Struct QueryEngine 

pub struct QueryEngine { /* private fields */ }

Query engine for semantic search

Implementations

impl QueryEngine

pub fn new( graph: Arc<Mutex<GraphStore>>, embedder: Arc<EmbeddingEngine>, ) -> Self

Create a new query engine

pub fn with_cache_capacity( graph: Arc<Mutex<GraphStore>>, embedder: Arc<EmbeddingEngine>, cache_capacity: usize, ) -> Self

Create a new query engine with custom cache capacity

pub fn with_context_budget(self, budget: usize) -> Self

Set the context token budget (builder pattern)

pub fn with_intelligence( self, intelligence: Box<dyn IntelligenceEngine>, ) -> Self

Set the intelligence engine for RAG (builder pattern)

pub fn set_intelligence(&mut self, intelligence: Box<dyn IntelligenceEngine>)

Set or replace the intelligence engine after creation

pub fn search(&self, query: &str, k: usize) -> Result<Vec<SearchResult>>

Search for relevant chunks using a hybrid (semantic + lexical) approach

Uses integer Reciprocal Rank Fusion (RRF) per CP-003 §7 for deterministic results: Score(d) = 1,000,000 / (k + rank(d))

pub fn search_semantic( &self, query: &str, k: usize, ) -> Result<Vec<SearchResult>>

Perform purely semantic search (vector only)

pub fn search_lexical(&self, query: &str, k: usize) -> Result<Vec<SearchResult>>

Perform purely lexical search (keyword only)

pub fn search_filtered( &self, query: &str, k: usize, filters: &[Filter], ) -> Result<Vec<SearchResult>>

Search with filters applied

Per CP-012: Supports filtering by document path, MIME type, and modification time.

pub fn search_cached(&self, query: &str, k: usize) -> Result<Vec<SearchResult>>

Search with caching

Per CP-020: Uses query cache for faster repeated queries.

pub fn invalidate_cache(&self) -> Result<()>

Invalidate the query cache

pub fn get_chunks_for_document(&self, doc_id: Uuid) -> Result<Vec<SearchResult>>

Get all chunks for a specific document

pub fn graph(&self) -> Arc<Mutex<GraphStore>>

Access the underlying graph store (for testing/debugging)

pub async fn generate_answer(&self, query: &str) -> Result<GenerationResult>

Generate an answer using the knowledge graph and an LLM

pub fn generate_proof_receipt( &self, query: &str, search_results: &[SearchResult], identity: &DeviceIdentity, ) -> Result<ProofReceipt>

Generate a cryptographic proof receipt for a query.

This creates a signed, verifiable record of exactly what context was available when a search was performed. The receipt includes Merkle proofs for each chunk, allowing independent verification.

pub async fn chat(&self, query: &str, history: &[Message]) -> Result<String>

Chat with context from the knowledge graph.

Performs RAG search, prepends conversation history to the context, and delegates to generate_answer for the LLM call.

impl QueryEngine

pub fn live_search( &self, query: &str, remote_results: &[VerifiedRemoteResult], max_results: usize, ) -> Result<Vec<MergedSearchResult>>

Perform a live search that dispatches the query to the local index and to remote peers in parallel, merging results via cross-source RRF.

This is the primary search entrypoint when the node has peer connections. Remote results that arrive within timeout are merged. Local results are always included regardless of remote availability.