duke-ai-kernel

AI kernel for the duke-messenger platform. Handles LLM routing, context management, and real-time messaging via WebSocket.

Quick Start

# 1. Clone and install
git clone https://github.com/duke-messenger/duke-ai-kernel
cd duke-ai-kernel
npm install

# 2. Configure environment
cp .env.example .env
# Edit .env with your API keys

# 3. Start the kernel
npm run dev

The kernel starts on http://localhost:3000 by default.

Features

• Unified LLM routing — route requests to OpenAI, Anthropic, or local Ollama models
• Context management — per-conversation history with configurable token-budget trimming
• Real-time streaming — stream responses to clients over WebSocket
• Plugin pipeline — extend with custom pre/post processors
• Worker concurrency — bounded async workers prevent rate-limit saturation

Stack

| Layer | Technology | |-------|-----------| | Runtime | Node.js 20+ | | Framework | Express 5 | | WebSocket | ws | | Language | TypeScript | | Testing | Vitest |

Scripts

npm run dev          # Start with hot reload
npm run build        # Compile TypeScript
npm start            # Run compiled output
npm run lint         # Check code style
npm run typecheck    # Type-check without building
npm test             # Run all tests
npm run test:coverage # Run tests with coverage report

Environment Variables

Copy .env.example to .env. At minimum you need:

OPENAI_API_KEY=sk-...
AUTH_SECRET=your-secret-here

See Environment Variables for the full reference.

Docker

# Start kernel + Redis
docker compose up -d

# Verify
curl http://localhost:3000/api/v1/healthz

Documentation

• API Reference
• Architecture & Data Flow

License

MIT © duke-messenger contributors