This project wraps a pydantic-ai agent as an OpenAI API-compatible service, allowing you to use your pydantic-ai agents with existing OpenAI client libraries and tools.

The service reimplements the OpenAI API interface standards to expose a pydantic-ai agent as if it were a large language model service. This enables seamless integration with existing applications that expect OpenAI's API format.

• OpenAI API-compatible endpoints
• Pydantic-ai agent integration
• RESTful API interface
• Drop-in replacement for OpenAI API clients
• API key authentication support
• Session-based request tracking
• Streaming and non-streaming responses

Clients can interact with this service using the same interface as OpenAI's API, making it compatible with existing OpenAI client libraries and tools.

• GET / - Health check
• GET /v1/models - List available models (returns single "pydantic-ai-agent" model)
• POST /v1/chat/completions - Create chat completions (OpenAI-compatible)

The service supports optional API key authentication. Set the API_KEY environment variable to enable authentication:

export API_KEY="your-secret-key"

When authentication is enabled, include the API key in the Authorization header:

curl -H "Authorization: Bearer your-secret-key" http://localhost:8000/v1/models

Environment variables:

• API_KEY: Optional API key for authentication
• AGENTO_SALT: Optional salt for session ID generation (defaults to "abc")

• pyproject.toml - Project configuration and dependencies
• src/main.py - Demo Application entry point wrapper