OPERATIONS.md

# TEQUMSA Space Operations

## Startup and Always-On Verification

1. Confirm the Space hardware is set to `CPU Upgrade`.
2. Confirm Space sleep is disabled in the Hugging Face settings UI.
3. Confirm `TEQUMSA_INTERNAL_SECRET`, `SUPABASE_URL`, `SUPABASE_SERVICE_ROLE_KEY`, and `SUPABASE_PROJECT_ID` are set.
4. Open `GET /api/health` and verify:
   - `status` is `ok`
   - `target_hardware` is `cpu-upgrade`
   - `sleep_policy` is `always-on`
   - `dependency_state.kernel_loaded` is `true`

## Runtime Error Recovery

1. Check the latest Hugging Face build and runtime logs.
2. Confirm the vendored kernel file still exists at `vendor/metacognitive_kernel_v18.py`.
3. Confirm Python dependencies installed from `requirements.txt` without version conflicts.
4. Restart the Space after secrets and variables are verified.
5. If the Space still fails, route traffic through gateway fallback mode until the Space health endpoint returns `ok`.

## Rollback Path

1. Revert the Space repo to the last healthy commit.
2. Restart the Space and confirm `/api/health` is live.
3. Keep `chat.lifeambassadorsint.org` active; it already falls back locally when the Space is unhealthy.

## Public Gateway Checks

1. Call `GET https://chat.lifeambassadorsint.org/health`.
2. Confirm the response mode is `live` when the Space is healthy.
3. Send a test `POST /chat` request and confirm `source` returns `hf_space`.
4. If `source` returns `local_fallback`, inspect Space health before changing gateway code.