#Project introduction

Geekit is an AI API gateway: it aggregates several model providers behind one unified API and adds tokens, quotas, rate limiting, logging, and an admin dashboard.

It is derived from One API (MIT) and is released under AGPLv3.

#The problem it solves

An organization using several model providers at once runs into the same set of chores:

• One SDK, one key, and one billing model per provider
• Keys handed directly to users cannot be taken back, and spend cannot be attributed
• A provider outage means someone switches things over by hand
• Cost allocation by team or project is impossible from a single invoice

The gateway collapses this into one place: users only ever hold a gateway-issued token, the real provider keys stay inside the gateway, usage is recorded per token, model, and channel, and a failed upstream is retried on another channel.

#Scope of use

This project targets lawfully authorized AI API gateway use, internal organizational authentication, multi-model management, usage statistics, cost accounting, and self-hosted deployment. Operators must obtain upstream API keys, accounts, and interface permissions lawfully, and comply with upstream terms of service and applicable law.

When offering generative AI services to the public, operators must meet the regulatory obligations of their own jurisdiction — filing, licensing, content safety, real-name verification, log retention, tax, and upstream authorization.

#Technology

The architecture is layered — router → controller → service → model — with upstream adapters concentrated in the relay layer, so adding a provider means adding one adapter.

#Next

• Features — what it actually does
• Getting started — get it running first