Function _embedding_search

code_review_graph/search.py:212–247

Run a vector similarity search using the embedding store. Returns list of ``(node_id, similarity_score)`` tuples. Gracefully returns an empty list if embeddings are not available.

(
    store: GraphStore,
    query: str,
    limit: int = 50,
    model: str | None = None,
    provider: str | None = None,
)

def _embedding_search(
    store: GraphStore,
    query: str,
    limit: int = 50,
    model: str | None = None,
    provider: str | None = None,
) -> list[tuple[int, float]]:
    """Run a vector similarity search using the embedding store.

    Returns list of ``(node_id, similarity_score)`` tuples.
    Gracefully returns an empty list if embeddings are not available.
    """
    try:
        from .embeddings import EmbeddingStore
    except ImportError:
        return []

    try:
        emb_store = EmbeddingStore(store.db_path, provider=provider, model=model)
        try:
            if not emb_store.available or emb_store.count() == 0:
                return []

            results = emb_store.search(query, limit=limit)
            # Map qualified names back to node IDs
            id_scores: list[tuple[int, float]] = []
            for qn, score in results:
                node = store.get_node(qn)
                if node:
                    id_scores.append((node.id, score))
            return id_scores
        finally:
            emb_store.close()
    except Exception as e:
        logger.warning("Embedding search failed: %s", e)
        return []


# ---------------------------------------------------------------------------