🧱 Tinkerflow‑AI — App Authority Wall

🛡️ GitDigital Badge Authority — Governing Authority Wall

🔖 Authority Badges

🧱 Tinkerflow‑AI — App Authority

🛡️ Governing Authority — GitDigital Badge Authority

Tinkerflow-AI

Tinkerflow AI Gateway

A unified, secure, and scalable proxy for multiple AI providers

Tinkerflow AI Gateway sits between your applications and AI services like OpenAI and Ollama, providing a single API endpoint with built-in authentication, rate limiting, metrics, and flexible model routing. It's designed for developers who want to manage multiple AI backends without changing their code.

🚀 Why Tinkerflow?

· Unified API – Use the same OpenAI-compatible interface for all providers.
· Production‑ready – Redis‑backed rate limiting, Prometheus metrics, and comprehensive tests.
· Extensible – Easily add new providers (Anthropic, Cohere, etc.) via simple modules.
· Self‑hosted – Full control over your data and costs.

✨ Features

Feature Description
🔑 API Key Authentication Simple shared‑secret security with header or query parameter support.
⏱️ Rate Limiting Sliding window rate limiting per key (in‑memory for dev, Redis for production).
🤖 Multi‑Provider Route requests to OpenAI, Ollama, or custom backends.
🧠 Model Mapping YAML configuration to map model names to providers (e.g., gpt‑4 → openai).
🔄 Streaming Full support for Server‑Sent Events (SSE) – compatible with OpenAI's streaming API.
📊 Metrics Prometheus metrics for requests, latencies, and errors.
🐍 Python Client Drop‑in replacement for the OpenAI Python library.
🧪 Tested 90%+ test coverage with GitHub Actions CI.

📦 Quick Start (2 minutes)

# Clone and enter
git clone https://github.com/RickCreator87/Tinkerflow-AI.git
cd Tinkerflow-AI

# Install dependencies
pip install -r requirements.txt

# Configure (copy and edit .env)
cp .env.example .env
# Edit with your OpenAI API key and gateway secret

# Run Redis (if not already)
docker run -d -p 6379:6379 redis

# Start the gateway
uvicorn gateway.main:app --reload

The gateway is now live at http://localhost:8000. Test it:

curl -X POST http://localhost:8000/v1/chat/completions \
  -H "Authorization: Bearer your-gateway-api-key" \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-3.5-turbo", "messages": [{"role": "user", "content": "Hello!"}]}'

🏗️ Architecture Overview

┌─────────┐      ┌───────────────────────┐      ┌──────────────┐
│ Client  │ ──►  │ Tinkerflow Gateway    │ ──►  │ OpenAI       │
│ (your   │      │ - Auth                 │      ├──────────────┤
│  app)   │      │ - Rate limiting        │ ──►  │ Ollama       │
└─────────┘      │ - Model routing        │      ├──────────────┤
                 │ - Metrics               │ ──►  │ (other)      │
                 └───────────────────────┘      └──────────────┘
                              │
                              ▼
                         ┌──────────┐
                         │ Redis    │
                         │ (rate    │
                         │  limits) │
                         └──────────┘

· All requests hit the gateway first.
· The gateway authenticates, rate‑limits, and routes based on model field.
· Responses (including streaming) are proxied transparently.
· Metrics are exposed for Prometheus.

📚 Documentation

All documentation is in the repository:

· Installation & Setup
· Configuration Reference
· Usage Examples
· Model Management
· Security Notes
· Troubleshooting
· Roadmap

🧩 Use Cases

· Centralized AI access for multiple microservices.
· Cost control – enforce rate limits per team or project.
· Local development – use Ollama locally, switch to OpenAI in production.
· Vendor‑agnostic – change providers without rewriting code.

🤝 Contributing

Contributions are welcome! Check the Roadmap for ideas or open an issue.

📄 License

MIT © RickCreator87

Built with ❤️ using FastAPI, Redis, and a lot of coffee.