Struct SemanticQueryCache 

pub struct SemanticQueryCache { /* private fields */ }

Semantic query cache

Implementations

impl SemanticQueryCache

pub fn new(threshold: f32) -> Self

Create a new semantic query cache with default max entries

pub fn with_capacity(threshold: f32, max_entries: usize) -> Self

Create a new semantic query cache with specified capacity

pub fn with_config( threshold: f32, max_entries: usize, index_config: SemanticIndexConfig, ) -> Self

Create with custom index config

pub fn lookup(&self, embedding: &[f32]) -> Option<SimilarityResult>

Lookup by semantic similarity

pub fn lookup_with_threshold( &self, embedding: &[f32], threshold: f32, ) -> Option<SimilarityResult>

Lookup with custom threshold

pub fn find_similar(&self, embedding: &[f32], k: usize) -> Vec<SimilarityResult>

Find k most similar queries

pub fn lookup_with_branch( &self, embedding: &[f32], branch: &BranchContext, ) -> Option<SimilarityResult>

Lookup with branch context filtering

Returns cached entry only if it’s compatible with the given branch context. This enables branch-aware caching for time-travel queries.

pub fn lookup_with_session( &self, embedding: &[f32], session: &String, ) -> Option<SimilarityResult>

Lookup with session affinity for agent conversations

Prioritizes entries from the same session for better conversation context.

pub fn lookup_with_context( &self, embedding: &[f32], branch: Option<&BranchContext>, session: Option<&String>, workload: AIWorkloadContext, ) -> Option<SimilarityResult>

Lookup with full AI context (branch + session + workload)

Most comprehensive lookup that considers:

• Branch context for time-travel
• Session affinity for agent conversations
• Workload type for TTL and priority

pub fn find_similar_in_branch( &self, embedding: &[f32], branch: &BranchContext, k: usize, ) -> Vec<SimilarityResult>

Find similar entries within a branch

pub fn invalidate_by_table(&self, table: &str) -> usize

Invalidate entries by table name

Used when WAL invalidation detects changes to a table.

pub fn invalidate_branch(&self, branch: &BranchId) -> usize

Invalidate entries by branch

pub fn insert( &self, query: impl Into<String>, embedding: Embedding, result: Value, ) -> VectorId

Insert a new entry

pub fn insert_with_ttl( &self, query: impl Into<String>, embedding: Embedding, result: Value, ttl: Duration, ) -> VectorId

Insert with TTL

pub fn insert_with_context( &self, query: impl Into<String>, embedding: Embedding, result: Value, branch: Option<BranchContext>, session: Option<String>, workload: AIWorkloadContext, tables: Vec<String>, ) -> VectorId

Insert with full AI context (branch, session, workload, tables)

This is the recommended insertion method for AI/Agent workloads as it enables branch-aware caching, session affinity, and workload-specific TTL management.

pub fn insert_rag_retrieval( &self, query: impl Into<String>, embedding: Embedding, result: Value, tables: Vec<String>, ) -> VectorId

Insert for RAG retrieval workload

Optimized TTL for fast-refresh retrieval phase.

pub fn insert_agent_response( &self, query: impl Into<String>, embedding: Embedding, result: Value, session: String, branch: Option<BranchContext>, ) -> VectorId

Insert for agent conversation

Session-aware with longer TTL for conversation context.

pub fn insert_tool_result( &self, query: impl Into<String>, embedding: Embedding, result: Value, ) -> VectorId

Insert deterministic tool result

Long TTL for deterministic tool calls (e.g., math, date formatting).

pub fn remove(&self, id: VectorId)

Remove an entry

pub fn cleanup_expired(&self)

Remove expired entries

pub fn clear(&self)

Clear all entries

pub fn len(&self) -> usize

Get entry count

pub fn is_empty(&self) -> bool

Check if empty

pub fn stats(&self) -> SemanticCacheStatsSnapshot

Get statistics