Sensei v5 Neural Network Server

This is the standalone Python server for the v5 Neural Network AI.

Setup Instructions

1. Create New Replit Project

• Create a new Python Replit (use the Python template)
• Copy ALL files from this folder to your new Replit

2. Configure Secrets

In your new Replit, add the following secret:

• SENSEI_SECRET - Use the same value as your main Board Game Dojo project

3. Install Dependencies

The dependencies will install automatically when you run the project. If needed manually: pip install flask onnxruntime numpy

4. Run the Server

Click "Run" - the server will start on port 5001.

5. Deploy

• Go to the "Deployments" tab
• Choose "Reserved VM" or "Autoscale"
• Deploy the project

6. Custom Domain Setup

To use sensei.boardgamedojo.com:

1. In Replit Deployments, click "Add Custom Domain"
2. Enter: sensei.boardgamedojo.com
3. Copy the CNAME value Replit provides
4. In your DNS provider (for boardgamedojo.com), add a CNAME record:
   • Name: sensei
   • Value: (the value from step 3)
5. Wait for DNS propagation (can take up to 24 hours)

Endpoints

• GET /health - Health check (returns status and model info)
• POST /evaluate - Evaluate a game state (requires HMAC authentication)

Environment Variables

• SENSEI_SECRET (required) - Shared secret for HMAC authentication
• PORT (optional) - Server port, defaults to 5001
• MODEL_VARIANT (optional) - "int8" or "float", defaults to "int8"

Files

• server.py - Main Flask server
• sensei_v5_int8.onnx - Quantized model (smaller, faster)
• sensei_v5_float.onnx - Full precision model (backup)
• requirements.txt - Python dependencies