feat(genai)!: migration to google-genai

.github/CONTRIBUTING.md

@@ -171,6 +171,16 @@ To run integration tests:
 make integration_tests
 ```
 
+#### Backend Selection for Integration Tests
+
+For `libs/genai`, integration tests can run against different backends using the `TEST_VERTEXAI` environment variable:
+
+- **Google AI only (default)**: `make integration_tests`
+- **Both Google AI and Vertex AI**: `TEST_VERTEXAI=1 make integration_tests`
+- **Vertex AI only**: `TEST_VERTEXAI=only make integration_tests`
+
+Vertex AI tests require the `GOOGLE_CLOUD_PROJECT` environment variable to be set. Tests will automatically skip if not configured.
+
 #### Annotating integration tests
 
 We annotate integration tests to separate those tests which heavily rely on GCP infrastructure. Especially for running those tests we have created a separate GCP project with all necessary infrastructure parts provisioned. To run the extended integration tests locally you will need to provision a GCP project and pass its configuration via env variables.

libs/genai/Makefile

@@ -67,5 +67,8 @@ help:
 	@echo 'lint                         - run linters'
 	@echo 'test                         - run unit tests'
 	@echo 'tests                        - run unit tests'
-	@echo 'integration_test             - run integration tests(NOTE: "export GOOGLE_API_KEY=..." is needed.)'
+	@echo 'integration_test             - run integration tests (NOTE: "export GOOGLE_API_KEY=..." is needed.)'
+	@echo '                               Use TEST_VERTEXAI=1 to test both Google AI and Vertex AI backends'
+	@echo '                               Use TEST_VERTEXAI=only to test only Vertex AI backend'
+	@echo '                               Vertex AI requires GOOGLE_CLOUD_PROJECT to be set'
 	@echo 'test TEST_FILE=<test_file>   - run all tests in file'

libs/genai/langchain_google_genai/__init__.py

@@ -1,60 +1,38 @@
-"""LangChain Google Generative AI Integration (GenAI).
+"""LangChain Google Gen AI integration.
 
-This module integrates Google's Generative AI models, specifically the Gemini series,
-with the LangChain framework. It provides classes for interacting with chat models and
-generating embeddings, leveraging Google's advanced AI capabilities.
+!!! note "Vertex AI Consolidated"
+
+    As of `langchain-google-genai 4.0.0`, this package uses the
+    [consolidated Google Gen AI SDK](https://googleapis.github.io/python-genai/)
+    instead of the legacy [`google-ai-generativelanguage`](https://googleapis.dev/python/generativelanguage/latest/)
+    SDK.
+
+    This brings support for Gemini models both via the Gemini API and Gemini API in
+    Vertex AI, superseding `langchain-google-vertexai`. Users should migrate to this
+    package for continued support of Google's Generative AI models.
+
+    Certain Vertex AI features are not yet supported in the consolidated SDK (and
+    subsequently this package); see [the docs](https://docs.langchain.com/oss/python/integrations/providers/google)
+    for more details.
+
+This module provides an interface to Google's Generative AI models, specifically the
+Gemini series, with the LangChain framework. It provides classes for interacting with
+chat models, generating embeddings, and more.
 
 **Chat Models**
 
 The `ChatGoogleGenerativeAI` class is the primary interface for interacting with
 Google's Gemini chat models. It allows users to send and receive messages using a
 specified Gemini model, suitable for various conversational AI applications.
 
-**LLMs**
-
-The `GoogleGenerativeAI` class is the primary interface for interacting with Google's
-Gemini LLMs. It allows users to generate text using a specified Gemini model.
-
 **Embeddings**
 
 The `GoogleGenerativeAIEmbeddings` class provides functionalities to generate embeddings
 using Google's models. These embeddings can be used for a range of NLP tasks, including
 semantic analysis, similarity comparisons, and more.
 
-**Using Chat Models**
-
-After setting up your environment with the required API key, you can interact with the
-Google Gemini models.
-
-```python
-from langchain_google_genai import ChatGoogleGenerativeAI
-
-llm = ChatGoogleGenerativeAI(model="gemini-2.5-pro")
-llm.invoke("Sing a ballad of LangChain.")
-```
-
-**Using LLMs**
-
-The package also supports generating text with Google's models.
-
-```python
-from langchain_google_genai import GoogleGenerativeAI
-
-llm = GoogleGenerativeAI(model="gemini-2.5-pro")
-llm.invoke("Once upon a time, a library called LangChain")
-```
-
-**Embedding Generation**
-
-The package also supports creating embeddings with Google's models, useful for textual
-similarity and other NLP applications.
-
-```python
-from langchain_google_genai import GoogleGenerativeAIEmbeddings
-
-embeddings = GoogleGenerativeAIEmbeddings(model="models/gemini-embedding-001")
-embeddings.embed_query("hello, world!")
-```
+See [the docs](https://docs.langchain.com/oss/python/integrations/providers/google) for
+more information on usage of this package.
 """
 
 from langchain_google_genai._enums import (
@@ -70,6 +48,7 @@
     GoogleVectorStore,
 )
 from langchain_google_genai.llms import GoogleGenerativeAI
+from langchain_google_genai.utils import create_context_cache
 
 __all__ = [
     "ChatGoogleGenerativeAI",
@@ -81,4 +60,5 @@
     "HarmCategory",
     "MediaResolution",
     "Modality",
+    "create_context_cache",
 ]