Omnigate

Omnigate is a secondary development project based on the open-source new-api codebase. The upstream project provides the core protocol adapters, usage tracking, billing, and user management. My work focused on branding, safer defaults, and a production deployment and operations setup for a single server.

A self-hosted model gateway needs more than application code. HTTPS, streaming proxy behavior, database and cache services, secret templates, backups, and a repeatable deployment process all need to work together.

The deployment builds the application from source and runs it with Caddy, PostgreSQL, and Redis through Docker Compose. Caddy terminates HTTPS without buffering streamed responses. The repository also includes environment templates, a database backup script with retention cleanup, and an operations log.

The service is running at omnigate.cc. Clients can use one base URL and one issued token with OpenAI-compatible tools, while the deployment files document how the service is configured, backed up, and updated.

This project taught me how to work through a mature open-source codebase before changing it, and how to keep custom work within a maintainable boundary. I also gained a more practical understanding of streamed responses behind a reverse proxy, and how HTTPS, PostgreSQL, Redis, database backups, and upgrades fit together in production. Handling the license and upstream attribution made me more careful about separating upstream capabilities from my own work.

• Source-built Docker Compose deployment with Caddy, PostgreSQL, and Redis
• Automatic HTTPS with response buffering disabled for streamed output
• Database backup script with configurable retention cleanup
• Project documentation clearly separates upstream features from custom work