Build a District Cooling Plant Maintenance Outage Scheduler — implementation

## Build
Evaluate planned chiller overhauls against district cooling customer contracts and ambient weather loads. Synthesize outage sequence plans that preserve minimum customer delta-T and chilling tonnage.

## User
District cooling operational dispatchers and chilled water network coordinators

## Workflow
1. Ingest scheduled chiller maintenance requests and customer contracted refrigeration tonnage limits.
2. Model cooling demand profiles using EnergyPlus epJSON chiller objects and ambient wet-bulb predictions.
3. Synthesize customer chilled water outage allocations to avoid critical pressure or cooling loss.
4. Compile an operations outage plan and dispatch notice for customer engineering review.

## Interface
Multi-pane dashboard displaying plant chiller staging, customer delivery pressure profiles, and maintenance sign-off approvals.

## Data
Use open-source synthetic normalized district cooling network logs with EnergyPlus epJSON HVAC templates stored in a GitHub mock repository.
Starter examples below are synthetic. They are not downloaded records.

## Source
[EnergyPlus](https://raw.githubusercontent.com/NREL/EnergyPlus/develop/README.md)
Extracts chiller plant modeling schemas and input object conventions from the EnergyPlus epJSON schema repository.

## Tools
anthropic-claude, pydantic-ai, mlflow.
anthropic-claude provides hosted structured analysis and bilingual notification drafts without running untrusted local model instances.
Hosted Claude API for structured schema parsing, constraint reasoning, and operations report drafting.

## Deliverables
Submit runnable source, pinned dependencies, setup commands, a working interface, example outputs, and a recorded end-to-end demo. Include tests and an action log.

## Acceptance
- Validates chiller maintenance windows against total peak customer cooling load demand.
- Flags violations where scheduled outages exceed available thermal energy storage buffer capacity.
- Produces structured JSON outage schedules including bilingual English-Arabic customer advisory drafts.
- Requires explicit human engineer confirmation before publishing simulated network dispatch actions.

## Starter Examples
```json
[
  {
    "input": {
      "chiller_id": "CH-01",
      "capacity_tonnage": 2500,
      "maintenance_window_hours": 8,
      "ambient_wet_bulb_c": 31.5,
      "connected_customers": [
        {
          "customer_id": "CUST-TWR-A",
          "demand_tonnage": 1800,
          "critical_facility": false
        }
      ],
      "storage_ton_hours": 12000
    },
    "expected_output": {
      "feasible": true,
      "backup_chiller": "CH-02",
      "storage_draw_tonnage": 600,
      "unmet_load_tonnage": 0,
      "customer_advisories": [
        {
          "customer_id": "CUST-TWR-A",
          "flow_derate_pct": 0,
          "advisory_en": "Scheduled maintenance CH-01 mitigated by thermal storage.",
          "advisory_ar": "صيانة مجدولة للمبرد CH-01 مغطاة بالتخزين الحراري."
        }
      ]
    }
  },
  {
    "input": {
      "chiller_id": "CH-02",
      "capacity_tonnage": 3000,
      "maintenance_window_hours": 12,
      "ambient_wet_bulb_c": 34,
      "connected_customers": [
        {
          "customer_id": "CUST-HOSPITAL",
          "demand_tonnage": 2800,
          "critical_facility": true
        }
      ],
      "storage_ton_hours": 2000
    },
    "expected_output": {
      "feasible": false,
      "backup_chiller": null,
      "storage_draw_tonnage": 2000,
      "unmet_load_tonnage": 800,
      "customer_advisories": [
        {
          "customer_id": "CUST-HOSPITAL",
          "flow_derate_pct": 28.57,
          "advisory_en": "Outage rejected: Critical deficit of 800 tons for hospital client.",
          "advisory_ar": "تم رفض الإيقاف: عجز حرج قدره 800 طن تبريد للمستشفى."
        }
      ]
    }
  }
]
```

## Safety
Simulates dispatch actions only; requires manual validation by licensed mechanical engineers before actual valve operations.
Use mock external writes. Require human review before any real deployment.