::OpenAIConfig>>,
db_client: &Arc<SurrealDbClient>,
) -> Result<String, AppError> {
// Use a client with timeouts and reuse
let client = reqwest::ClientBuilder::new()
@@ -119,12 +123,14 @@ async fn fetch_text_from_url(
let content = structured_content
.replace(|c: char| c.is_control(), " ")
.replace(" ", " ");
process_web_content(content, openai_client).await
process_web_content(content, openai_client, &db_client).await
}
pub async fn process_web_content(
content: String,
openai_client: &Arc<async_openai::Client<async_openai::config::OpenAIConfig>>,
db_client: &Arc<SurrealDbClient>,
) -> Result<String, AppError> {
const MAX_TOKENS: usize = 122000;
const SYSTEM_PROMPT: &str = r#"
@@ -155,6 +161,7 @@ pub async fn process_web_content(
"#;
let bpe = o200k_base()?;
let settings = SystemSettings::get_current(db_client).await?;
// Process content in chunks if needed
let truncated_content = if bpe.encode_with_special_tokens(&content).len() > MAX_TOKENS {
@@ -164,7 +171,7 @@ pub async fn process_web_content(
};
let request = CreateChatCompletionRequestArgs::default()
.model("gpt-4o-mini")
.model(&settings.processing_model)
.temperature(0.0)
.max_tokens(16200u32)
.messages([
@@ -174,13 +181,13 @@ pub async fn process_web_content(
.build()?;
let response = openai_client.chat().create(request).await?;
// Extract and return the content
response
.choices
.first()
.and_then(|choice| choice.message.content.as_ref())
.map(|content| content.to_owned())
.ok_or(AppError::LLMParsing("No content in response".into()))
.and_then(|choice| choice.message.content.clone())
.ok_or(AppError::LLMParsing("No content found in LLM response".into()))
}
fn truncate_content(

108
crates/ingestion-pipeline/src/utils/llm_instructions.rs
View File

@@ -1,81 +1,41 @@
use serde_json::{json, Value};
use common::storage::types::system_prompts::DEFAULT_INGRESS_ANALYSIS_SYSTEM_PROMPT;
use serde_json::json;
pub static INGRESS_ANALYSIS_SYSTEM_MESSAGE: &str = r#"
You are an AI assistant. You will receive a text content, along with user instructions 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 instructions 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:
{
"knowledge_entities": [
{
"key": "unique-key-1",
"name": "Entity Name",
"description": "A detailed description of the entity.",
"entity_type": "TypeOfEntity"
},
// More entities...
],
"relationships": [
{
"type": "RelationshipType",
"source": "unique-key-1 or UUID from existing database",
"target": "unique-key-1 or UUID from existing database"
},
// More relationships...
]
}
Guidelines:
1. Do NOT generate any IDs or UUIDs. Use a unique `key` for each knowledge entity.
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"
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 static INGRESS_ANALYSIS_SYSTEM_MESSAGE: &str = DEFAULT_INGRESS_ANALYSIS_SYSTEM_PROMPT;
pub fn get_ingress_analysis_schema() -> Value {
pub fn get_ingress_analysis_schema() -> serde_json::Value {
json!({
"type": "object",
"properties": {
"knowledge_entities": {
"type": "array",
"items": {
"type": "object",
"properties": {
"key": { "type": "string" },
"name": { "type": "string" },
"description": { "type": "string" },
"entity_type": {
"type": "string",
"enum": ["idea", "project", "document", "page", "textsnippet"]
}
"type": "object",
"properties": {
"knowledge_entities": {
"type": "array",
"items": {
"type": "object",
"properties": {
"key": { "type": "string" },
"name": { "type": "string" },
"description": { "type": "string" },
"entity_type": { "type": "string" }
},
"required": ["key", "name", "description", "entity_type"],
"additionalProperties": false
}
},
"required": ["key", "name", "description", "entity_type"],
"additionalProperties": false
}
"relationships": {
"type": "array",
"items": {
"type": "object",
"properties": {
"type": { "type": "string" },
"source": { "type": "string" },
"target": { "type": "string" }
},
"required": ["type", "source", "target"],
"additionalProperties": false
}
}
},
"relationships": {
"type": "array",
"items": {
"type": "object",
"properties": {
"type": {
"type": "string",
"enum": ["RelatedTo", "RelevantTo", "SimilarTo"]
},
"source": { "type": "string" },
"target": { "type": "string" }
},
"required": ["type", "source", "target"],
"additionalProperties": false
}
}
},
"required": ["knowledge_entities", "relationships"],
"additionalProperties": false
"required": ["knowledge_entities", "relationships"],
"additionalProperties": false
})
}

38
templates/admin/edit_ingestion_prompt_modal.html Normal file
View File

@@ -0,0 +1,38 @@
{% extends "modal_base.html" %}
{% block form_attributes %}
hx-patch="/update-ingestion-prompt"
hx-target="#system_prompt_section"
hx-swap="outerHTML"
{% endblock %}
{% block modal_content %}
<h3 class="text-lg font-bold mb-4">Edit Ingestion Prompt</h3>
<div class="form-control">
<textarea name="ingestion_system_prompt" class="textarea textarea-bordered h-96 w-full font-mono text-sm">{{
settings.ingestion_system_prompt }}</textarea>
<p class="text-xs text-gray-500 mt-1">System prompt used for content processing and ingestion</p>
</div>
{% endblock %}
{% block primary_actions %}
<button type="button" class="btn btn-outline mr-2" id="reset_prompt_button">
Reset to Default
</button>
<textarea id="default_prompt_content" style="display:none;">{{ default_ingestion_prompt }}</textarea>
<script>
document.getElementById('reset_prompt_button').addEventListener('click', function () {
const defaultContent = document.getElementById('default_prompt_content').value;
document.querySelector('textarea[name=ingestion_system_prompt]').value = defaultContent;
});
</script>
<button type="submit" class="btn btn-primary">
<span class="htmx-indicator hidden">
<span class="loading loading-spinner loading-xs mr-2"></span>
</span>
Save Changes
</button>
{% endblock %}

38
templates/admin/edit_query_prompt_modal.html Normal file
View File

@@ -0,0 +1,38 @@
{% extends "modal_base.html" %}
{% block form_attributes %}
hx-patch="/update-query-prompt"
hx-target="#system_prompt_section"
hx-swap="outerHTML"
{% endblock %}
{% block modal_content %}
<h3 class="text-lg font-bold mb-4">Edit System Prompt</h3>
<div class="form-control">
<textarea name="query_system_prompt" class="textarea textarea-bordered h-96 w-full font-mono text-sm">{{
settings.query_system_prompt }}</textarea>
<p class="text-xs text-gray-500 mt-1">System prompt used for answering user queries</p>
</div>
{% endblock %}
{% block primary_actions %}
<button type="button" class="btn btn-outline mr-2" id="reset_prompt_button">
Reset to Default
</button>
<textarea id="default_prompt_content" style="display:none;">{{ default_query_prompt }}</textarea>
<script>
document.getElementById('reset_prompt_button').addEventListener('click', function () {
const defaultContent = document.getElementById('default_prompt_content').value;
document.querySelector('textarea[name=query_system_prompt]').value = defaultContent;
});
</script>
<button type="submit" class="btn btn-primary">
<span class="htmx-indicator hidden">
<span class="loading loading-spinner loading-xs mr-2"></span>
</span>
Save Changes
</button>
{% endblock %}

56
templates/auth/admin_panel.html
View File

@@ -23,7 +23,6 @@
</div>
</div>
<!-- Settings in Fieldset -->
<div class="grid grid-cols-1 sm:grid-cols-2 gap-6">
<fieldset class="fieldset p-4 shadow rounded-box">
@@ -39,6 +38,61 @@
</label>
<div id="registration-status" class="text-sm mt-2"></div>
</fieldset>
<fieldset class="fieldset p-4 shadow rounded-box">
<legend class="fieldset-legend">AI Models</legend>
{% block model_settings_form %}
<form hx-patch="/update-model-settings" hx-swap="outerHTML">
<div class="form-control mb-4">
<label class="label">
<span class="label-text">Query Model</span>
</label>
<select name="query_model" class="select select-bordered w-full">
<option value="gpt-4o-mini" {% if settings.query_model=="gpt-4o-mini" %}selected{% endif %}>GPT-4o Mini
</option>
<option value="gpt-4o" {% if settings.query_model=="gpt-4o" %}selected{% endif %}>GPT-4o</option>
<option value="gpt-3.5-turbo" {% if settings.query_model=="gpt-3.5-turbo" %}selected{% endif %}>GPT-3.5
Turbo</option>
</select>
<p class="text-xs text-gray-500 mt-1">Model used for answering user queries</p>
</div>
<div class="form-control my-4">
<label class="label">
<span class="label-text">Processing Model</span>
</label>
<select name="processing_model" class="select select-bordered w-full">
<option value="gpt-4o-mini" {% if settings.processing_model=="gpt-4o-mini" %}selected{% endif %}>GPT-4o Mini
</option>
<option value="gpt-4o" {% if settings.processing_model=="gpt-4o" %}selected{% endif %}>GPT-4o</option>
<option value="gpt-3.5-turbo" {% if settings.processing_model=="gpt-3.5-turbo" %}selected{% endif %}>GPT-3.5
Turbo</option>
</select>
<p class="text-xs text-gray-500 mt-1">Model used for content processing and ingestion</p>
</div>
<button type="submit" class="btn btn-primary btn-sm">Save Model Settings</button>
</form>
{% endblock %}
</fieldset>
{% block system_prompt_section %}
<div id="system_prompt_section">
<fieldset class="fieldset p-4 shadow rounded-box">
<legend class="fieldset-legend">System Prompts</legend>
<div class="flex gap-2 flex-col sm:flex-row">
<button type="button" class="btn btn-primary btn-sm" hx-get="/edit-query-prompt" hx-target="#modal"
hx-swap="innerHTML">
Edit Query Prompt
</button>
<button type="button" class="btn btn-primary btn-sm" hx-get="/edit-ingestion-prompt" hx-target="#modal"
hx-swap="innerHTML">
Edit Ingestion Prompt
</button>
</div>
</fieldset>
</div>
{% endblock %}
</div>
</main>
{% endblock %}

3
templates/documentation/get_started.html
View File

@@ -16,9 +16,6 @@
<pre class="overflow-x-auto text-sm">
<code class="break-words whitespace-pre-wrap">
OPENAI_API_KEY: your_api_key
SMTP_EMAIL_RELAYER: your_email_relayer
SMTP_USERNAME: your_smtp_username
SMTP_PASSWORD: your_smtp_password
DB_ADDRESS: your_db_address
DB_USER: your_db_user
DB_PASSWORD: your_db_password

5
todo.md
View File

@@ -1,8 +1,13 @@
\[\] archive ingressed webpage
\[\] openai api key in config
\[\] option to set models, query and processing
\[\] template customization?
\[\] configs primarily get envs
\[\] filtering on categories
\[\] three js graph explorer
\[\] three js vector explorer
\[\] integrate templates in release build
\[\] integrate assets folder in release build
\[x\] add user_id to ingress objects
\[x\] admin controls re registration
\[x\] chat functionality