Merge branch 'main' into fix-missing-scope-exception

libs/community/langchain_google_community/drive.py

@@ -27,6 +27,8 @@ class GoogleDriveLoader(BaseLoader, BaseModel):
     """Path to the credentials file."""
     token_path: Path = Path.home() / ".credentials" / "token.json"
     """Path to the token file."""
+    credentials: Any = None
+    """Your own google credentials created via your own mechanism"""
     folder_id: Optional[str] = None
     """The folder id to load from."""
     document_ids: Optional[List[str]] = None
@@ -276,6 +278,11 @@ def _load_credentials(self) -> Any:
         if self.token_path.exists():
             creds = Credentials.from_authorized_user_file(str(self.token_path), SCOPES)
 
+        if self.credentials:
+            # use whatever was passed to us
+            creds = self.credentials
+            return creds
+
         if not creds or not creds.valid:
             if creds and creds.expired and creds.refresh_token:
                 creds.refresh(Request())

libs/community/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langchain-google-community"
-version = "2.0.2"
+version = "2.0.3"
 description = "An integration package connecting miscellaneous Google's products and LangChain"
 authors = []
 readme = "README.md"

libs/genai/langchain_google_genai/_image_utils.py

@@ -8,6 +8,7 @@
 from typing import Any, Dict
 from urllib.parse import urlparse
 
+import filetype  # type: ignore[import]
 import requests
 from google.ai.generativelanguage_v1beta.types import Part
 
@@ -87,7 +88,13 @@ def load_part(self, image_string: str) -> Part:
             raise ValueError(msg)
 
         inline_data: Dict[str, Any] = {"data": bytes_}
+
         mime_type, _ = mimetypes.guess_type(image_string)
+        if not mime_type:
+            kind = filetype.guess(bytes_)
+            if kind:
+                mime_type = kind.mime
+
         if mime_type:
             inline_data["mime_type"] = mime_type
 

libs/genai/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langchain-google-genai"
-version = "2.0.5"
+version = "2.0.6"
 description = "An integration package connecting Google's genai package and LangChain"
 authors = []
 readme = "README.md"
@@ -15,6 +15,7 @@ python = ">=3.9,<4.0"
 langchain-core = ">=0.3.15,<0.4"
 google-generativeai = "^0.8.0"
 pydantic = ">=2,<3"
+filetype = "^1.2.0"
 
 [tool.poetry.group.test]
 optional = true

libs/genai/tests/integration_tests/test_chat_models.py

@@ -158,6 +158,33 @@ def test_chat_google_genai_invoke_multimodal() -> None:
         assert len(chunk.content.strip()) > 0
 
 
+def test_chat_google_genai_invoke_multimodal_by_url() -> None:
+    messages: list = [
+        HumanMessage(
+            content=[
+                {
+                    "type": "text",
+                    "text": "Guess what's in this picture! You have 3 guesses.",
+                },
+                {
+                    "type": "image_url",
+                    "image_url": "https://picsum.photos/seed/picsum/200/300",
+                },
+            ]
+        ),
+    ]
+    llm = ChatGoogleGenerativeAI(model=_VISION_MODEL)
+    response = llm.invoke(messages)
+    assert isinstance(response.content, str)
+    assert len(response.content.strip()) > 0
+
+    # Try streaming
+    for chunk in llm.stream(messages):
+        print(chunk)  # noqa: T201
+        assert isinstance(chunk.content, str)
+        assert len(chunk.content.strip()) > 0
+
+
 def test_chat_google_genai_invoke_multimodal_multiple_messages() -> None:
     messages: list = [
         HumanMessage(content="Hi there"),

libs/vertexai/langchain_google_vertexai/_anthropic_utils.py

@@ -31,7 +31,9 @@
 from pydantic import BaseModel
 
 if TYPE_CHECKING:
-    from anthropic.types import RawMessageStreamEvent  # type: ignore
+    from anthropic.types import (
+        RawMessageStreamEvent,  # type: ignore[unused-ignore, import-not-found]
+    )
 
 _message_type_lookups = {
     "human": "user",

libs/vertexai/langchain_google_vertexai/_image_utils.py

@@ -4,6 +4,7 @@
 import os
 import re
 from enum import Enum
+from functools import cached_property
 from typing import Dict, Optional, Union
 from urllib.parse import urlparse
 
@@ -43,6 +44,10 @@ def __init__(
         """
         self._project = project
 
+    @cached_property
+    def _storage_client(self):
+        return storage.Client(project=self._project)
+
     def load_bytes(self, image_string: str) -> bytes:
         """Routes to the correct loader based on the image_string.
 
@@ -198,7 +203,7 @@ def _blob_from_gcs(self, gcs_uri: str) -> storage.Blob:
             storage.Blob
         """
 
-        gcs_client = storage.Client(project=self._project)
+        gcs_client = self._storage_client
         blob = storage.Blob.from_string(gcs_uri, gcs_client)
         blob.reload(client=gcs_client)
         return blob

libs/vertexai/langchain_google_vertexai/model_garden.py

@@ -148,7 +148,7 @@ def __init__(self, **kwargs: Any) -> None:
 
     @model_validator(mode="after")
     def validate_environment(self) -> Self:
-        from anthropic import (  # type: ignore
+        from anthropic import (  # type: ignore[unused-ignore, import-not-found]
             AnthropicVertex,
             AsyncAnthropicVertex,
         )

libs/vertexai/langchain_google_vertexai/vectorstores/document_storage.py

@@ -292,10 +292,15 @@ def mset(self, key_value_pairs: Sequence[Tuple[str, Document]]) -> None:
             entities = []
             for key, document in zip(keys, documents):
                 entity = self._client.entity(
-                    key=key, exclude_from_indexes=self.exclude_from_indexes
+                    key=key,
+                    exclude_from_indexes=self.exclude_from_indexes
+                    if self.exclude_from_indexes
+                    else [],
                 )
                 metadata_entity = self._client.entity(
                     exclude_from_indexes=self.exclude_from_indexes
+                    if self.exclude_from_indexes
+                    else []
                 )
                 metadata_entity.update(document.metadata)
                 entity[self._text_property_name] = document.page_content