openai

Core Concept

Integration guide for using OpenAI's API in Python projects, particularly with LangChain and LangGraph.

Critical Reference Files

IMPORTANT: Before working with OpenAI API, read:

• Core Rules - Environment setup and ChatOpenAI usage

Environment Setup

Required Variables

Always assume these environment variables are set:

• OPENAI_BASE_URL - API endpoint
• OPENAI_API_KEY - Authentication key

Docker Integration

When using Docker Compose, propagate from host environment:

services:
  myapp:
    environment:
      - OPENAI_BASE_URL
      - OPENAI_API_KEY

ChatOpenAI Usage

Type Checking Workaround

The model parameter raises type checking warnings. Use type ignore comment:

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="gpt-4",  # type: ignore[unknown-argument]
    temperature=0.7,
)

Integration with LangGraph

When using OpenAI with LangGraph:

1. Read ../langgraph/SKILL.md for LangGraph patterns
2. Use structured output with Pydantic (see ../langgraph/references/core-rules.md)
3. Always use async calls with ainvoke()

Example:

from langchain_openai import ChatOpenAI
from langchain_core.output_parsers import PydanticOutputParser

llm = ChatOpenAI(
    model="gpt-4",  # type: ignore[unknown-argument]
)

parser = PydanticOutputParser(pydantic_object=ResponseModel)
result = await parser.ainvoke(await (prompt | llm).ainvoke(inputs))

Best Practices

1. Never hardcode credentials - Always use environment variables
2. Use type ignore for model parameter - Known LangChain issue
3. Prefer async calls - Use ainvoke() over invoke()
4. Structured outputs - Always parse with Pydantic in LangGraph
5. Error handling - Wrap API calls in try/except for rate limits

Integration Points

With Python

• Follow ../python/SKILL.md for general practices
• See ../python/references/secrets.md for handling API keys

With LangGraph

• See ../langgraph/SKILL.md for graph integration
• Use structured output patterns from ../langgraph/references/core-rules.md

With Docker

• Propagate environment variables in docker-compose
• See ../docker/SKILL.md for container setup