Function semantic_search_nodes

code_review_graph/tools/query.py:376–448

Search for nodes by name, keyword, or semantic similarity. Uses hybrid search (FTS5 BM25 + vector embeddings merged via Reciprocal Rank Fusion) as the primary search path, with graceful fallback to keyword matching. Args: query: Search string to match against node names and

(
    query: str,
    kind: str | None = None,
    limit: int = 20,
    repo_root: str | None = None,
    context_files: list[str] | None = None,
    model: str | None = None,
    provider: str | None = None,
    detail_level: str = "standard",
)

def semantic_search_nodes(
    query: str,
    kind: str | None = None,
    limit: int = 20,
    repo_root: str | None = None,
    context_files: list[str] | None = None,
    model: str | None = None,
    provider: str | None = None,
    detail_level: str = "standard",
) -> dict[str, Any]:
    """Search for nodes by name, keyword, or semantic similarity.

    Uses hybrid search (FTS5 BM25 + vector embeddings merged via Reciprocal
    Rank Fusion) as the primary search path, with graceful fallback to
    keyword matching.

    Args:
        query: Search string to match against node names and qualified names.
        kind: Optional filter by node kind (File, Class, Function, Type, Test).
        limit: Maximum results to return (default: 20).
        repo_root: Repository root path. Auto-detected if omitted.
        context_files: Optional list of file paths. Nodes in these files
            receive a relevance boost.
        detail_level: "standard" (full output) or "minimal" (summary only).

    Returns:
        Ranked list of matching nodes.
    """
    store, root = _get_store(repo_root)
    try:
        results = hybrid_search(
            store, query, kind=kind, limit=limit, context_files=context_files,
            model=model, provider=provider,
        )

        search_mode = "hybrid"
        if not results:
            search_mode = "keyword"

        summary = f"Found {len(results)} node(s) matching '{query}'" + (
            f" (kind={kind})" if kind else ""
        )

        if detail_level == "minimal":
            minimal_results = [
                {
                    k: r[k]
                    for k in ("name", "kind", "file_path", "score")
                    if k in r
                }
                for r in results[:5]
            ]
            return {
                "status": "ok",
                "query": query,
                "search_mode": search_mode,
                "summary": summary,
                "results": minimal_results,