Routerly launches self-hosted LLM gateway

This is a real infra play, not just another proxy wrapper. The pitch is attractive because it makes routing and spend control explicit, but the project will live or die on whether the policy engine beats simpler static rules in practice.

• –Nine policies cover cheapest, fastest, healthiest, most capable, context, fairness, rate limits, budget-remaining, and LLM-driven selection.
• –No database, Redis, or PostgreSQL makes it much easier to self-host on a small footprint.
• –OpenAI and Anthropic compatibility lowers adoption friction across Cursor, LangChain, Open WebUI, and similar clients.
• –The best fit is multi-tenant SaaS and local-first teams that need per-project tokens, hard caps, and failover.
• –The project is still early, and the site/repo licensing copy is inconsistent, both of which are rough edges serious users will notice fast.