opencode-models (tmiland-lab self-hosting API, static surface)

Public static API for the tmiland-lab/opencode self-hosting fork, published via GitHub Pages at https://tmiland-lab.github.io/opencode-models.

Endpoints

Path	Source	Use
`/api.json`	hourly sync from `https://models.opencode.ai/api.json` (`.github/workflows/sync.yml`)	`OPENCODE_MODELS_URL=https://tmiland-lab.github.io/opencode-models`
`/api.json.pinned`	manual pin of a known-good catalog	disaster fallback: point the env var here if upstream ships a bad catalog

Refresh cadence is ours (hourly cron, dispatchable).
Fork-specific entries (local Ollama models, gateway models with our caps) get overlaid here later — upstream can never throttle or reshape the catalog under us.
Stateful parts (LLM gateway with per-key quotas) live on the vps1 host as a systemd service, not here — GitHub hosts only the static surface.

Create public repo tmiland-lab/opencode-models with this layout.
Settings → Pages → Deploy from branch → main → /docs.
Run sync workflow once via dispatch, verify https://tmiland-lab.github.io/opencode-models/api.json returns 200.
Keep https://vps1.tmiland.com/opencode-models/api.json as fallback mirror (already live, hourly cron on vps1).