quota-rate-limit-pattern

What problem does it solve?

This Skill solves the problem of incorrectly throttling API traffic by handling rate limits (burst), concurrent in-flight requests, and daily token quotas with distinct mechanisms so you can reliably prevent overload and return actionable 429 errors.

Core Features & Use Cases

• 3-layer rate limiting: Combines slowapi RPM limits, asyncio Semaphore concurrency admission, and DB-backed daily token checks to cover different failure modes.
• Reason-code 429 responses: Returns HTTP 429 with a structured error payload and a specific code for rpm_limit, concurrent_limit, or daily_token_limit so clients can retry appropriately.
• Operational visibility: Integrates a Prometheus counter that labels rejections by reason, enabling fast debugging of which layer is blocking traffic.
• Distributed-environment guidance: Documents limits of worker-local enforcement and the required Redis/admission-controller alternatives when moving beyond single-worker deployments.

Quick Start

Ask your AI to generate a gateway quota module that enforces slowapi RPM, per-user asyncio concurrency slots, and a DB daily token SUM check, returning OpenAI-compatible 429 error responses with code set to the exact rejected layer.