OmniRoute + OpenCode: local gateway setup tutorial

This tutorial connects OpenCode to OmniRoute, an open source AI gateway that runs on the user's machine. The demonstrated request passes through the local gateway to an external provider and back. It does not demonstrate running model inference locally or offline.

The walkthrough starts with a global npm installation and opens the dashboard on port 20128. The presenter describes local SQLite storage for keys, usage and history, with encrypted credentials and no telemetry. These are claims made in the video, rather than an independent security assessment.

For the coding assistant setup, the presenter connects the OpenCode provider without authentication, then adds a custom provider in the OpenCode client. The configuration uses localhost:20128/v1, a newly generated gateway API key and an exact model ID copied from the dashboard. The provider connection and the client's gateway key are separate steps. A greeting tests Big Pickle before the presenter adds MiniMax as another selectable model.

The dashboard also includes usage analytics, provider health and API format translation. Its Combos page configures provider chains and routing strategies. The presenter says automatic fallback can move requests to another provider when a free tier reaches its limit. Available models and provider quotas still determine what the setup can use.