tidying stuff up, dto for search

Cargo.lock

@@ -2194,7 +2194,7 @@ dependencies = [
 ]
 
 [[package]]
-name = "eval"
+name = "evaluations"
 version = "0.1.0"
 dependencies = [
  "anyhow",

Cargo.toml

@@ -7,7 +7,7 @@ members = [
   "ingestion-pipeline",
   "retrieval-pipeline",
   "json-stream-parser",
-  "eval"
+  "evaluations"
 ]
 resolver = "2"
 
@@ -80,7 +80,7 @@ implicit_clone = "warn"
 redundant_clone = "warn"
 
 # Security-focused lints
-integer_arithmetic = "warn"
+arithmetic_side_effects = "warn"
 indexing_slicing = "warn"
 unwrap_used = "warn"
 expect_used = "warn"
@@ -90,7 +90,7 @@ todo = "warn"
 
 # Async/Network lints
 async_yields_async = "warn"
-await_holding_invalid_state = "warn"
+await_holding_invalid_type = "warn"
 rc_buffer = "warn"
 
 # Maintainability-focused lints
@@ -109,6 +109,8 @@ wildcard_dependencies = "warn"
 missing_docs_in_private_items = "warn"
 
 # Allow noisy lints that don't add value for this project
-manual_must_use = "allow"
 needless_raw_string_hashes = "allow"
 multiple_bound_locations = "allow"
+cargo_common_metadata = "allow"
+multiple-crate-versions = "allow"
+module_name_repetition = "allow"

common/migrations/20251210_add_embedding_backend_to_system_settings.surql

@@ -0,0 +1,8 @@
+-- Add embedding_backend field to system_settings for visibility of active backend
+
+DEFINE FIELD IF NOT EXISTS embedding_backend ON system_settings TYPE option<string>;
+
+-- Set default to 'openai' for existing installs to preserve backward compatibility
+UPDATE system_settings:current SET
+    embedding_backend = 'openai'
+WHERE embedding_backend == NONE;

common/src/storage/types/knowledge_entity.rs

@@ -428,6 +428,103 @@ impl KnowledgeEntity {
         info!("Re-embedding process for knowledge entities completed successfully.");
         Ok(())
     }
+
+    /// Re-creates embeddings for all knowledge entities using an `EmbeddingProvider`.
+    ///
+    /// This variant uses the application's configured embedding provider (FastEmbed, OpenAI, etc.)
+    /// instead of directly calling OpenAI. Used during startup when embedding configuration changes.
+    pub async fn update_all_embeddings_with_provider(
+        db: &SurrealDbClient,
+        provider: &crate::utils::embedding::EmbeddingProvider,
+    ) -> Result<(), AppError> {
+        let new_dimensions = provider.dimension();
+        info!(
+            dimensions = new_dimensions,
+            backend = provider.backend_label(),
+            "Starting re-embedding process for all knowledge entities"
+        );
+
+        // Fetch all entities first
+        let all_entities: Vec<KnowledgeEntity> = db.select(Self::table_name()).await?;
+        let total_entities = all_entities.len();
+        if total_entities == 0 {
+            info!("No knowledge entities to update. Just updating the index.");
+            KnowledgeEntityEmbedding::redefine_hnsw_index(db, new_dimensions).await?;
+            return Ok(());
+        }
+        info!(entities = total_entities, "Found entities to process");
+
+        // Generate all new embeddings in memory
+        let mut new_embeddings: HashMap<String, (Vec<f32>, String)> = HashMap::new();
+        info!("Generating new embeddings for all entities...");
+
+        for (i, entity) in all_entities.iter().enumerate() {
+            if i > 0 && i % 100 == 0 {
+                info!(progress = i, total = total_entities, "Re-embedding progress");
+            }
+
+            let embedding_input = format!(
+                "name: {}, description: {}, type: {:?}",
+                entity.name, entity.description, entity.entity_type
+            );
+
+            let embedding = provider
+                .embed(&embedding_input)
+                .await
+                .map_err(|e| AppError::InternalError(format!("Embedding failed: {e}")))?;
+
+            // Safety check: ensure the generated embedding has the correct dimension.
+            if embedding.len() != new_dimensions {
+                let err_msg = format!(
+                    "CRITICAL: Generated embedding for entity {} has incorrect dimension ({}). Expected {}. Aborting.",
+                    entity.id, embedding.len(), new_dimensions
+                );
+                error!("{}", err_msg);
+                return Err(AppError::InternalError(err_msg));
+            }
+            new_embeddings.insert(entity.id.clone(), (embedding, entity.user_id.clone()));
+        }
+        info!("Successfully generated all new embeddings.");
+
+        // Perform DB updates in a single transaction
+        info!("Applying embedding updates in a transaction...");
+        let mut transaction_query = String::from("BEGIN TRANSACTION;");
+
+        for (id, (embedding, user_id)) in new_embeddings {
+            let embedding_str = format!(
+                "[{}]",
+                embedding
+                    .iter()
+                    .map(|f| f.to_string())
+                    .collect::<Vec<_>>()
+                    .join(",")
+            );
+            transaction_query.push_str(&format!(
+                "UPSERT type::thing('knowledge_entity_embedding', '{id}') SET \
+                    entity_id = type::thing('knowledge_entity', '{id}'), \
+                    embedding = {embedding}, \
+                    user_id = '{user_id}', \
+                    created_at = IF created_at != NONE THEN created_at ELSE time::now() END, \
+                    updated_at = time::now();",
+                id = id,
+                embedding = embedding_str,
+                user_id = user_id
+            ));
+        }
+
+        transaction_query.push_str(&format!(
+            "DEFINE INDEX OVERWRITE idx_embedding_knowledge_entity_embedding ON TABLE knowledge_entity_embedding FIELDS embedding HNSW DIMENSION {};",
+            new_dimensions
+        ));
+
+        transaction_query.push_str("COMMIT TRANSACTION;");
+
+        // Execute the entire atomic operation
+        db.query(transaction_query).await?;
+
+        info!("Re-embedding process for knowledge entities completed successfully.");
+        Ok(())
+    }
 }
 
 #[cfg(test)]

common/src/storage/types/system_settings.rs

@@ -13,6 +13,9 @@ pub struct SystemSettings {
     pub processing_model: String,
     pub embedding_model: String,
     pub embedding_dimensions: u32,
+    /// Active embedding backend ("openai", "fastembed", "hashed"). Read-only, synced from config.
+    #[serde(default)]
+    pub embedding_backend: Option<String>,
     pub query_system_prompt: String,
     pub ingestion_system_prompt: String,
     pub image_processing_model: String,
@@ -49,6 +52,57 @@ impl SystemSettings {
             "Something went wrong updating the settings".into(),
         ))
     }
+
+    /// Syncs SystemSettings with the active embedding provider's properties.
+    /// Updates embedding_backend, embedding_model, and embedding_dimensions if they differ.
+    /// Returns true if any settings were changed.
+    pub async fn sync_from_embedding_provider(
+        db: &SurrealDbClient,
+        provider: &crate::utils::embedding::EmbeddingProvider,
+    ) -> Result<(Self, bool), AppError> {
+        let mut settings = Self::get_current(db).await?;
+        let mut needs_update = false;
+
+        let backend_label = provider.backend_label().to_string();
+        let provider_dimensions = provider.dimension() as u32;
+        let provider_model = provider.model_code();
+
+        // 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,
+                new_dimensions = provider_dimensions,
+                "Embedding dimensions changed, updating SystemSettings"
+            );
+            settings.embedding_dimensions = provider_dimensions;
+            needs_update = true;
+        }
+
+        // Sync model if provider has one
+        if let Some(model) = provider_model {
+            if settings.embedding_model != model {
+                tracing::info!(
+                    old_model = %settings.embedding_model,
+                    new_model = %model,
+                    "Embedding model changed, updating SystemSettings"
+                );
+                settings.embedding_model = model;
+                needs_update = true;
+            }
+        }
+
+        if needs_update {
+            settings = Self::update(db, settings).await?;
+        }
+
+        Ok((settings, needs_update))
+    }
 }
 
 #[cfg(test)]

common/src/storage/types/text_chunk.rs

@@ -323,6 +323,106 @@ impl TextChunk {
         info!("Re-embedding process for text chunks completed successfully.");
         Ok(())
     }
+
+    /// Re-creates embeddings for all text chunks using an `EmbeddingProvider`.
+    ///
+    /// This variant uses the application's configured embedding provider (FastEmbed, OpenAI, etc.)
+    /// instead of directly calling OpenAI. Used during startup when embedding configuration changes.
+    pub async fn update_all_embeddings_with_provider(
+        db: &SurrealDbClient,
+        provider: &crate::utils::embedding::EmbeddingProvider,
+    ) -> Result<(), AppError> {
+        let new_dimensions = provider.dimension();
+        info!(
+            dimensions = new_dimensions,
+            backend = provider.backend_label(),
+            "Starting re-embedding process for all text chunks"
+        );
+
+        // Fetch all chunks first
+        let all_chunks: Vec<TextChunk> = db.select(Self::table_name()).await?;
+        let total_chunks = all_chunks.len();
+        if total_chunks == 0 {
+            info!("No text chunks to update. Just updating the index.");
+            TextChunkEmbedding::redefine_hnsw_index(db, new_dimensions).await?;
+            return Ok(());
+        }
+        info!(chunks = total_chunks, "Found chunks to process");
+
+        // Generate all new embeddings in memory
+        let mut new_embeddings: HashMap<String, (Vec<f32>, String, String)> = HashMap::new();
+        info!("Generating new embeddings for all chunks...");
+        
+        for (i, chunk) in all_chunks.iter().enumerate() {
+            if i > 0 && i % 100 == 0 {
+                info!(progress = i, total = total_chunks, "Re-embedding progress");
+            }
+            
+            let embedding = provider
+                .embed(&chunk.chunk)
+                .await
+                .map_err(|e| AppError::InternalError(format!("Embedding failed: {e}")))?;
+
+            // Safety check: ensure the generated embedding has the correct dimension.
+            if embedding.len() != new_dimensions {
+                let err_msg = format!(
+                    "CRITICAL: Generated embedding for chunk {} has incorrect dimension ({}). Expected {}. Aborting.",
+                    chunk.id, embedding.len(), new_dimensions
+                );
+                error!("{}", err_msg);
+                return Err(AppError::InternalError(err_msg));
+            }
+            new_embeddings.insert(
+                chunk.id.clone(),
+                (embedding, chunk.user_id.clone(), chunk.source_id.clone()),
+            );
+        }
+        info!("Successfully generated all new embeddings.");
+
+        // Perform DB updates in a single transaction against the embedding table
+        info!("Applying embedding updates in a transaction...");
+        let mut transaction_query = String::from("BEGIN TRANSACTION;");
+
+        for (id, (embedding, user_id, source_id)) in new_embeddings {
+            let embedding_str = format!(
+                "[{}]",
+                embedding
+                    .iter()
+                    .map(ToString::to_string)
+                    .collect::<Vec<_>>()
+                    .join(",")
+            );
+            write!(
+                &mut transaction_query,
+                "UPSERT type::thing('text_chunk_embedding', '{id}') SET \
+                    chunk_id = type::thing('text_chunk', '{id}'), \
+                    source_id = '{source_id}', \
+                    embedding = {embedding}, \
+                    user_id = '{user_id}', \
+                    created_at = IF created_at != NONE THEN created_at ELSE time::now() END, \
+                    updated_at = time::now();",
+                id = id,
+                embedding = embedding_str,
+                user_id = user_id,
+                source_id = source_id
+            )
+            .map_err(|e| AppError::InternalError(e.to_string()))?;
+        }
+
+        write!(
+            &mut transaction_query,
+            "DEFINE INDEX OVERWRITE idx_embedding_text_chunk_embedding ON TABLE text_chunk_embedding FIELDS embedding HNSW DIMENSION {};",
+            new_dimensions
+        )
+        .map_err(|e| AppError::InternalError(e.to_string()))?;
+
+        transaction_query.push_str("COMMIT TRANSACTION;");
+
+        db.query(transaction_query).await?;
+
+        info!("Re-embedding process for text chunks completed successfully.");
+        Ok(())
+    }
 }
 
 #[cfg(test)]

common/src/utils/config.rs

@@ -2,6 +2,19 @@ use config::{Config, ConfigError, Environment, File};
 use serde::Deserialize;
 use std::env;
 
+/// Selects the embedding backend for vector generation.
+#[derive(Clone, Deserialize, Debug, Default, PartialEq)]
+#[serde(rename_all = "lowercase")]
+pub enum EmbeddingBackend {
+    /// Use OpenAI-compatible API for embeddings.
+    OpenAI,
+    /// Use FastEmbed local embeddings (default).
+    #[default]
+    FastEmbed,
+    /// Use deterministic hashed embeddings (for testing).
+    Hashed,
+}
+
 #[derive(Clone, Deserialize, Debug, PartialEq)]
 #[serde(rename_all = "lowercase")]
 pub enum StorageKind {
@@ -60,6 +73,8 @@ pub struct AppConfig {
     pub fastembed_max_length: Option<usize>,
     #[serde(default)]
     pub retrieval_strategy: Option<String>,
+    #[serde(default)]
+    pub embedding_backend: EmbeddingBackend,
 }
 
 /// Default data directory for persisted assets.
@@ -127,6 +142,7 @@ impl Default for AppConfig {
             fastembed_show_download_progress: None,
             fastembed_max_length: None,
             retrieval_strategy: None,
+            embedding_backend: EmbeddingBackend::default(),
         }
     }
 }

common/src/utils/embedding.rs

@@ -235,6 +235,34 @@ impl EmbeddingProvider {
             },
         })
     }
+
+    /// Creates an embedding provider based on application configuration.
+    ///
+    /// 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> {
+        use crate::utils::config::EmbeddingBackend;
+
+        match config.embedding_backend {
+            EmbeddingBackend::OpenAI => {
+                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 => {
+                // 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),
+        }
+    }
 }
 
 // Helper functions for hashed embeddings

html-router/src/html_state.rs

@@ -1,4 +1,5 @@
 use common::storage::{db::SurrealDbClient, store::StorageManager};
+use common::utils::embedding::EmbeddingProvider;
 use common::utils::template_engine::{ProvidesTemplateEngine, TemplateEngine};
 use common::{create_template_engine, storage::db::ProvidesDb, utils::config::AppConfig};
 use retrieval_pipeline::{reranking::RerankerPool, RetrievalStrategy};
@@ -16,6 +17,7 @@ pub struct HtmlState {
     pub config: AppConfig,
     pub storage: StorageManager,
     pub reranker_pool: Option<Arc<RerankerPool>>,
+    pub embedding_provider: Arc<EmbeddingProvider>,
 }
 
 impl HtmlState {
@@ -26,6 +28,7 @@ impl HtmlState {
         storage: StorageManager,
         config: AppConfig,
         reranker_pool: Option<Arc<RerankerPool>>,
+        embedding_provider: Arc<EmbeddingProvider>,
     ) -> Result<Self, Box<dyn std::error::Error>> {
         let template_engine = create_template_engine!("templates");
         debug!("Template engine created for html_router.");
@@ -38,6 +41,7 @@ impl HtmlState {
             config,
             storage,
             reranker_pool,
+            embedding_provider,
         })
     }
 

html-router/src/routes/admin/handlers.rs

@@ -151,21 +151,46 @@ pub async fn update_model_settings(
 
     let current_settings = SystemSettings::get_current(&state.db).await?;
 
-    // Determine if re-embedding is required
-    let reembedding_needed = input
-        .embedding_dimensions
-        .is_some_and(|new_dims| new_dims != current_settings.embedding_dimensions);
+    // Check if using FastEmbed - if so, embedding model/dimensions cannot be changed via UI
+    let uses_local_embeddings = current_settings
+        .embedding_backend
+        .as_deref()
+        .is_some_and(|b| b == "fastembed" || b == "hashed");
+
+    // For local embeddings, ignore any embedding model/dimension changes from the form
+    let (final_embedding_model, final_embedding_dimensions, reembedding_needed) =
+        if uses_local_embeddings {
+            // Keep current values - they're controlled by config, not the admin UI
+            info!(
+                backend = ?current_settings.embedding_backend,
+                "Embedding model/dimensions controlled by config, ignoring form input"
+            );
+            (
+                current_settings.embedding_model.clone(),
+                current_settings.embedding_dimensions,
+                false,
+            )
+        } else {
+            // OpenAI backend - allow changes from form
+            let reembedding_needed = input
+                .embedding_dimensions
+                .is_some_and(|new_dims| new_dims != current_settings.embedding_dimensions);
+            (
+                input.embedding_model,
+                input
+                    .embedding_dimensions
+                    .unwrap_or(current_settings.embedding_dimensions),
+                reembedding_needed,
+            )
+        };
 
     let new_settings = SystemSettings {
         query_model: input.query_model,
         processing_model: input.processing_model,
         image_processing_model: input.image_processing_model,
         voice_processing_model: input.voice_processing_model,
-        embedding_model: input.embedding_model,
-        // Use new dimensions if provided, otherwise retain the current ones.
-        embedding_dimensions: input
-            .embedding_dimensions
-            .unwrap_or(current_settings.embedding_dimensions),
+        embedding_model: final_embedding_model,
+        embedding_dimensions: final_embedding_dimensions,
         ..current_settings.clone()
     };
 

html-router/src/routes/chat/message_response_stream.rs

@@ -132,6 +132,7 @@ pub async fn get_response_stream(
     let retrieval_result = match retrieval_pipeline::retrieve_entities(
         &state.db,
         &state.openai_client,
+        Some(&*state.embedding_provider),
         &user_message.content,
         &user.id,
         config,

html-router/src/routes/knowledge/handlers.rs

@@ -288,6 +288,7 @@ pub async fn suggest_knowledge_relationships(
             retrieval_pipeline::retrieve_entities(
                 &state.db,
                 &state.openai_client,
+                Some(&*state.embedding_provider),
                 &query,
                 &user.id,
                 config,

html-router/src/routes/search/handlers.rs

@@ -1,13 +1,21 @@
-use std::{fmt, str::FromStr, time::Duration};
+use std::{
+    collections::{HashMap, HashSet},
+    fmt, str::FromStr,
+};
 
 use axum::{
     extract::{Query, State},
     response::IntoResponse,
 };
-use common::storage::types::{conversation::Conversation, user::User};
+use common::storage::types::{
+    conversation::Conversation,
+    text_content::{deserialize_flexible_id, TextContent},
+    user::User,
+    StoredObject,
+};
 use retrieval_pipeline::{RetrievalConfig, SearchResult, SearchTarget, StrategyOutput};
 use serde::{de, Deserialize, Deserializer, Serialize};
-use tokio::time::error::Elapsed;
+use surrealdb::RecordId;
 
 use crate::{
     html_state::HtmlState,
@@ -31,6 +39,113 @@ where
     }
 }
 
+fn source_id_suffix(source_id: &str) -> String {
+    let start = source_id.len().saturating_sub(8);
+    source_id[start..].to_string()
+}
+
+fn truncate_label(value: &str, max_chars: usize) -> String {
+    let mut end = None;
+    let mut count = 0;
+    for (idx, _) in value.char_indices() {
+        if count == max_chars {
+            end = Some(idx);
+            break;
+        }
+        count += 1;
+    }
+
+    match end {
+        Some(idx) => format!("{}...", &value[..idx]),
+        None => value.to_string(),
+    }
+}
+
+fn first_non_empty_line(text: &str, max_chars: usize) -> Option<String> {
+    for line in text.lines() {
+        let trimmed = line.trim();
+        if !trimmed.is_empty() {
+            return Some(truncate_label(trimmed, max_chars));
+        }
+    }
+    None
+}
+
+#[derive(Deserialize)]
+struct UrlInfoLabel {
+    #[serde(default)]
+    title: String,
+    #[serde(default)]
+    url: String,
+}
+
+#[derive(Deserialize)]
+struct FileInfoLabel {
+    #[serde(default)]
+    file_name: String,
+}
+
+#[derive(Deserialize)]
+struct SourceLabelRow {
+    #[serde(deserialize_with = "deserialize_flexible_id")]
+    id: String,
+    #[serde(default)]
+    url_info: Option<UrlInfoLabel>,
+    #[serde(default)]
+    file_info: Option<FileInfoLabel>,
+    #[serde(default)]
+    context: Option<String>,
+    #[serde(default)]
+    category: String,
+    #[serde(default)]
+    text: String,
+}
+
+fn build_source_label(row: &SourceLabelRow) -> String {
+    const MAX_LABEL_CHARS: usize = 80;
+
+    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_label(trimmed, MAX_LABEL_CHARS);
+        }
+    }
+
+    if let Some(text_label) = first_non_empty_line(&row.text, MAX_LABEL_CHARS) {
+        return text_label;
+    }
+
+    let category = row.category.trim();
+    if !category.is_empty() {
+        return truncate_label(category, MAX_LABEL_CHARS);
+    }
+
+    format!("Text snippet: {}", source_id_suffix(&row.id))
+}
+
+fn fallback_source_label(source_id: &str) -> String {
+    format!("Text snippet: {}", source_id_suffix(source_id))
+}
+
 #[derive(Deserialize)]
 pub struct SearchParams {
     #[serde(default, deserialize_with = "empty_string_as_none")]
@@ -42,6 +157,7 @@ pub struct SearchParams {
 struct TextChunkForTemplate {
     id: String,
     source_id: String,
+    source_label: String,
     chunk: String,
     score: f32,
 }
@@ -54,6 +170,7 @@ struct KnowledgeEntityForTemplate {
     description: String,
     entity_type: String,
     source_id: String,
+    source_label: String,
     score: f32,
 }
 
@@ -89,14 +206,21 @@ pub async fn search_result_handler(
             } else {
                 // Use retrieval pipeline Search strategy
                 let config = RetrievalConfig::for_search(SearchTarget::Both);
+
+                // Checkout a reranker lease if pool is available
+                let reranker_lease = match &state.reranker_pool {
+                    Some(pool) => Some(pool.checkout().await),
+                    None => None,
+                };
+                
                 let result = retrieval_pipeline::pipeline::run_pipeline(
                     &state.db,
                     &state.openai_client,
-                    None, // No embedding provider in HtmlState
+                    Some(&state.embedding_provider),
                     trimmed_query,
                     &user.id,
                     config,
-                    None, // No reranker for now
+                    reranker_lease,
                 )
                 .await?;
 
@@ -105,17 +229,74 @@ pub async fn search_result_handler(
                     _ => SearchResult::new(vec![], vec![]),
                 };
 
+                let mut source_ids = HashSet::new();
+                for chunk_result in &search_result.chunks {
+                    source_ids.insert(chunk_result.chunk.source_id.clone());
+                }
+                for entity_result in &search_result.entities {
+                    source_ids.insert(entity_result.entity.source_id.clone());
+                }
+
+                let source_label_map = if source_ids.is_empty() {
+                    HashMap::new()
+                } else {
+                    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(TextContent::table_name(), id))
+                            }
+                        })
+                        .collect();
+                    let mut response = state
+                        .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", TextContent::table_name()))
+                        .bind(("user_id", user.id.clone()))
+                        .bind(("record_ids", record_ids))
+                        .await?;
+                    let contents: Vec<SourceLabelRow> = response.take(0)?;
+
+                    tracing::debug!(
+                        source_id_count = source_ids.len(),
+                        label_row_count = contents.len(),
+                        "Resolved search 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!("{}:{}", TextContent::table_name(), content.id),
+                            label,
+                        );
+                    }
+
+                    labels
+                };
+
                 let mut combined_results: Vec<SearchResultForTemplate> =
                     Vec::with_capacity(search_result.chunks.len() + search_result.entities.len());
 
                 // Add chunk results
                 for chunk_result in search_result.chunks {
+                    let source_label = source_label_map
+                        .get(&chunk_result.chunk.source_id)
+                        .cloned()
+                        .unwrap_or_else(|| fallback_source_label(&chunk_result.chunk.source_id));
                     combined_results.push(SearchResultForTemplate {
                         result_type: "text_chunk".to_string(),
                         score: chunk_result.score,
                         text_chunk: Some(TextChunkForTemplate {
                             id: chunk_result.chunk.id,
                             source_id: chunk_result.chunk.source_id,
+                            source_label,
                             chunk: chunk_result.chunk.chunk,
                             score: chunk_result.score,
                         }),
@@ -125,6 +306,10 @@ pub async fn search_result_handler(
 
                 // Add entity results
                 for entity_result in search_result.entities {
+                    let source_label = source_label_map
+                        .get(&entity_result.entity.source_id)
+                        .cloned()
+                        .unwrap_or_else(|| fallback_source_label(&entity_result.entity.source_id));
                     combined_results.push(SearchResultForTemplate {
                         result_type: "knowledge_entity".to_string(),
                         score: entity_result.score,
@@ -135,6 +320,7 @@ pub async fn search_result_handler(
                             description: entity_result.entity.description,
                             entity_type: format!("{:?}", entity_result.entity.entity_type),
                             source_id: entity_result.entity.source_id,
+                            source_label,
                             score: entity_result.score,
                         }),
                     });

html-router/templates/admin/base.html

@@ -36,9 +36,12 @@
       <div id="system_prompt_section" class="nb-panel p-4">
         <div class="text-sm font-semibold mb-3">System Prompts</div>
         <div class="flex gap-2 flex-col sm:flex-row">
-          <button type="button" class="nb-btn btn-sm" hx-get="/edit-query-prompt" hx-target="#modal" hx-swap="innerHTML">Edit Query Prompt</button>
-          <button type="button" class="nb-btn btn-sm" hx-get="/edit-ingestion-prompt" hx-target="#modal" hx-swap="innerHTML">Edit Ingestion Prompt</button>
-          <button type="button" class="nb-btn btn-sm" hx-get="/edit-image-prompt" hx-target="#modal" hx-swap="innerHTML">Edit Image Prompt</button>
+          <button type="button" class="nb-btn btn-sm" hx-get="/edit-query-prompt" hx-target="#modal"
+            hx-swap="innerHTML">Edit Query Prompt</button>
+          <button type="button" class="nb-btn btn-sm" hx-get="/edit-ingestion-prompt" hx-target="#modal"
+            hx-swap="innerHTML">Edit Ingestion Prompt</button>
+          <button type="button" class="nb-btn btn-sm" hx-get="/edit-image-prompt" hx-target="#modal"
+            hx-swap="innerHTML">Edit Image Prompt</button>
         </div>
       </div>
       {% endblock %}
@@ -52,7 +55,8 @@
             <div class="text-sm opacity-80 mb-1">Query Model</div>
             <select name="query_model" class="nb-select w-full">
               {% for model in available_models.data %}
-              <option value="{{model.id}}" {% if settings.query_model==model.id %} selected {% endif %}>{{model.id}}</option>
+              <option value="{{model.id}}" {% if settings.query_model==model.id %} selected {% endif %}>{{model.id}}
+              </option>
               {% endfor %}
             </select>
             <p class="text-xs opacity-70 mt-1">Current: <span class="font-mono">{{settings.query_model}}</span></p>
@@ -63,7 +67,8 @@
             <div class="text-sm opacity-80 mb-1">Processing Model</div>
             <select name="processing_model" class="nb-select w-full">
               {% for model in available_models.data %}
-              <option value="{{model.id}}" {% if settings.processing_model==model.id %} selected {% endif %}>{{model.id}}</option>
+              <option value="{{model.id}}" {% if settings.processing_model==model.id %} selected {% endif %}>
+                {{model.id}}</option>
               {% endfor %}
             </select>
             <p class="text-xs opacity-70 mt-1">Current: <span class="font-mono">{{settings.processing_model}}</span></p>
@@ -74,10 +79,12 @@
             <div class="text-sm opacity-80 mb-1">Image Processing Model</div>
             <select name="image_processing_model" class="nb-select w-full">
               {% for model in available_models.data %}
-              <option value="{{model.id}}" {% if settings.image_processing_model==model.id %} selected {% endif %}>{{model.id}}</option>
+              <option value="{{model.id}}" {% if settings.image_processing_model==model.id %} selected {% endif %}>
+                {{model.id}}</option>
               {% endfor %}
             </select>
-            <p class="text-xs opacity-70 mt-1">Current: <span class="font-mono">{{settings.image_processing_model}}</span></p>
+            <p class="text-xs opacity-70 mt-1">Current: <span
+                class="font-mono">{{settings.image_processing_model}}</span></p>
           </div>
 
           <!-- Voice Processing Model -->
@@ -85,39 +92,66 @@
             <div class="text-sm opacity-80 mb-1">Voice Processing Model</div>
             <select name="voice_processing_model" class="nb-select w-full">
               {% for model in available_models.data %}
-              <option value="{{model.id}}" {% if settings.voice_processing_model==model.id %} selected {% endif %}>{{model.id}}</option>
+              <option value="{{model.id}}" {% if settings.voice_processing_model==model.id %} selected {% endif %}>
+                {{model.id}}</option>
               {% endfor %}
             </select>
-            <p class="text-xs opacity-70 mt-1">Current: <span class="font-mono">{{settings.voice_processing_model}}</span></p>
+            <p class="text-xs opacity-70 mt-1">Current: <span
+                class="font-mono">{{settings.voice_processing_model}}</span></p>
           </div>
 
           <!-- Embedding Model -->
           <div>
             <div class="text-sm opacity-80 mb-1">Embedding Model</div>
+            {% if settings.embedding_backend == "fastembed" or settings.embedding_backend == "hashed" %}
+            <input type="text" name="embedding_model" class="nb-input w-full opacity-60 cursor-not-allowed"
+              value="{{settings.embedding_model}}" disabled />
+            <p class="text-xs opacity-70 mt-1">Model: <span class="font-mono">{{settings.embedding_model}}
+                ({{settings.embedding_dimensions}} dims)</span></p>
+            <p class="text-xs text-info mt-1">ℹ️ Embedding model is controlled by config when using <span
+                class="font-mono">{{settings.embedding_backend}}</span> backend.</p>
+            {% else %}
             <select name="embedding_model" class="nb-select w-full">
               {% for model in available_models.data %}
-              <option value="{{model.id}}" {% if settings.embedding_model==model.id %} selected {% endif %}>{{model.id}}</option>
+              <option value="{{model.id}}" {% if settings.embedding_model==model.id %} selected {% endif %}>{{model.id}}
+              </option>
               {% endfor %}
             </select>
-            <p class="text-xs opacity-70 mt-1">Current: <span class="font-mono">{{settings.embedding_model}} ({{settings.embedding_dimensions}} dims)</span></p>
+            <p class="text-xs opacity-70 mt-1">Current: <span class="font-mono">{{settings.embedding_model}}
+                ({{settings.embedding_dimensions}} dims)</span></p>
+            {% endif %}
           </div>
 
           <!-- Embedding Dimensions -->
           <div>
             <div class="text-sm opacity-80 mb-1" for="embedding_dimensions">Embedding Dimensions</div>
-            <input type="number" id="embedding_dimensions" name="embedding_dimensions" class="nb-input w-full" value="{{ settings.embedding_dimensions }}" required />
+            {% if settings.embedding_backend == "fastembed" or settings.embedding_backend == "hashed" %}
+            <input type="number" id="embedding_dimensions" name="embedding_dimensions"
+              class="nb-input w-full opacity-60 cursor-not-allowed" value="{{ settings.embedding_dimensions }}"
+              disabled />
+            <p class="text-xs text-info mt-1">ℹ️ Dimensions are fixed for <span
+                class="font-mono">{{settings.embedding_backend}}</span> backend. Set <span
+                class="font-mono">EMBEDDING_BACKEND=openai</span> to use OpenAI embeddings.</p>
+            {% else %}
+            <input type="number" id="embedding_dimensions" name="embedding_dimensions" class="nb-input w-full"
+              value="{{ settings.embedding_dimensions }}" required />
+            {% endif %}
           </div>
 
           <!-- Alert -->
+          {% if settings.embedding_backend != "fastembed" and settings.embedding_backend != "hashed" %}
           <div id="embedding-change-alert" class="nb-panel p-3 bg-warning/20 hidden">
-            <div class="text-sm"><strong>Warning:</strong> Changing dimensions will require re-creating all embeddings. Look up your model's required dimensions or use a model that allows specifying them.</div>
+            <div class="text-sm"><strong>Warning:</strong> Changing dimensions will require re-creating all embeddings.
+              Look up your model's required dimensions or use a model that allows specifying them.</div>
           </div>
+          {% endif %}
 
           <div class="flex justify-end">
             <button type="submit" class="nb-btn nb-cta btn-sm">Save Model Settings</button>
           </div>
         </form>
 
+        {% if settings.embedding_backend != "fastembed" and settings.embedding_backend != "hashed" %}
         <script>
           // Rebind after HTMX swaps
           (() => {
@@ -135,6 +169,7 @@
             }
           })();
         </script>
+        {% endif %}
         {% endblock %}
       </div>
 
@@ -143,7 +178,8 @@
         <label class="flex items-center gap-3">
           {% block registration_status_input %}
           <form hx-patch="/toggle-registrations" hx-swap="outerHTML" hx-trigger="change">
-          <input name="registration_open" type="checkbox" class="nb-checkbox" {% if settings.registrations_enabled %}checked{% endif %} />
+            <input name="registration_open" type="checkbox" class="nb-checkbox" {% if settings.registrations_enabled
+              %}checked{% endif %} />
           </form>
           {% endblock %}
           <span class="text-sm">Enable Registrations</span>
@@ -153,4 +189,4 @@
     </section>
   </div>
 </div>
-{% endblock %}
+{% endblock %}

html-router/templates/search/response.html

@@ -22,7 +22,7 @@
     </div>
 
     <div class="flex-1 min-w-0">
-      <h3 class="text-lg font-extrabold mb-1 leading-snug">
+      <h3 class="text-lg font-extrabold mb-2 leading-snug">
         <a hx-get="/content/{{ tc.id }}/read" hx-target="#modal" hx-swap="innerHTML" class="nb-link">
           {% set title_text = tc.highlighted_url_title
           | default(tc.url_info.title if tc.url_info else none, true)
@@ -72,6 +72,44 @@
         </span>
       </div>
     </div>
+    {% elif result.result_type == "text_chunk" %}
+    {% set chunk = result.text_chunk %}
+    <div
+      class="w-10 h-10 flex-shrink-0 self-start mt-1 grid place-items-center border-2 border-neutral bg-base-100 shadow-[4px_4px_0_0_#000]">
+      <div class="tooltip tooltip-right" data-tip="Text Chunk">
+        {% include "icons/bars_icon.html" %}
+      </div>
+    </div>
+
+    <div class="flex-1 min-w-0">
+      <h3 class="text-lg font-extrabold mb-1 leading-snug">
+        <a hx-get="/content/{{ chunk.source_id }}/read" hx-target="#modal" hx-swap="innerHTML" class="nb-link">
+          Chunk from {{ chunk.source_label }}
+        </a>
+      </h3>
+
+      <div class="markdown-content prose-tufte-compact text-base-content/80 mb-4 overflow-hidden line-clamp-6"
+        data-content="{{ chunk.chunk | escape }}">
+        {{ chunk.chunk | escape }}
+      </div>
+
+      <div class="text-xs flex flex-wrap gap-x-4 gap-y-2 items-center">
+        <a hx-get="/content/{{ chunk.source_id }}/read" hx-target="#modal" hx-swap="innerHTML"
+          class="nb-link uppercase tracking-wide">
+          View full document
+        </a>
+
+        <span class="inline-flex items-center min-w-0">
+          <span class="uppercase tracking-wide opacity-60 mr-2">Source</span>
+          <span class="nb-badge truncate max-w-xs" title="{{ chunk.source_label }}">{{ chunk.source_label }}</span>
+        </span>
+
+        <span class="inline-flex items-center">
+          <span class="uppercase tracking-wide opacity-60 mr-2">Score</span>
+          <span class="nb-badge">{{ result.score | round(3) }}</span>
+        </span>
+      </div>
+    </div>
     {% elif result.result_type == "knowledge_entity" %}
     {% set entity = result.knowledge_entity %}
     <div
@@ -84,15 +122,12 @@
     <div class="flex-1 min-w-0">
       <h3 class="text-lg font-extrabold mb-1 leading-snug">
         <a hx-get="/knowledge-entity/{{ entity.id }}" hx-target="#modal" hx-swap="innerHTML" class="nb-link">
-          {% set entity_title = entity.highlighted_name | default(entity.name, true) %}
-          {{ entity_title }}
+          {{ entity.name }}
         </a>
       </h3>
 
       <div class="prose prose-tufte-compact text-base-content/80 mb-3 overflow-hidden line-clamp-6">
-        {% if entity.highlighted_description %}
-        {{ entity.highlighted_description }}
-        {% elif entity.description %}
+        {% if entity.description %}
         {{ entity.description | escape }}
         {% else %}
         <span class="italic opacity-60">No description available.</span>
@@ -107,14 +142,14 @@
 
         {% if entity.source_id %}
         <span class="inline-flex items-center min-w-0">
-          <span class="uppercase tracking-wide opacity-60 mr-2">Source ID</span>
-          <span class="nb-badge truncate max-w-xs" title="{{ entity.source_id }}">{{ entity.source_id }}</span>
+          <span class="uppercase tracking-wide opacity-60 mr-2">Source</span>
+          <span class="nb-badge truncate max-w-xs" title="{{ entity.source_label }}">{{ entity.source_label }}</span>
         </span>
         {% endif %}
 
         <span class="inline-flex items-center">
           <span class="uppercase tracking-wide opacity-60 mr-2">Score</span>
-          <span class="nb-badge">{{ result.score }}</span>
+          <span class="nb-badge">{{ result.score | round(3) }}</span>
         </span>
       </div>
     </div>
@@ -133,4 +168,4 @@
   <p class="text-lg font-semibold">Enter a term above to search your knowledge base.</p>
   <p class="text-sm opacity-70">Results will appear here.</p>
 </div>
-{% endif %}
+{% endif %}

ingestion-pipeline/src/pipeline/services.rs

@@ -189,7 +189,7 @@ impl PipelineServices for DefaultPipelineServices {
         match retrieval_pipeline::retrieve_entities(
             &self.db,
             &self.openai_client,
-            // embedding_provider_ref,
+            Some(&*self.embedding_provider),
             &input_text,
             &content.user_id,
             config,

main/src/main.rs

@@ -2,18 +2,19 @@ use api_router::{api_routes_v1, api_state::ApiState};
 use axum::{extract::FromRef, Router};
 use common::{
     storage::{
-        db::SurrealDbClient,
-        indexes::ensure_runtime_indexes,
-        store::StorageManager,
-        types::system_settings::SystemSettings,
+        db::SurrealDbClient, indexes::ensure_runtime_indexes, store::StorageManager,
+        types::{
+            knowledge_entity::KnowledgeEntity, system_settings::SystemSettings,
+            text_chunk::TextChunk,
+        },
     },
-    utils::config::get_config,
+    utils::{config::get_config, embedding::EmbeddingProvider},
 };
 use html_router::{html_routes, html_state::HtmlState};
 use ingestion_pipeline::{pipeline::IngestionPipeline, run_worker_loop};
 use retrieval_pipeline::reranking::RerankerPool;
 use std::sync::Arc;
-use tracing::{error, info};
+use tracing::{error, info, warn};
 use tracing_subscriber::{fmt, prelude::*, EnvFilter};
 
 use tokio::task::LocalSet;
@@ -44,8 +45,6 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
 
     // Ensure db is initialized
     db.apply_migrations().await?;
-    let settings = SystemSettings::get_current(&db).await?;
-    ensure_runtime_indexes(&db, settings.embedding_dimensions as usize).await?;
 
     let session_store = Arc::new(db.create_session_store().await?);
     let openai_client = Arc::new(async_openai::Client::with_config(
@@ -54,6 +53,57 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
             .with_api_base(&config.openai_base_url),
     ));
 
+    // Create embedding provider based on config before syncing settings.
+    let embedding_provider =
+        Arc::new(EmbeddingProvider::from_config(&config, Some(openai_client.clone())).await?);
+    info!(
+        embedding_backend = ?config.embedding_backend,
+        embedding_dimension = embedding_provider.dimension(),
+        "Embedding provider initialized"
+    );
+
+    // Sync SystemSettings with provider's dimensions/model/backend
+    let (settings, dimensions_changed) =
+        SystemSettings::sync_from_embedding_provider(&db, &embedding_provider).await?;
+
+    // Now ensure runtime indexes with the correct (synced) dimensions
+    ensure_runtime_indexes(&db, settings.embedding_dimensions as usize).await?;
+
+    // If dimensions changed, re-embed existing data to keep queries working.
+    if dimensions_changed {
+        warn!(
+            new_dimensions = settings.embedding_dimensions,
+            "Embedding configuration changed; re-embedding existing data"
+        );
+
+        // Re-embed text chunks
+        info!("Re-embedding TextChunks");
+        if let Err(e) = TextChunk::update_all_embeddings_with_provider(
+            &db,
+            &embedding_provider,
+        )
+        .await
+        {
+            error!("Failed to re-embed TextChunks: {}. Search results may be stale.", e);
+        }
+
+        // Re-embed knowledge entities
+        info!("Re-embedding KnowledgeEntities");
+        if let Err(e) = KnowledgeEntity::update_all_embeddings_with_provider(
+            &db,
+            &embedding_provider,
+        )
+        .await
+        {
+            error!(
+                "Failed to re-embed KnowledgeEntities: {}. Search results may be stale.",
+                e
+            );
+        }
+
+        info!("Re-embedding complete.");
+    }
+
     let reranker_pool = RerankerPool::maybe_from_config(&config)?;
 
     // Create global storage manager
@@ -66,6 +116,7 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
         storage.clone(),
         config.clone(),
         reranker_pool.clone(),
+        embedding_provider.clone(),
     )
     .await?;
 
@@ -114,9 +165,6 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
             .await
             .unwrap(),
         );
-        let settings = SystemSettings::get_current(&worker_db)
-            .await
-            .expect("failed to load system settings");
 
         // Initialize worker components
         let openai_client = Arc::new(async_openai::Client::with_config(
@@ -125,14 +173,11 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
                 .with_api_base(&config.openai_base_url),
         ));
 
-        // Create embedding provider for ingestion
+        // Create embedding provider based on config
         let embedding_provider = Arc::new(
-            common::utils::embedding::EmbeddingProvider::new_openai(
-                openai_client.clone(),
-                settings.embedding_model,
-                settings.embedding_dimensions,
-            )
-            .expect("failed to create embedding provider"),
+            EmbeddingProvider::from_config(&config, Some(openai_client.clone()))
+                .await
+                .expect("failed to create embedding provider"),
         );
         let ingestion_pipeline = Arc::new(
             IngestionPipeline::new(
@@ -226,6 +271,12 @@ mod tests {
             .await
             .expect("failed to build storage manager");
 
+        // Use hashed embeddings for tests to avoid external dependencies
+        let embedding_provider = Arc::new(
+            common::utils::embedding::EmbeddingProvider::new_hashed(384)
+                .expect("failed to create hashed embedding provider"),
+        );
+
         let html_state = HtmlState::new_with_resources(
             db.clone(),
             openai_client,
@@ -233,6 +284,7 @@ mod tests {
             storage.clone(),
             config.clone(),
             None,
+            embedding_provider,
         )
         .await
         .expect("failed to build html state");

main/src/server.rs

@@ -3,7 +3,8 @@ use std::sync::Arc;
 use api_router::{api_routes_v1, api_state::ApiState};
 use axum::{extract::FromRef, Router};
 use common::{
-    storage::db::SurrealDbClient, storage::store::StorageManager, utils::config::get_config,
+    storage::{db::SurrealDbClient, store::StorageManager, types::system_settings::SystemSettings},
+    utils::{config::get_config, embedding::EmbeddingProvider},
 };
 use html_router::{html_routes, html_state::HtmlState};
 use retrieval_pipeline::reranking::RerankerPool;
@@ -49,6 +50,20 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
     // Create global storage manager
     let storage = StorageManager::new(&config).await?;
 
+    // Create embedding provider based on config
+    let embedding_provider = Arc::new(
+        EmbeddingProvider::from_config(&config, Some(openai_client.clone())).await?,
+    );
+    info!(
+        embedding_backend = ?config.embedding_backend,
+        embedding_dimension = embedding_provider.dimension(),
+        "Embedding provider initialized"
+    );
+
+    // Sync SystemSettings with provider's dimensions/backend for visibility
+    let (_settings, _dimensions_changed) =
+        SystemSettings::sync_from_embedding_provider(&db, &embedding_provider).await?;
+
     let html_state = HtmlState::new_with_resources(
         db,
         openai_client,
@@ -56,6 +71,7 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
         storage.clone(),
         config.clone(),
         reranker_pool,
+        embedding_provider,
     )
     .await?;
 

main/src/worker.rs

@@ -1,10 +1,12 @@
 use std::sync::Arc;
 
 use common::{
-    storage::db::SurrealDbClient, storage::store::StorageManager, utils::config::get_config,
+    storage::db::SurrealDbClient, storage::store::StorageManager,
+    utils::{config::get_config, embedding::EmbeddingProvider},
 };
 use ingestion_pipeline::{pipeline::IngestionPipeline, run_worker_loop};
 use retrieval_pipeline::reranking::RerankerPool;
+use tracing::info;
 use tracing_subscriber::{fmt, prelude::*, EnvFilter};
 
 #[tokio::main]
@@ -37,9 +39,14 @@ async fn main() -> Result<(), Box<dyn std::error::Error>> {
 
     let reranker_pool = RerankerPool::maybe_from_config(&config)?;
 
-    // Create embedding provider for ingestion
-    let embedding_provider =
-        Arc::new(common::utils::embedding::EmbeddingProvider::new_fastembed(None).await?);
+    // Create embedding provider based on config
+    let embedding_provider = Arc::new(
+        EmbeddingProvider::from_config(&config, Some(openai_client.clone())).await?,
+    );
+    info!(
+        embedding_backend = ?config.embedding_backend,
+        "Embedding provider initialized for worker"
+    );
 
     // Create global storage manager
     let storage = StorageManager::new(&config).await?;

retrieval-pipeline/src/lib.rs

@@ -21,11 +21,29 @@ use tracing::instrument;
 pub enum StrategyOutput {
     Entities(Vec<RetrievedEntity>),
     Chunks(Vec<RetrievedChunk>),
+    Search(SearchResult),
+}
+
+/// Unified search result containing both chunks and entities
+#[derive(Debug, Clone)]
+pub struct SearchResult {
+    pub chunks: Vec<RetrievedChunk>,
+    pub entities: Vec<RetrievedEntity>,
+}
+
+impl SearchResult {
+    pub fn new(chunks: Vec<RetrievedChunk>, entities: Vec<RetrievedEntity>) -> Self {
+        Self { chunks, entities }
+    }
+
+    pub fn is_empty(&self) -> bool {
+        self.chunks.is_empty() && self.entities.is_empty()
+    }
 }
 
 pub use pipeline::{
     retrieved_entities_to_json, PipelineDiagnostics, PipelineStageTimings, RetrievalConfig,
-    RetrievalStrategy, RetrievalTuning,
+    RetrievalStrategy, RetrievalTuning, SearchTarget,
 };
 
 // Captures a supporting chunk plus its fused retrieval score for downstream prompts.
@@ -48,6 +66,7 @@ pub struct RetrievedEntity {
 pub async fn retrieve_entities(
     db_client: &SurrealDbClient,
     openai_client: &async_openai::Client<async_openai::config::OpenAIConfig>,
+    embedding_provider: Option<&common::utils::embedding::EmbeddingProvider>,
     input_text: &str,
     user_id: &str,
     config: RetrievalConfig,
@@ -56,7 +75,7 @@ pub async fn retrieve_entities(
     pipeline::run_pipeline(
         db_client,
         openai_client,
-        None,
+        embedding_provider,
         input_text,
         user_id,
         config,
@@ -252,4 +271,49 @@ mod tests {
             "Chunk results should contain relevant snippets"
         );
     }
+
+    #[tokio::test]
+    async fn test_search_strategy_returns_search_result() {
+        let db = setup_test_db().await;
+        let user_id = "search_user";
+        let chunk = TextChunk::new(
+            "search_src".into(),
+            "Async Rust programming uses Tokio runtime for concurrent tasks.".into(),
+            user_id.into(),
+        );
+
+        TextChunk::store_with_embedding(chunk.clone(), chunk_embedding_primary(), &db)
+            .await
+            .expect("Failed to store chunk");
+
+        let config = RetrievalConfig::for_search(pipeline::SearchTarget::Both);
+        let openai_client = Client::new();
+        let results = pipeline::run_pipeline_with_embedding(
+            &db,
+            &openai_client,
+            None,
+            test_embedding(),
+            "async rust programming",
+            user_id,
+            config,
+            None,
+        )
+        .await
+        .expect("Search strategy retrieval failed");
+
+        let search_result = match results {
+            StrategyOutput::Search(sr) => sr,
+            other => panic!("expected Search output, got {:?}", other),
+        };
+
+        // Should return chunks (entities may be empty if none stored)
+        assert!(
+            !search_result.chunks.is_empty(),
+            "Search strategy should return chunks"
+        );
+        assert!(
+            search_result.chunks.iter().any(|c| c.chunk.chunk.contains("Tokio")),
+            "Search results should contain relevant chunks"
+        );
+    }
 }

retrieval-pipeline/src/pipeline/config.rs

@@ -12,6 +12,21 @@ pub enum RetrievalStrategy {
     RelationshipSuggestion,
     /// Entity retrieval for context during content ingestion
     Ingestion,
+    /// Unified search returning both chunks and entities
+    Search,
+}
+
+/// Configures which result types to include in Search strategy
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize, Default)]
+#[serde(rename_all = "snake_case")]
+pub enum SearchTarget {
+    /// Return only text chunks
+    ChunksOnly,
+    /// Return only knowledge entities
+    EntitiesOnly,
+    /// Return both chunks and entities (default)
+    #[default]
+    Both,
 }
 
 impl Default for RetrievalStrategy {
@@ -37,6 +52,7 @@ impl std::str::FromStr for RetrievalStrategy {
             }
             "relationship_suggestion" => Ok(Self::RelationshipSuggestion),
             "ingestion" => Ok(Self::Ingestion),
+            "search" => Ok(Self::Search),
             other => Err(format!("unknown retrieval strategy '{other}'")),
         }
     }
@@ -48,6 +64,7 @@ impl fmt::Display for RetrievalStrategy {
             RetrievalStrategy::Default => "default",
             RetrievalStrategy::RelationshipSuggestion => "relationship_suggestion",
             RetrievalStrategy::Ingestion => "ingestion",
+            RetrievalStrategy::Search => "search",
         };
         f.write_str(label)
     }
@@ -140,6 +157,8 @@ impl Default for RetrievalTuning {
 pub struct RetrievalConfig {
     pub strategy: RetrievalStrategy,
     pub tuning: RetrievalTuning,
+    /// Target for Search strategy (chunks, entities, or both)
+    pub search_target: SearchTarget,
 }
 
 impl RetrievalConfig {
@@ -147,6 +166,7 @@ impl RetrievalConfig {
         Self {
             strategy: RetrievalStrategy::Default,
             tuning,
+            search_target: SearchTarget::default(),
         }
     }
 
@@ -154,11 +174,16 @@ impl RetrievalConfig {
         Self {
             strategy,
             tuning: RetrievalTuning::default(),
+            search_target: SearchTarget::default(),
         }
     }
 
     pub fn with_tuning(strategy: RetrievalStrategy, tuning: RetrievalTuning) -> Self {
-        Self { strategy, tuning }
+        Self {
+            strategy,
+            tuning,
+            search_target: SearchTarget::default(),
+        }
     }
 
     /// Create config for chat retrieval with strategy selection support
@@ -175,6 +200,15 @@ impl RetrievalConfig {
     pub fn for_ingestion() -> Self {
         Self::with_strategy(RetrievalStrategy::Ingestion)
     }
+
+    /// Create config for unified search (chunks and/or entities)
+    pub fn for_search(target: SearchTarget) -> Self {
+        Self {
+            strategy: RetrievalStrategy::Search,
+            tuning: RetrievalTuning::default(),
+            search_target: target,
+        }
+    }
 }
 
 impl Default for RetrievalConfig {
@@ -182,6 +216,7 @@ impl Default for RetrievalConfig {
         Self {
             strategy: RetrievalStrategy::default(),
             tuning: RetrievalTuning::default(),
+            search_target: SearchTarget::default(),
         }
     }
 }

retrieval-pipeline/src/pipeline/mod.rs

@@ -3,7 +3,7 @@ mod diagnostics;
 mod stages;
 mod strategies;
 
-pub use config::{RetrievalConfig, RetrievalStrategy, RetrievalTuning};
+pub use config::{RetrievalConfig, RetrievalStrategy, RetrievalTuning, SearchTarget};
 pub use diagnostics::{
     AssembleStats, ChunkEnrichmentStats, CollectCandidatesStats, EntityAssemblyTrace,
     PipelineDiagnostics,
@@ -17,7 +17,7 @@ use std::time::{Duration, Instant};
 use tracing::info;
 
 use stages::PipelineContext;
-use strategies::{DefaultStrategyDriver, IngestionDriver, RelationshipSuggestionDriver};
+use strategies::{DefaultStrategyDriver, IngestionDriver, RelationshipSuggestionDriver, SearchStrategyDriver};
 
 // Export StrategyOutput publicly from this module
 // (it's defined in lib.rs but we re-export it here)
@@ -181,6 +181,24 @@ pub async fn run_pipeline(
             .await?;
             Ok(StrategyOutput::Entities(run.results))
         }
+        RetrievalStrategy::Search => {
+            let search_target = config.search_target;
+            let driver = SearchStrategyDriver::new(search_target);
+            let run = execute_strategy(
+                driver,
+                db_client,
+                openai_client,
+                embedding_provider,
+                None,
+                input_text,
+                user_id,
+                config,
+                reranker,
+                false,
+            )
+            .await?;
+            Ok(StrategyOutput::Search(run.results))
+        }
     }
 }
 
@@ -246,6 +264,24 @@ pub async fn run_pipeline_with_embedding(
             .await?;
             Ok(StrategyOutput::Entities(run.results))
         }
+        RetrievalStrategy::Search => {
+            let search_target = config.search_target;
+            let driver = SearchStrategyDriver::new(search_target);
+            let run = execute_strategy(
+                driver,
+                db_client,
+                openai_client,
+                embedding_provider,
+                Some(query_embedding),
+                input_text,
+                user_id,
+                config,
+                reranker,
+                false,
+            )
+            .await?;
+            Ok(StrategyOutput::Search(run.results))
+        }
     }
 }
 

retrieval-pipeline/src/pipeline/strategies.rs

@@ -88,3 +88,63 @@ impl StrategyDriver for IngestionDriver {
         Ok(ctx.take_entity_results())
     }
 }
+
+use crate::SearchResult;
+use super::config::SearchTarget;
+
+/// Search strategy driver that retrieves both chunks and entities
+pub struct SearchStrategyDriver {
+    target: SearchTarget,
+}
+
+impl SearchStrategyDriver {
+    pub fn new(target: SearchTarget) -> Self {
+        Self { target }
+    }
+}
+
+impl StrategyDriver for SearchStrategyDriver {
+    type Output = SearchResult;
+
+    fn stages(&self) -> Vec<BoxedStage> {
+        match self.target {
+            SearchTarget::ChunksOnly => vec![
+                Box::new(EmbedStage),
+                Box::new(ChunkVectorStage),
+                Box::new(ChunkRerankStage),
+                Box::new(ChunkAssembleStage),
+            ],
+            SearchTarget::EntitiesOnly => vec![
+                Box::new(EmbedStage),
+                Box::new(CollectCandidatesStage),
+                Box::new(GraphExpansionStage),
+                Box::new(RerankStage),
+                Box::new(AssembleEntitiesStage),
+            ],
+            SearchTarget::Both => vec![
+                Box::new(EmbedStage),
+                // Chunk retrieval path
+                Box::new(ChunkVectorStage),
+                Box::new(ChunkRerankStage),
+                Box::new(ChunkAssembleStage),
+                // Entity retrieval path (runs after chunk stages)
+                Box::new(CollectCandidatesStage),
+                Box::new(GraphExpansionStage),
+                Box::new(RerankStage),
+                Box::new(AssembleEntitiesStage),
+            ],
+        }
+    }
+
+    fn finalize(&self, ctx: &mut PipelineContext<'_>) -> Result<Self::Output, AppError> {
+        let chunks = match self.target {
+            SearchTarget::EntitiesOnly => Vec::new(),
+            _ => ctx.take_chunk_results(),
+        };
+        let entities = match self.target {
+            SearchTarget::ChunksOnly => Vec::new(),
+            _ => ctx.take_entity_results(),
+        };
+        Ok(SearchResult::new(chunks, entities))
+    }
+}