Function resolve_stored_document_parser_engine

lightrag/parser/routing.py:1207–1234

Resolve the parser engine key for a full_docs row during processing. Returns a registry key: the internal ``reuse``/``passthrough`` handlers for the no-op formats, or a real engine name for ``pending_parse``. Never raises and never filters an unknown stored engine — the parse worker

(
    file_path: str | Path,
    content_data: dict[str, Any] | None,
)

def resolve_stored_document_parser_engine(
    file_path: str | Path,
    content_data: dict[str, Any] | None,
) -> str:
    """Resolve the parser engine key for a full_docs row during processing.

    Returns a registry key: the internal ``reuse``/``passthrough`` handlers for
    the no-op formats, or a real engine name for ``pending_parse``.  Never
    raises and never filters an unknown stored engine — the parse worker
    decides fallback/validation (so its warning path actually fires).
    """
    if content_data:
        doc_format = content_data.get("parse_format", FULL_DOCS_FORMAT_RAW)
        # All lightrag rows reuse the already-parsed sidecar (sidecar optional;
        # ReuseParser tolerates a missing one). Reached on resume/retry.
        if doc_format == FULL_DOCS_FORMAT_LIGHTRAG:
            return PARSER_ENGINE_REUSE
        # Anything not pending is already-extracted content (raw direct-insert
        # or legacy-extracted RAW) -> pass through verbatim.
        if doc_format != FULL_DOCS_FORMAT_PENDING_PARSE:
            return PARSER_ENGINE_PASSTHROUGH
        # PENDING_PARSE: honour the stored engine verbatim; fall back to
        # filename rules only when it is absent.
        pending_engine = normalize_parser_engine(content_data.get("parse_engine"))
        if pending_engine:
            return pending_engine

    return resolve_file_parser_engine(file_path)