Support custom API base URL and API key (OpenAI-compatible provider)

Feature Request

Allow commandcode to use a custom API base URL and API key from external providers (e.g. any OpenAI-compatible gateway), instead of being hardwired to api.commandcode.ai.

Motivation

Right now all models are routed through Command Code's own backend (api.commandcode.ai). Users on other providers — or self-hosted / relay gateways that expose an OpenAI-compatible /v1 endpoint — cannot point commandcode at them. The only knob that exists, COMMANDCODE_API_URL, switches between Command Code's own prod/staging/local environments; it does not accept an external provider base URL.

Proposed behavior

Something like the existing opencode / Claude Code config model:

• A config file and/or env vars, e.g.:
  • COMMANDCODE_BASE_URL=https://my-provider.example.com/v1
  • COMMANDCODE_API_KEY=sk-...
• Optional custom model list (id → display name, context window, modalities) when pointing at a non-Command-Code backend.
• Priority: CLI flag > config file > env var > current auth.

Use case

I maintain an opencode.jsonc with a custom provider (npm: @ai-sdk/openai-compatible, baseURL + apiKey) and would love to drive commandcode with the same endpoint/key instead of a separate Command Code subscription key.

Environment

• commandcode 1.11.0
• Linux, Node v26

Inizia tracciando la configurazione esistente di COMMANDCODE_API_URL e il percorso di autenticazione attuale. Implementa le fonti proposte per COMMANDCODE_BASE_URL e COMMANDCODE_API_KEY, inclusi CLI, file di configurazione, ambiente e precedenza dell’autenticazione attuale, oltre ai metadati opzionali per i modelli personalizzati. Il lavoro è completato quando è possibile usare un endpoint esterno compatibile con OpenAI e una chiave senza una chiave di abbonamento a Command Code.