OwlRoraRouting and Observability for Reliable AI

Route Once. Reach All.

Get started

Deployment guide

Implementation status

Protocol-native gateway

Preserve OpenAI, Anthropic, and Gemini request semantics while routing to compatible upstream deployments.

Policy-driven reliability

Apply target selection, failover, retry, circuit, health, stickiness, timeout, budget, rate, and concurrency policy.

Multi-tenant management

Keep deployment and organization authority explicit across users, JWT issuers, Management API keys, Gateway API keys, BYOK resources, and grants.

Self-hosted modular monolith

Run one Rust server with an embedded React console, PostgreSQL durable state, Redis coordination, and no hidden control-plane service.

Current delivery boundary

Source and release boundary

This site follows repository main and can describe capabilities newer than a selected binary. Before deployment, pin exact server and CLI releases, verify their source and release notes, and pin the server image by immutable digest. Do not interpret a release as completion of the target specification.

OwlRora's normative target architecture lives under spec/. The target specification is not yet complete as a delivered product. See Implementation status for the evidence-based boundary between released capability, newer source revisions, and remaining target work.

Start with the right path

Evaluate the current source tree: Getting started
Plan a production topology: Deployment
Configure every server setting: Configuration
Operate upgrades, backups, and recovery: Production operations
Understand trust and secret boundaries: Security model