CausewayAI is a backend system designed for causal analysis and retrieval-augmented generation (RAG) using FastAPI, Qdrant, and NetworkX. It leverages Google's Gemini models and embeddinggemma for advanced query understanding and response generation.

• Graph Strategy: Uses a knowledge graph to traverse causal logic and retrieve relevant calls.
• Filter Strategy: Uses metadata filtering to find relevant conversations based on domain, topic, and outcome.
• Streaming Responses: Returns Server-Sent Events (SSE) for real-time interaction.

• Python 3.10+
• Docker & Docker Compose (optional but recommended)
• Google API Key (for Gemini)
• Hugging Face Token (for embeddinggemma)

Create a .env file in the root directory:

GOOGLE_API_KEY=your_google_api_key
HF_TOKEN=your_hugging_face_token

1. Install dependencies:

   pip install -r requirements.txt

2. Run the server:

   uvicorn main:app --reload

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

1. Build and run the container:

   docker-compose up --build

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

• GET /health: Health check.
• POST /chat: Main chat endpoint.
  • Payload:

    {
      "query": "Why are customers churning?",
      "history": [],
      "model_type": "graph"  // or "filter"
    }

  • Response: Streaming SSE events (status, concepts, sources, token, error).

Use the provided client script to test the engine:

python client_test.py "Why are customers churning?" --model graph