Not So Fast AI
BuildingWhat it is
Cost-aware AI router: every task goes to the cheapest model that can do it well — local first, flagship only when needed — plus a market for unused credits.
The problem
Sending every AI task to the most expensive model wastes money, and one provider failing stops the work.
How it works
- 1Each task is classified by difficulty
- 2It is routed to the cheapest tier that can do it well: local, cheap cloud, high tier, flagship
- 3If a provider fails, the next one takes over automatically
- 4Cost and savings are tracked per task
Stack
PythonFastAPIReactViteLLM APIs
Private web app
The live control panel runs on my own PC, behind a login, and is not public. The public part is this page.
Status + What's next
Building
A market where users can list unused credits and others buy them below list price, with a wallet and ledger.