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.