google_genai.py — langchain Source File

Source Code

"""Derivations of standard content blocks from Google (GenAI) content."""

import base64
import re
from collections.abc import Iterator
from typing import Any, cast

from langchain_core.messages import AIMessage, AIMessageChunk
from langchain_core.messages import content as types
from langchain_core.messages.content import Citation, create_citation

try:
    import filetype  # type: ignore[import-not-found]

    _HAS_FILETYPE = True
except ImportError:
    _HAS_FILETYPE = False


def _bytes_to_b64_str(bytes_: bytes) -> str:
    """Convert bytes to base64 encoded string."""
    return base64.b64encode(bytes_).decode("utf-8")


def translate_grounding_metadata_to_citations(
    grounding_metadata: dict[str, Any],
) -> list[Citation]:
    """Translate Google AI grounding metadata to LangChain Citations.

    Args:
        grounding_metadata: Google AI grounding metadata containing web search
            queries, grounding chunks, and grounding supports.

    Returns:
        List of Citation content blocks derived from the grounding metadata.

    Example:
        >>> metadata = {
        ...     "web_search_queries": ["UEFA Euro 2024 winner"],
        ...     "grounding_chunks": [
        ...         {
        ...             "web": {
        ...                 "uri": "https://uefa.com/euro2024",
        ...                 "title": "UEFA Euro 2024 Results",
        ...             }
        ...         }
        ...     ],
        ...     "grounding_supports": [
        ...         {
        ...             "segment": {
        ...                 "start_index": 0,
        ...                 "end_index": 47,
        ...                 "text": "Spain won the UEFA Euro 2024 championship",
        ...             },
        ...             "grounding_chunk_indices": [0],
        ...         }
        ...     ],
        ... }
        >>> citations = translate_grounding_metadata_to_citations(metadata)
        >>> len(citations)
// ... (507 more lines)