pydantic-ai-langfuse

pydantic-ai-langfuse extends pydantic-ai-slim to integrate Langfuse tracking into your OpenAI model interactions. By incorporating our Langfuse OpenAI model settings, you can easily label, track, and filter your generations using enriched metadata.

Installation

Ensure you have the required dependencies installed:

pip install pydantic-ai-langfuse

Environment Setup

Before running your model, you need to set up the following environment variables:

• OPENAI_API_KEY: Your OpenAI API key.
• LANGFUSE_PUBLIC_KEY: Your Langfuse public key.
• LANGFUSE_SECRET_KEY: Your Langfuse secret key.
• LANGFUSE_HOST: Your Langfuse host endpoint.

Quickstart

Below is a complete Python example showing how to set up and use the LangfuseOpenAIModel with extra model settings. This example uses the synchronous run_sync method with basic error handling and retries built in.

import os

from langfuse.openai import AsyncOpenAI
from pydantic_ai import Agent

from pydantic_ai_langfuse import LangfuseOpenAIModel

for var in [
    "OPENAI_API_KEY",
    "LANGFUSE_PUBLIC_KEY",
    "LANGFUSE_SECRET_KEY",
    "LANGFUSE_HOST",
]:
    if var not in os.environ:
        raise OSError(f"Missing env variable: {var}")

weather_agent = Agent(
    model=LangfuseOpenAIModel("gpt-4o", openai_client=AsyncOpenAI()),
    system_prompt="Be concise: reply with one sentence.",
    retries=2,
)

result = weather_agent.run_sync(
    "What the weather like in Medolago BG?",
    model_settings={
        "name": "weather_query",
        "metadata": {"location": "medolago", "query_type": "weather"},
        "session_id": "testoneditest",
        "user_id": "user123",
        "tags": ["weather", "italy", "query"],
    },
)

print("Response:", result.data)