A lightweight LLM gateway written in Go, inspired by LiteLLM.
Features:
- OpenAI-compatible REST API (ChatCompletion) so existing OpenAI SDKs can point to this gateway.
- Route requests to multiple back-ends: OpenAI, Azure OpenAI, AWS Bedrock (more coming).
- Simple round-robin load balancing between multiple deployments of the same provider.
- Optional per-key rate limiting (token bucket, in-memory for now).
- YAML configuration file similar to
litellm_config.yaml
.
⚠️ Enterprise-only features of LiteLLM are out-of-scope for this project.
# install deps
go mod tidy
# run server with sample config
go run ./cmd/server --config examples/config.yaml
See examples/config.yaml
for available options.
- Basic router
- Metrics / Prometheus
- Pluggable auth & logging callbacks