Hey,
I recently found Aurora (a Go-based LLM gateway) on the internet. I liked the lightweight concept but found it lacked essential production features, advanced routing, and a flexible management interface. I deeply upgraded it into a stable, feature-rich fork for my own infrastructure.
Most importantly, it completely fixes the recent early-September OpenCode API updates that enforced strict session headers and broke local coding tools (Goose, Pi, Claude Code) with HTTP 400 errors. I built a Session Hub engine inside the gateway that automates header transformation, keeps connections flawless, and preserves upstream Prompt Cache.
The project is fully functional, documented, and available on Docker Hub (entbtw/aurora:latest).
Github - https://github.com/entitybtw/aurora
Core Features:
- Fixed OpenCode API: Automatically maps local client traffic into stable, valid session IDs via Session Hub (
mapmode) to prevent HTTP 400 drops and keep upstream Prompt Cache warm. - Multi-IP Host Networking: Supports advanced per-provider
bind_ipconfigurations and load-balanced pools, making it perfect for strict network routing and multi-account setups. - Web UI Administrative Dashboard: Full runtime management for 14+ providers and routing strategies with instant persistence to disk config files. Eliminates the need for .env editing or container restarts.
- Visual Fallback Chains: Configure, edit, and prioritize backup models and fallback rules dynamically.
Check out the repository for detailed orchestration guides, full environment tables, and exact API curl examples. Feedback on the Go architecture and session handling layer is highly welcome!
Top comments (0)