wtf is tokhub?

TokHub is a monitoring and gateway system for people who run or use AI API relay services, the middlemen that sit between you and providers like OpenAI, Anthropic, or Gemini and forward your requests to them. It combines a public status page, a ranked list of providers, a workspace for individual users, an admin backend, and a gateway that speaks the same language as OpenAI's API, all in one open source project you can run on your own server with Docker. The problem it aims to fix is that most status pages only say a service is "up" or "down" without saying whether the failure is in DNS, a security certificate, login credentials, the list of available models, or the actual reply generation. TokHub checks health in three layers instead. The first layer tests basic network reachability: resolving the address, connecting, and checking the security handshake. The second layer calls the provider's model list endpoint to confirm the login key works and the requested model actually exists. The third layer sends a minimal real chat request and checks that a proper reply comes back, catching cases where a service looks reachable but is not actually generating anything useful. These three checks are combined into an overall status such as healthy, degraded, or down, along with a health score. Users can save public channels they like or add their own private ones, each with its own address, model, and quota. Everyone gets a personal OpenAI-compatible gateway that can route requests across several upstream providers, choosing the best one by lowest latency, highest success rate, or lowest cost, and automatically skipping ones that are currently failing. Security keys are encrypted, gateway keys are only shown once at creation, and the system blocks requests aimed at internal or local network addresses to reduce abuse. The backend is written in Go with PostgreSQL and Redis, the frontend in React and TypeScript, and the whole project is released under the Apache 2.0 license.