OmniRoutePhenotype AI Gateway Fork

OpenAI-compatible routing, retries, load balancing, fallbacks — with governance-grade observability.

Smart Routing

Provider failover, weighted load balancing, quota-aware routing. Drops to a fast secondary when the primary hits rate limits.

First-class plugin model — declare a provider once and reuse across OpenAI-compatible clients without code changes.

Hash-chained audit, governance checks, threat model. Built for fork governance at scale.

Per-request cost rollups with attribution to model + tier. Persistent in SQLite; exportable to Tracera.

Single binary, SQLite by default, no external dependencies. Runs on macOS, Linux, Raspberry Pi.

Provider-error-classification with exponential backoff. Quota recovery follow-up hook for retry-and-quota fault tolerance.