← Back to the board
Cerberus
Rate limiting as a service, atomic down to the Redis script.
The problem
Rate limiters that check then update in two steps break under load: two requests both see room, and both get through.
50K+ checks/sec
What it is designed for, at sub-1.5ms p99.
Race-free
Each decision happens atomically inside one Redis script.
4 packages
SDKs and middleware for adding it to an existing service.
Under the hood
- Sliding-window and token-bucket limiters in Redis Lua behind a 3-tier cache.
- Designed for 50K+ checks/sec at sub-1.5ms p99.
- 4 SDK and middleware packages.
Stack
← PreviousPythonFastAPIRedis LuaPostgresPrometheus
DOOMA
A public AI courtroom. Stake it, argue it, let the crowd pick a side, get paid out by escrow.
Next →Capline
Spend limits for AI agents that a jailbreak can't talk its way past.