Test AI assisted BRICK data modeling

[bbartling]

🎯 Goal

Test an AI-assisted workflow where an LLM helps a human tag an exported Open-FDD data model JSON by adding Brick metadata, then re-importing it to update the Brick model and re-running validation.

---

🛠 What To Do

1. Export a site data model from the Open-FDD FastAPI CRUD app in JSON format (sites/equipment/points).
2. Ask an LLM to add a new key to the JSON structure for each point, for example a brick object that includes:
   • Brick class/tag (e.g., brick:Supply_Air_Temperature_Sensor)
   • External time-series reference (Timescale key / column name / measurement id)
   • Raw BACnet point reference (device + object type/instance or human-readable point name)
3. Review and correct the AI output (human-in-the-loop).
4. Import the updated JSON back into Open-FDD so it updates the existing Brick model.
5. Re-run model validation and any available rule-runner checks to confirm the model is usable.

---

📎 Deliverable

Post a short Markdown summary including:

• Sanitized example of the JSON before/after (small snippet is fine)
• What the AI got right vs wrong
• Where the AI saved time vs created extra work
• Any validation errors or successes after re-import