// the find
ENTERPILOT/GoModel
AI gateway written in Go. Lightweight unified OpenAI-compatible API for OpenAI, Anthropic, Gemini, Groq, xAI & Ollama. LiteLLM alternative with observability, guardrails, streaming, costs and usage tracking.
GoModel is an AI gateway written in Go that sits in front of OpenAI, Anthropic, Gemini, and 20+ other providers, exposing a unified OpenAI-compatible API. It handles auth, cost tracking, caching, rate limiting, failover, and an admin dashboard in a single binary. Aimed at teams that want LiteLLM-style multi-provider routing without Python's overhead.
Single static Go binary with genuinely low resource usage — the benchmarks in the repo are self-reproducible, not marketing screenshots. The feature set is unusually complete for a young project: semantic response caching, hierarchical user paths with per-key budgets, circuit breakers, and an MCP gateway all in one place. The ADR documents (0001–0008) show actual design thinking about ingress frames, capability models, and cache semantics rather than just bolting features together. Dashboard is embedded in the binary with pre-built assets, so there's nothing extra to deploy.
The dashboard JS bundle is a compiled artifact checked into the repo — you can't audit or modify the frontend without knowing the separate build pipeline, and there's no source in-tree. Storage is PostgreSQL or MongoDB, which means you're running real infrastructure for what is often a sidecar concern; Redis is optional for caching but the compose file already pulls in three databases. The project is young (< 1018 stars, sub-1.0 versioning) and the roadmap calls out commercial features, so the open-source/paid split is unclear and could shift. Streaming passthrough for Anthropic's native API was flagged as having compatibility findings as recently as July 2026 (docs/dev/2026-07-17_anthropic-sdk-compat-findings.md), meaning edge cases in that path are still being worked out.