Skip to main content Module embed Copy item path Source HnswResult Query result from HNSW search. Model2VecEmbedder Neural-quality embedder using Model2Vec (distilled sentence transformer). TrigramEmbedder Lightweight, zero-dependency embedder using character n-gram hashing. EmbedProvider Embedding engine trait. Implementations can use ONNX, API calls, etc. best_embedder Create the best available embedder: Model2Vec if possible, fallback to trigram. build_hnsw_index Build an HNSW index from embeddings and save to disk.
Returns the number of vectors indexed. code_embedder Singleton lazy-init code embedder (Arc-based, shared across threads). cosine_similarity Dot-product similarity — equivalent to cosine when vectors are L2-normalized. doc_embedder Singleton lazy-init doc embedder (Arc-based, shared across threads). embedding_count Count the number of embeddings in the binary file at root/.infigraph/embeddings.bin. init_embedder Factory: select Model2Vec if available, otherwise fall back to TrigramEmbedder. invalidate_embeddings_cache Invalidate the embeddings cache (call after save_embeddings or update_embeddings). invalidate_hnsw_cache Invalidate the HNSW cache. load_embeddings Load symbol embeddings from a binary file. load_embeddings_cached Load embeddings with process-lifetime caching. Returns cached data if the
file hasn’t been modified since last load. Falls back to load_embeddings
on any cache miss. path_to_context Extract meaningful context from a file path by filtering out common directory names. rich_symbol_text Build a rich text representation of a symbol for embedding, including file context. rich_symbol_text_full Extended rich text representation with parameter and return type info. save_embeddings Save symbol embeddings to a binary file.
Format: [count:u32] then for each entry: [id_len:u32][id_bytes][dim:u32][f32 * dim] search_hnsw Search the HNSW index, returning top-k results by inner product similarity.
Returns None if no valid index exists (caller should fall back to brute-force). symbol_text Build a text representation of a symbol for embedding. update_embeddings Incrementally update embeddings for changed files.