快速开始

LangSmith 是一个用于构建生产级 LLM 应用程序的平台。它允许您密切监控和评估您的应用程序的性能和成本，并对其进行优化。

兼容说明

如果您使用 Python 开发自己的 AI 应用，可以继续阅读本文。
如果您使用的是 LangChain 或 LangGraph，它们都与 LangSmith 无缝集成，则可以阅读使用 LangChain 进行跟踪或使用 LangGraph 进行跟踪的指南。
Lighthouse 完全兼容 LangSmith SDK，您可以使用 LangSmith Python SDK 来采集 LLM 应用的可观测性数据。

安装依赖项

前置条件

确保您的环境满足以下要求：

Python 3.7 或更高版本
已创建 Lighthouse LLM 应用
已获取应用的密钥和公钥

pip install -U langsmith openai

创建 API 密钥

要创建 API 密钥，请前往 Lighthouse 「LLM 应用/应用分析/新建 LLM 应用」页面，然后单击「新建 API Key」（创建 API 密钥）。

设置您的环境

export LANGSMITH_TRACING=true
export LANGSMITH_ENDPOINT="<your-lighthouse-endpoint>"
export LANGSMITH_API_KEY="<your-lighthouse-api-key>"
export LANGSMITH_PROJECT="<your-lighthouse-project>"

export OPENAI_API_URL="<your-openai-api>"
export OPENAI_API_KEY="<your-openai-api-key>"

配置建议

推荐使用环境变量方式进行配置，这样可以更好地管理敏感信息。

定义您的应用程序

以下是一个简单的 RAG 应用示例，您也可以替换为自己的代码：

from openai import OpenAI

openai_client = OpenAI()

# This is the retriever we will use in RAG
# This is mocked out, but it could be anything we want
def retriever(query: str):
    results = ["Harrison worked at Kensho"]
    return results

# This is the end-to-end RAG chain.
# It does a retrieval step then calls OpenAI
def rag(question):
    docs = retriever(question)
    system_message = """Answer the users question using only the provided information below:
    
    {docs}""".format(docs="\n".join(docs))
    
    return openai_client.chat.completions.create(
        messages=[
            {"role": "system", "content": system_message},
            {"role": "user", "content": question},
        ],
        model="gpt-4o-mini",
    )

跟踪 OpenAI 调用

from openai import OpenAI
from langsmith.wrappers import wrap_openai
import os

openai_client = wrap_openai(OpenAI(base_url=os.environ["OPENAI_API_URL"],api_key=os.environ["OPENAI_API_KEY"]))

# This is the retriever we will use in RAG
# This is mocked out, but it could be anything we want
def retriever(query: str):
    results = ["Harrison worked at Kensho"]
    return results

# This is the end-to-end RAG chain.
# It does a retrieval step then calls OpenAI
def rag(question):
    docs = retriever(question)
    system_message = """Answer the users question using only the provided information below:
    
    {docs}""".format(docs="\n".join(docs))
    
    return openai_client.chat.completions.create(
        messages=[
            {"role": "system", "content": system_message},
            {"role": "user", "content": question},
        ],
        model="glm-4-flash",
    )

现在，当您按如下方式调用应用程序时：

rag("where did harrison work")

将在 Lighthouse 新建的项目中生成 OpenAI 调用的跟踪。它应该看起来像这样： alt text

跟踪整个应用程序

from openai import OpenAI
from langsmith import traceable
from langsmith.wrappers import wrap_openai
import os

openai_client = wrap_openai(OpenAI(base_url=os.environ["BASE_URL"],api_key=os.environ["OPENAI_API_KEY"]))

# This is the retriever we will use in RAG
# This is mocked out, but it could be anything we want
def retriever(query: str):
    results = ["Harrison worked at Kensho"]
    return results

# This is the end-to-end RAG chain.
# It does a retrieval step then calls OpenAI
@traceable
def rag(question):
    docs = retriever(question)
    system_message = """Answer the users question using only the provided information below:
    
    {docs}""".format(docs="\n".join(docs))
    
    return openai_client.chat.completions.create(
        messages=[
            {"role": "system", "content": system_message},
            {"role": "user", "content": question},
        ],
        model="glm-4-flash",
    )

result = rag("where did harrison work in 2025 ")
print(result.choices[0].message.content)

这将生成整个管道的跟踪（将 OpenAI 调用作为子运行）- 它应该看起来像这样 alt text

框架支持

支持的框架

Lighthouse 通过 LangSmith SDK 支持以下主流框架：

OpenAI
LangChain
LangGraph