GoModel is a LiteLLM alternative for the post-compromise era
GoModel hit Show HN April 21 (147 points) as an open-source AI gateway written in Go. One OpenAI-compatible endpoint that fans out to OpenAI, Anthropic, Gemini, Groq, xAI, and Ollama. Aliases, scoped workflows, exact-match caching, audit logs, per-user usage tracking. Docker quick start.
The interesting framing is in the founder's writeup: they say they started this because LiteLLM got compromised. Not "we want a faster proxy" β "we do not trust the dominant Python proxy anymore and want to rebuild it in Go with a smaller blast radius." Whether or not you fully buy the threat model, this is the kind of supply-chain anxiety that has been spreading through agent infra teams since the OX Security MCP disclosure.
This matters because every team running production agents is one rogue dependency away from a bad day. A typed, statically-compiled, single-binary gateway that you can audit is a very different operational profile from a Python package with hundreds of transitive dependencies. Whether the market cares enough to switch is the open question, but the wedge is clear.
GitHub at github.com/ENTERPILOT/GoModel, project page at gomodel.enterpilot.io. If you are running anything past a toy prompt count, worth a look.
← Back to all articles
The interesting framing is in the founder's writeup: they say they started this because LiteLLM got compromised. Not "we want a faster proxy" β "we do not trust the dominant Python proxy anymore and want to rebuild it in Go with a smaller blast radius." Whether or not you fully buy the threat model, this is the kind of supply-chain anxiety that has been spreading through agent infra teams since the OX Security MCP disclosure.
This matters because every team running production agents is one rogue dependency away from a bad day. A typed, statically-compiled, single-binary gateway that you can audit is a very different operational profile from a Python package with hundreds of transitive dependencies. Whether the market cares enough to switch is the open question, but the wedge is clear.
GitHub at github.com/ENTERPILOT/GoModel, project page at gomodel.enterpilot.io. If you are running anything past a toy prompt count, worth a look.
Comments