vibe12-ai-commissioning-api

name	vibe12-ai-commissioning-api
description	Use when validating MQTT→cloud telemetry, calling Vibe12 Dashboard Lambda APIs for commissioning status, BRICK time-series refs, points registry, or auth from Codex/Cursor/OpenClaw. Triggers on: commissioning API, cloud_ingest_ok, telemetry flow, brick_timeseries_ref, external_ref, validate_cloud_pipeline, series_flowing, DashboardUrl, Bearer token, demo bens-office.

Vibe12 AI commissioning API (Codex-transferable)

HTTP JSON on the DashboardUrl (Lambda Function URL). Full curl patterns: references/codex-api-cookbook.md.

Auth (every session)

export PATH="$HOME/.local/bin:$PATH"
cd ~/py-bacnet-stacks-playground/vibe_code_apps_12
URL=$(aws cloudformation describe-stacks --stack-name vibe12cloud --region us-east-2 \
  --query "Stacks[0].Outputs[?OutputKey=='DashboardUrl'].OutputValue" --output text | tr -d '\r')
URL="${URL%/}"
# Password from samconfig.toml WebPassword (never commit)
TOKEN=$(curl -sS -X POST "$URL/api/auth/login" -H 'Content-Type: application/json' \
  -d '{"username":"engineer","password":"YOUR_PASSWORD"}' | python3 -c "import sys,json; print(json.load(sys.stdin)['token'])")
AUTH=(-H "Authorization: Bearer $TOKEN")

Or run ./scripts/validate_cloud_pipeline.sh (loads password from samconfig.toml).

Gate: is data flowing?

curl -sS "$URL/a../edge_backup/status/demo/bens-office?window_minutes=20" "${AUTH[@]}" \
  | python3 -c "import sys,json; d=json.load(sys.stdin); assert d['cloud_ingest_ok']; print(d['series_flowing'], '/', d['series_total'])"

Pass: cloud_ingest_ok: true and series_flowing ≥ 1 for the job's required points.

Endpoints (agent cheat sheet)

Method	Path	Use
GET	/api/health	Deploy revision, no auth
POST	/api/auth/login	JWT
GET	/a../edge_backup/status/{site}/{building}	Primary gate + ai_hints + recommended_actions
GET	/api/telemetry/flow/{site}/{building}	Per-series flowing, last_value
GET	/api/brick/timeseries-ref/{site}/{building}	BRICK refs for modeling
GET	/api/points/{site}/{building}	Registry
GET	/api/brick/{site}/{building}	BRICK graph
GET/POST	/api/data-model/{site}/{building}/...	Canonical model, TTL, import
POST	/api/playground/test-rule	FDD test window
POST	/api/playground/go-live	FDD backfill

Query: window_minutes (default 15) on flow/commissioning.

BRICK time-series ref (DynamoDB)

Each sample includes:

• series_id / external_ref → DynamoDB device_id (partition key)
• entity_id → brick:{site}/{building}/point/... for graph / SparkQL
• brick_timeseries_ref JSON → mqtt_topic, dynamodb.table_key, brick_class

Ingest writes these automatically; agents read via /api/brick/timeseries-ref/....

Human + agent workflow

1. Human: SSH, discover, enable points.csv
2. Agent: Ansible deploy → ./scripts/validate_cloud_pipeline.sh
3. Human: confirm points match job
4. Agent: BRICK graph / data-model APIs
5. Human + agent: Rule Lab → test-rule → go-live

Failure hints

Symptom	Check
cloud_ingest_ok: false	IoT policy publish ARNs; memory/integrations/aws-iot.md
Pi publishes, cloud empty	PUBACK 135 on Pi — policy not hierarchical
series_total: 0	Wrong API path parsing (fixed rev 8+) or no ingest yet
401	WebPassword vs samconfig

Related

• docs/ai-commissioning-api.md
• skills/vibe12-ansible-edge/
• skills/vibe12-cloud-deploy/