Skip to content

Latest commit

 

History

History
118 lines (90 loc) · 2.57 KB

README.md

File metadata and controls

118 lines (90 loc) · 2.57 KB

Conversation Optimizer

An AI-powered conversation optimization service that uses Monte Carlo Tree Search (MCTS) and Large Language Models to suggest optimal conversational moves and evaluate conversation quality.

Features

  • Conversation optimization using MCTS
  • Multi-dimensional conversation evaluation
  • FastAPI-based REST API
  • Caching and rate limiting for LLM calls
  • Async support for improved performance

Setup

  1. Clone the repository:
git clone https://github.com/yourusername/conversation-optimizer.git
cd conversation-optimizer
  1. Install uv (if not already installed):
pip install uv
  1. Create a virtual environment and install dependencies:
uv venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
uv pip install -e .
  1. Create a .env file in the project root with your OpenAI API key:
OPENAI_API_KEY=your_api_key_here
LLM_MODEL=gpt-3.5-turbo  # Optional: Change the model

Usage

  1. Start the API server:
uvicorn src.api.main:app --reload

  1. The API will be available at http://localhost:8000

  2. Access the API documentation at http://localhost:8000/docs

API Endpoints

POST /api/v1/optimize

Optimize a conversation by suggesting the best next moves.

Request:

{
    "conversation": [
        {"speaker": "you", "text": "Hi, I wanted to discuss our project strategy."},
        {"speaker": "opponent", "text": "I'm concerned about the risks involved."}
    ],
    "goal": "Discuss project strategy while addressing risks"
}

POST /api/v1/evaluate

Evaluate a conversation based on multiple dimensions.

Request:

{
    "conversation": [
        {"speaker": "you", "text": "Hi, I wanted to discuss our project strategy."},
        {"speaker": "opponent", "text": "I'm concerned about the risks involved."}
    ],
    "goal": "Discuss project strategy while addressing risks"
}

Development

  1. Install development dependencies:
uv pip install -e ".[dev]"
  1. Run tests:
pytest
  1. Format code:
black src/
isort src/
  1. Run linting:
ruff check src/
mypy src/

Configuration

The application can be configured through environment variables or a .env file:

  • OPENAI_API_KEY: Your OpenAI API key (required)
  • LLM_MODEL: The LLM model to use (default: gpt-3.5-turbo)
  • SIMULATION_DEPTH: MCTS simulation depth (default: 4)
  • SIMULATION_BREADTH: MCTS simulation breadth (default: 5)
  • MCTS_ITERATIONS: Number of MCTS iterations (default: 20)

See src/api/core/config.py for all available settings.

License

MIT License