v0.1.0

First public release of OrcaRouter Lite — the open-source, single-workspace edition of OrcaRouter.

A self-hostable LLM gateway that routes requests across providers (OpenAI, Anthropic, Google, xAI, DeepSeek, and more) using real benchmark data, with a built-in dashboard, cost tracking, and OpenAI-compatible API.

Highlights

Routing

• Four routing strategies wired end-to-end: balanced, fastest, cheapest, quality — each functionally distinct
• model="auto" cascade powered by LiteLLM Router with catalog hygiene
• Real benchmark-driven quality routing via Artificial Analysis (AA) scores
• Hosted OrcaRouter fallback for unreachable providers, with savings attribution

Providers

• Env-configured provider keys auto-surfaced (OpenAI, Anthropic, Google, xAI, DeepSeek, etc.)
• O2 hosted integration with pinned model list and AA-matched quality scores
• AA scores matched to namespaced/dotted catalog ids (GPT-5.x, Llama 4, Qwen3, Yi, Phi)

Dashboard

• Brand-aligned UI matching orcarouter.ai
• True p50/p99 latency computed from raw samples (banker's rounding to match server)
• Hosted-auto savings tile, unreachable-models tile, top-N from orcarouter.ai/models
• Timezone-correct "X ago" timestamps

Reliability

• Streaming SSE: client disconnect closes upstream and logs 499
• DB session leak fixed in auth middleware error paths
• LiteLLM cost plumbed through; sibling versions deduped; dev-hostile defaults removed

Deploy

• Production-ready fly.toml and railway.json with verified docs
• Hardened Dockerfile builder stage with native build deps + .dockerignore
• ASGI module-level app instance for any standard server

Docs & i18n

• Full README translations: en, zh, ja, ko, de, es, fr, it, pt, ru, vi, hi
• DEMO.md and BENCHMARK.md walkthroughs

Install

```bash
git clone https://github.com/Continuum-AI-Corp/OrcaRouter-Lite.git
cd OrcaRouter-Lite

follow README quickstart

```

See README.md for the full quickstart and deploy guides.