Merge branch 'main' into vi-test

libs/community/langchain_google_community/bq_storage_vectorstores/bigquery.py

@@ -301,7 +301,7 @@ def _create_search_query(
         if table_to_query is not None:
             embeddings_query = f"""
             with embeddings as (
-            SELECT {self.embedding_field}, ROW_NUMBER() OVER() as row_num
+            SELECT {self.embedding_field}, row_num
             from `{table_to_query}`
             )"""
 
@@ -390,14 +390,16 @@ def _create_temp_bq_table(
         df = pd.DataFrame([])
 
         df[self.embedding_field] = embeddings
+        df["row_num"] = list(range(len(df)))
         table_id = (
             f"{self.project_id}."
             f"{self.temp_dataset_name}."
             f"{self.table_name}_{uuid.uuid4().hex}"
         )
 
         schema = [
-            bigquery.SchemaField(self.embedding_field, "FLOAT64", mode="REPEATED")
+            bigquery.SchemaField(self.embedding_field, "FLOAT64", mode="REPEATED"),
+            bigquery.SchemaField("row_num", "INT64"),
         ]
         table_ref = bigquery.Table(table_id, schema=schema)
         table = self._bq_client.create_table(table_ref)
@@ -483,7 +485,7 @@ def batch_search(
             )
 
         if queries is not None:
-            embeddings = self.embedding.embed_documents(queries)
+            embeddings = [self.embedding.embed_query(query) for query in queries]
 
         if embeddings is None:
             raise ValueError("Could not obtain embeddings - value is None.")

libs/genai/langchain_google_genai/_function_utils.py

@@ -342,8 +342,6 @@ def _get_items_from_schema(schema: Union[Dict, List, str]) -> Dict[str, Any]:
         items["type_"] = _get_type_from_schema(schema)
         if items["type_"] == glm.Type.OBJECT and "properties" in schema:
             items["properties"] = _get_properties_from_schema_any(schema["properties"])
-        if "title" in schema:
-            items["title"] = schema
         if "title" in schema or "description" in schema:
             items["description"] = (
                 schema.get("description") or schema.get("title") or ""

libs/genai/langchain_google_genai/chat_models.py

@@ -1461,7 +1461,11 @@ def create_cached_content(
 
     @property
     def _supports_tool_choice(self) -> bool:
-        return "gemini-1.5-pro" in self.model or "gemini-1.5-flash" in self.model
+        return (
+            "gemini-1.5-pro" in self.model
+            or "gemini-1.5-flash" in self.model
+            or "gemini-2" in self.model
+        )
 
 
 def _get_tool_name(

libs/genai/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langchain-google-genai"
-version = "2.0.6"
+version = "2.0.7"
 description = "An integration package connecting Google's genai package and LangChain"
 authors = []
 readme = "README.md"

libs/genai/tests/integration_tests/test_standard.py

@@ -11,6 +11,36 @@
 from langchain_google_genai import ChatGoogleGenerativeAI
 
 rate_limiter = InMemoryRateLimiter(requests_per_second=0.25)
+rate_limiter_2_0 = InMemoryRateLimiter(requests_per_second=0.1)
+
+
+class TestGeminiAI2Standard(ChatModelIntegrationTests):
+    @property
+    def chat_model_class(self) -> Type[BaseChatModel]:
+        return ChatGoogleGenerativeAI
+
+    @property
+    def chat_model_params(self) -> dict:
+        return {
+            "model": "models/gemini-2.0-flash-exp",
+            "rate_limiter": rate_limiter_2_0,
+        }
+
+    @pytest.mark.xfail(reason="with_structured_output with JSON schema not supported.")
+    async def test_structured_output_async(self, model: BaseChatModel) -> None:
+        await super().test_structured_output_async(model)
+
+    @pytest.mark.xfail(reason="with_structured_output with JSON schema not supported.")
+    def test_structured_output(self, model: BaseChatModel) -> None:
+        super().test_structured_output(model)
+
+    @pytest.mark.xfail(reason="with_structured_output with JSON schema not supported.")
+    def test_structured_output_pydantic_2_v1(self, model: BaseChatModel) -> None:
+        super().test_structured_output_pydantic_2_v1(model)
+
+    @pytest.mark.xfail(reason="investigate")
+    def test_bind_runnables_as_tools(self, model: BaseChatModel) -> None:
+        super().test_bind_runnables_as_tools(model)
 
 
 class TestGeminiAIStandard(ChatModelIntegrationTests):

libs/vertexai/langchain_google_vertexai/_utils.py

@@ -131,7 +131,11 @@ def _missing_(cls, value: Any) -> "GoogleModelFamily":
             "medlm-large-1.5-001",
             "medlm-large-1.5@001",
         }
-        if "gemini-1.5" in model_name or model_name in gemini_advanced_models:
+        if (
+            "gemini-1.5" in model_name
+            or model_name in gemini_advanced_models
+            or "gemini-2" in model_name
+        ):
             return GoogleModelFamily.GEMINI_ADVANCED
         if "gemini" in model_name:
             return GoogleModelFamily.GEMINI

libs/vertexai/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langchain-google-vertexai"
-version = "2.0.8"
+version = "2.0.9"
 description = "An integration package connecting Google VertexAI and LangChain"
 authors = []
 readme = "README.md"

libs/vertexai/tests/integration_tests/test_standard.py

@@ -12,6 +12,33 @@
 rate_limiter = InMemoryRateLimiter(requests_per_second=0.5)
 
 
+@pytest.mark.first
+class TestGemini2AIStandard(ChatModelIntegrationTests):
+    @property
+    def chat_model_class(self) -> Type[BaseChatModel]:
+        return ChatVertexAI
+
+    @property
+    def chat_model_params(self) -> dict:
+        return {
+            "model_name": "gemini-2.0-flash-exp",
+            "rate_limiter": rate_limiter,
+            "temperature": 0,
+        }
+
+    @property
+    def supports_image_inputs(self) -> bool:
+        return True
+
+    @property
+    def supports_video_inputs(self) -> bool:
+        return True
+
+    @property
+    def supports_audio_inputs(self) -> bool:
+        return True
+
+
 @pytest.mark.first
 class TestGeminiAIStandard(ChatModelIntegrationTests):
     @property