ChatGoogleGenerativeAI

This docs will help you get started with Google AI chat models. For detailed documentation of all ChatGoogleGenerativeAI features and configurations head to the API reference.

Google AI offers a number of different chat models. For information on the latest models, their features, context windows, etc. head to the Google AI docs.

Google AI vs Google Cloud Vertex AI

Google's Gemini models are accessible through Google AI and through Google Cloud Vertex AI. Using Google AI just requires a Google account and an API key. Using Google Cloud Vertex AI requires a Google Cloud account (with term agreements and billing) but offers enterprise features like customer encription key, virtual private cloud, and more.

To learn more about the key features of the two APIs see the Google docs.

Overview

Integration details

Class	Package	Local	Serializable	JS support	Package downloads	Package latest
ChatGoogleGenerativeAI	langchain-google-genai	❌	beta	✅

Model features

Tool calling	Structured output	JSON mode	Image input	Audio input	Video input	Token-level streaming	Native async	Token usage	Logprobs
✅	✅	❌	✅	✅	✅	✅	✅	✅	❌

Setup

To access Google AI models you'll need to create a Google Acount account, get a Google AI API key, and install the langchain-google-genai integration package.

Credentials

Head to https://ai.google.dev/gemini-api/docs/api-key to generate a Google AI API key. Once you've done this set the GOOGLE_API_KEY environment variable:

import getpass
import os

if "GOOGLE_API_KEY" not in os.environ:
    os.environ["GOOGLE_API_KEY"] = getpass.getpass("Enter your Google AI API key: ")

If you want to get automated tracing of your model calls you can also set your LangSmith API key by uncommenting below:

# os.environ["LANGSMITH_API_KEY"] = getpass.getpass("Enter your LangSmith API key: ")
# os.environ["LANGSMITH_TRACING"] = "true"

Installation

The LangChain Google AI integration lives in the langchain-google-genai package:

%pip install -qU langchain-google-genai

Instantiation

Now we can instantiate our model object and generate chat completions:

from langchain_google_genai import ChatGoogleGenerativeAI

llm = ChatGoogleGenerativeAI(
    model="gemini-1.5-pro",
    temperature=0,
    max_tokens=None,
    timeout=None,
    max_retries=2,
    # other params...
)

API Reference:ChatGoogleGenerativeAI

Invocation

messages = [
    (
        "system",
        "You are a helpful assistant that translates English to French. Translate the user sentence.",
    ),
    ("human", "I love programming."),
]
ai_msg = llm.invoke(messages)
ai_msg

AIMessage(content="J'adore programmer. \n", response_metadata={'prompt_feedback': {'block_reason': 0, 'safety_ratings': []}, 'finish_reason': 'STOP', 'safety_ratings': [{'category': 'HARM_CATEGORY_SEXUALLY_EXPLICIT', 'probability': 'NEGLIGIBLE', 'blocked': False}, {'category': 'HARM_CATEGORY_HATE_SPEECH', 'probability': 'NEGLIGIBLE', 'blocked': False}, {'category': 'HARM_CATEGORY_HARASSMENT', 'probability': 'NEGLIGIBLE', 'blocked': False}, {'category': 'HARM_CATEGORY_DANGEROUS_CONTENT', 'probability': 'NEGLIGIBLE', 'blocked': False}]}, id='run-eef5b138-1da6-4226-9cfe-ab9073ddd77e-0', usage_metadata={'input_tokens': 21, 'output_tokens': 5, 'total_tokens': 26})

print(ai_msg.content)

J'adore programmer.

Chaining

We can chain our model with a prompt template like so:

from langchain_core.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate.from_messages(
    [
        (
            "system",
            "You are a helpful assistant that translates {input_language} to {output_language}.",
        ),
        ("human", "{input}"),
    ]
)

chain = prompt | llm
chain.invoke(
    {
        "input_language": "English",
        "output_language": "German",
        "input": "I love programming.",
    }
)

API Reference:ChatPromptTemplate

AIMessage(content='Ich liebe das Programmieren. \n', response_metadata={'prompt_feedback': {'block_reason': 0, 'safety_ratings': []}, 'finish_reason': 'STOP', 'safety_ratings': [{'category': 'HARM_CATEGORY_SEXUALLY_EXPLICIT', 'probability': 'NEGLIGIBLE', 'blocked': False}, {'category': 'HARM_CATEGORY_HATE_SPEECH', 'probability': 'NEGLIGIBLE', 'blocked': False}, {'category': 'HARM_CATEGORY_HARASSMENT', 'probability': 'NEGLIGIBLE', 'blocked': False}, {'category': 'HARM_CATEGORY_DANGEROUS_CONTENT', 'probability': 'NEGLIGIBLE', 'blocked': False}]}, id='run-fbb35f30-4937-4a81-ae68-f7cb35721a0c-0', usage_metadata={'input_tokens': 16, 'output_tokens': 7, 'total_tokens': 23})

Safety Settings

Gemini models have default safety settings that can be overridden. If you are receiving lots of "Safety Warnings" from your models, you can try tweaking the safety_settings attribute of the model. For example, to turn off safety blocking for dangerous content, you can construct your LLM as follows:

from langchain_google_genai import (
    ChatGoogleGenerativeAI,
    HarmBlockThreshold,
    HarmCategory,
)

llm = ChatGoogleGenerativeAI(
    model="gemini-1.5-pro",
    safety_settings={
        HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_NONE,
    },
)

API Reference:ChatGoogleGenerativeAI | HarmBlockThreshold | HarmCategory

For an enumeration of the categories and thresholds available, see Google's safety setting types.

API reference

For detailed documentation of all ChatGoogleGenerativeAI features and configurations head to the API reference: https://python.langchain.com/api_reference/google_genai/chat_models/langchain_google_genai.chat_models.ChatGoogleGenerativeAI.html

Related

• Chat model conceptual guide
• Chat model how-to guides