Agent API Cookiecutter

Set the OPENAI_API_KEY environment variable to provide authentication for OpenAI services used by the agent API. This key is required for making requests to OpenAI models.

OPENAI_API_KEY=

The application uses pydantic-settings to manage configuration. Settings are loaded from environment variables or a .env file located in the project root. The Settings class is instantiated as a singleton named settings and can be imported throughout the package to access configuration values.

To customize settings, you can define a .env file in your project directory. The configuration is set to ignore extra environment variables that are not explicitly defined in the Settings class.

from {{cookiecutter.project_slug}}.config import settings

# Accessing a setting (assuming 'API_KEY' was defined in Settings)
api_key = settings.API_KEY

The docker-compose.yaml file defines the deployment configuration for the agent-api service. By default, it maps port 8000 on the host to port 8000 in the container and loads environment variables from a .env file located in the project root. The service is built from the local Dockerfile and is attached to a dedicated network named {{ cookiecutter.package_name }}-network.

services:
  agent-api:
    container_name: {{ cookiecutter.package_name }}
    build:
      context: .
      dockerfile: Dockerfile
    ports:
      - "8000:8000"
    env_file:
      - .env
    networks:
      - {{ cookiecutter.package_name }}-network

volumes:
  data:

networks:
  {{ cookiecutter.package_name }}-network:
    name: {{ cookiecutter.package_name }}-network

The API can be started directly using Python if the file is executed as a script. It uses uvicorn to serve the FastAPI application on 0.0.0.0:8000 by default.

python src/{{cookiecutter.project_slug}}/infrastructure/api/main.py

The Agent API provides several POST endpoints to interact with the agent's core services: chat, evaluation, document ingestion, and memory management. The API is built with FastAPI and includes an interactive documentation page at /docs.

# Example endpoint interactions (conceptual):

# Chat with the agent
POST /chat
Payload: ChatRequest

# Evaluate the agent
POST /eval
Payload: EvalRequest

# Ingest documents into the agent's knowledge base
POST /eval
Payload: IngestDocumentsRequest

# Reset the agent's memory
POST /reset-memory
Payload: ResetMemoryRequest