ChatGroq 集成

使用 LangChain Python 集成 ChatGroq 聊天模型。

This page makes reference to Groq, an AI hardware and software company. For information on how to use Grok models (provided by xAI), 请参阅 xAI provider page.

API 参考

有关所有 features and configuration options, 请前往 ChatGroq API reference.

For a list of all Groq models, visit their docs.

概述

集成详情

类	包	可序列化	JS 支持	下载量	版本
ChatGroq	langchain-groq	beta	✅

模型功能

Tool calling	Structured output	Image input	音频输入	视频输入	Token-level streaming	原生异步	Token usage	Logprobs
✅	✅	✅	❌	❌	✅	✅	✅	✅

设置

要访问 Groq 模型，您需要创建一个 Groq 账户，获取 API 密钥，并安装 langchain-groq 集成包。

凭证

Head to the Groq console 注册 Groq 并生成 API 密钥。 完成后设置 GROQ_API_KEY 环境变量：

import getpass
import os

if "GROQ_API_KEY" not in os.environ:
    os.environ["GROQ_API_KEY"] = getpass.getpass("Enter your Groq API key: ")

要启用模型调用的自动追踪，请设置您的 LangSmith API key:

os.environ["LANGSMITH_API_KEY"] = getpass.getpass("请输入您的 LangSmith API 密钥: ")
os.environ["LANGSMITH_TRACING"] = "true"

安装

LangChain 的 Groq 集成位于 langchain-groq 包中：

pip install -qU langchain-groq

实例化

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

Reasoning Format

If you choose to set a reasoning_format, you must ensure that the model you are using supports it. You can find a list of supported models in the Groq documentation.

from langchain_groq import ChatGroq

llm = ChatGroq(
    model="qwen/qwen3-32b",
    temperature=0,
    max_tokens=None,
    reasoning_format="parsed",
    timeout=None,
    max_retries=2,
    # 其他参数...
)

调用

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'aime la programmation.", additional_kwargs={'reasoning_content': 'Okay, so I need to translate the sentence "I love programming." into French. Let me think about how to approach this. \n\nFirst, I know that "I" in French is "Je." That\'s straightforward. Now, the verb "love" in French is "aime" when referring to oneself. So, "I love" would be "J\'aime." \n\nNext, the word "programming." In French, programming is "la programmation." But wait, in French, when you talk about loving an activity, you often use the definite article. So, it would be "la programmation." \n\nPutting it all together, "I love programming" becomes "J\'aime la programmation." That sounds right. I think that\'s the correct translation. \n\nI should double-check to make sure I\'m not missing anything. Maybe I can think of similar phrases. For example, "I love reading" is "J\'aime lire," but when it\'s a noun, like "I love music," it\'s "J\'aime la musique." So, yes, using "la programmation" makes sense here. \n\nI don\'t think I need to change anything else. The sentence structure in French is Subject-Verb-Object, just like in English, so "J\'aime la programmation" should be correct. \n\nI guess another way to say it could be "J\'adore la programmation," using "adore" instead of "aime," but "aime" is more commonly used in this context. So, sticking with "J\'aime la programmation" is probably the best choice.\n'}, response_metadata={'token_usage': {'completion_tokens': 346, 'prompt_tokens': 23, 'total_tokens': 369, 'completion_time': 1.447541218, 'prompt_time': 0.000983386, 'queue_time': 0.009673684, 'total_time': 1.448524604}, 'model_name': 'deepseek-r1-distill-llama-70b', 'system_fingerprint': 'fp_e98d30d035', 'finish_reason': 'stop', 'logprobs': None}, id='run--5679ae4f-f4e8-4931-bcd5-7304223832c0-0', usage_metadata={'input_tokens': 23, 'output_tokens': 346, 'total_tokens': 369})

print(ai_msg.content)

J'aime la programmation.

Vision

Groq supports vision capabilities with select models, allowing you to send images along with text prompts.

Vision-capable models

• meta-llama/llama-4-scout-17b-16e-instruct
• meta-llama/llama-4-maverick-17b-128e-instruct

For the latest list of vision-capable models, check the Groq documentation.

from langchain_groq import ChatGroq
from langchain.messages import HumanMessage

llm = ChatGroq(model="meta-llama/llama-4-scout-17b-16e-instruct")

message = HumanMessage(
    content=[
        {"type": "text", "text": "Describe this image in detail."},
        {
            "type": "image_url",
            "image_url": {"url": "https://example.com/image.jpg"},
        },
    ]
)

response = llm.invoke([message])
print(response.content)

Image URL requirements

Groq fetches images from URLs directly. Ensure your image URLs:

• Are publicly accessible (no authentication required)
• Return the image directly (no redirects)
• Maximum image size: 20MB per request

---

API 参考

有关所有 ChatGroq 功能和配置的详细文档，请前往 API reference.

---

Connect these docs to Claude, VSCode, and more via MCP for real-time answers.

Edit this page on GitHub or file an issue.