Equipment Gap Questionnaire

Equipment Gap Questionnaire

Conduct an interactive questionnaire with building operators to fill gaps in equipment survey data. This skill extracts available system information from local documents, identifies what's missing compared to Audette's equipment survey schema, and interviews the operator using natural language Q&A until all required fields are complete.

Required MCP Connections

No other MCPs required - Uses local file reading (Read tool) for document extraction.

Workflow Overview

1. Extract available data from documents/existing survey
2. Identify gaps relative to equipment-schema.md
3. Interview operator using AskUserQuestion
4. Parse natural language responses → map to schema fields
5. Ask follow-up questions for incomplete sections
6. Repeat until survey is 100% complete
7. Confirm final survey with operator
8. Submit to Audette

Pre-flight

Call switch_customer_account with the Audette customer account UID from the system prompt. This is required before any Audette write operations — omitting it causes HTTP 401.

If no account UID is in the system prompt, call list_customer_accounts and ask the user to select one.

Step 1: Identify Building

The building UID is in the system prompt if this asset has been linked to Audette. If not, call list_buildings to find it by name or address, or ask the user directly. You need the building_model_uid — verify it exists by calling get_building_model_details.

Step 2: Gather Available Equipment Data

2a. Check for Existing Survey

Call get_equipment_survey with the building_model_uid. If a survey exists, load it as the starting point. Note which fields are already populated vs. null.

2b. Extract from Documents (Optional)

If project documents are available, use Grep/Read to search for system descriptions.

Use Glob to find relevant documents:

**/*.pdf
**/PCA*.pdf
**/CNA*.pdf
**/*equipment*.pdf
**/*mechanical*.pdf

Use Grep with search patterns:

When Grep finds matches: Use Read to get the surrounding context (pages/sections). Extract specifications like capacity, efficiency, fuel type, installation dates.

Aggregate findings into notes about each system category.

2c. Load Equipment Schema

Read the equipment schema from the audette-equipment-survey skill: skills/audette-equipment-survey/references/equipment-schema.md

This contains all sections, fields, valid enum values, and field descriptions you'll need for gap analysis and question generation.

Step 3: Perform Gap Analysis

For each major equipment section in the schema, determine its completeness:

Critical vs. Optional Fields

Critical fields (required if _exists = true):

• *_type (equipment type enum)
• *_size or *_capacity (impacts carbon model significantly)
• *_average_installation_year (determines remaining useful life)

Optional but helpful fields:

• *_heating_type, *_cooling_type (refines model)
• *_terminal_units (distribution method)
• *_count (for elevators, rooftop units)

Gap Prioritization

Organize gaps into:

1. Critical unknowns: _exists is null (don't know if system exists)
2. High priority: _exists = true but type or size missing
3. Medium priority: Installation year, efficiency details missing
4. Low priority: Secondary characteristics missing

Step 4: Generate Targeted Questions

For each gap, generate a natural language question appropriate for a building operator.

Read references/question-templates.md now — it contains example questions for each equipment section, accounting for typical operator responses.

Question Design Principles

DO:

• Use plain language, not technical jargon (unless operator is highly technical)
• Ask about observable equipment ("What type of boilers do you have?")
• Accept ranges and estimates ("About how old are the boilers? Best guess is fine.")
• Ask for counts and sizes together ("How many chillers, and what's the capacity?")
• Provide examples in parentheses ("Natural gas, electric, oil, or other?")

DON'T:

• Ask for exact enum values (operator won't know "condensing_gas_boiler" enum)
• Require precision when estimates work (installation year ±2 years is fine)
• Use acronyms without explanation (say "domestic hot water" not just "DHW")
• Ask multiple unrelated questions in one (break into separate questions)

Example Question Transformations

Step 5: Conduct Interactive Interview

Use AskUserQuestion to interview the building operator. Work through gaps in priority order (critical → high → medium → low).

Interview Flow

1. Start with critical unknowns (whether systems exist)

   • "Does the building have a central heating system, or is heating suite-level?"
   • "Is there a central chiller or cooling system?"

2. For confirmed systems, gather details

   • "You mentioned a central boiler. What type? (Gas, electric, oil?)"
   • "How many boilers are there, and what's the capacity of each?"
   • "About how old are the boilers? Even a rough estimate helps."

3. Parse responses and map to schema

   • "2 Weil-McLain gas boilers at 500 MBH each" →

     central_plant_heater_exists: true
     central_plant_heater_type: "gas_boiler"  (Weil-McLain typically makes non-condensing)
     central_plant_heater_size: 83.3  (1000 MBH ÷ 12 = 83.3 tons)
     

4. Ask follow-ups for ambiguity

   • Response: "Gas boilers"
   • Follow-up: "Are these condensing boilers (high-efficiency, AFUE ≥90%) or standard boilers?"

5. Confirm and move to next gap

   • "Got it — 2 standard gas boilers, 500 MBH each, installed around 2005. Now, about the cooling system..."

Handling Natural Language Responses

Building operators will respond conversationally. Parse key information:

Clarification Strategy

When responses are ambiguous:

If manufacturer/model mentioned: Use it to infer type

• "Weil-McLain" → likely standard gas boiler (not condensing)
• "Navien" → likely condensing gas boiler or tankless water heater
• "Carrier" → could be rooftop unit or split system (ask follow-up)

If capacity units unclear: Ask for clarification

• "500" → "Is that 500 MBH, 500 kW, or 500 BTU/hr?"

If age is vague: Accept estimates

• "Pretty old" → "Best guess — installed in the 1990s, 2000s, or earlier?"

If distribution method unclear:

• "How is heat distributed? Radiators, baseboard, forced air, or something else?"

Step 6: Validate Completeness

After each section, verify all required fields are populated.

Completeness Checklist

For each section where _exists = true, confirm:

• Type field populated with valid enum value
• Size/capacity field populated (if applicable)
• Installation year populated (estimate acceptable)
• Distribution/terminal units specified (for central plants)

Handle Edge Cases

Multiple systems of different types:

• If building has 2 boilers + 1 heat pump, ask: "Which is the primary heating system?" or "Do boilers handle majority of the load, or heat pump?"
• Survey schema expects one primary system per section. If truly hybrid, document in notes and choose the dominant system for the survey.

Partial retrofits:

• If half the building has new equipment and half has old: "What year was the most recent equipment installed? We'll use that as the average."

Unknown details:

• If operator genuinely doesn't know (e.g., "Not sure about chiller capacity"), note it:
  • Option 1: Estimate based on building size (e.g., "For a 50,000 SF building, typical chiller is 150-200 tons. Does that sound right?")
  • Option 2: Mark for site visit / nameplate verification

Step 7: Review and Confirm

Once all gaps are filled, present the complete survey back to the operator for confirmation.

Confirmation Format

## Equipment Survey Summary

I've gathered the following information. Please review and confirm before I submit to Audette:

### Heating System ✓
- **Type:** Gas boiler (standard, non-condensing)
- **Capacity:** 83.3 tons (1,000 MBH total — two 500 MBH boilers)
- **Distribution:** Hydronic baseboard radiators
- **Installation Year:** 2005

### Cooling System ✓
- **Type:** Rooftop units (split system air conditioners)
- **Capacity:** 75 tons (15 units @ 5 tons each)
- **Installation Year:** 2010

### Domestic Hot Water ✓
- **Configuration:** Suite-level (not centrally distributed)
- **Type:** Electric storage water heaters
- **Capacity:** 4.5 kW element rating ÷ 3.517 = **1.28 tons** per unit (50 gal tank; nameplate rating used directly — if only tank volume had been known, the fallback rule of thumb `gallons ÷ 40` would give ≈1.25 tons, consistent with the nameplate figure here)
- **Installation Year:** 2015

[... continue for all sections ...]

**Does this look correct? Any changes needed before I submit?**

If operator identifies errors or additions, revise the affected sections and re-confirm.

Step 8: Submit to Audette

Once confirmed, call submit_equipment_survey with the complete survey payload.

Submission Format

Build the JSON payload according to the equipment schema. All top-level sections must be present. For sections where equipment doesn't exist, set _exists: false and all other fields to null.

Example payload structure:

{
  "building_model_uid": "abc-123-def",
  "air_handling_equipment": {
    "air_handling_equipment_exists": true,
    "air_handling_equipment_type": "packaged_air_handling_unit",
    "air_handling_equipment_heating_type": "hydronic",
    "air_handling_equipment_cooling_type": "direct_expansion",
    "air_handling_equipment_supply_air_rate": 12000.0,
    "air_handling_equipment_average_installation_year": 2008
  },
  "central_plant_cooler": {
    "central_plant_cooler_exists": false,
    "central_plant_cooler_type": null,
    "central_plant_cooler_terminal_units": null,
    "central_plant_cooler_size": null,
    "central_plant_cooler_average_installation_year": null
  },
  "central_plant_heater": {
    "central_plant_heater_exists": true,
    "central_plant_heater_type": "gas_boiler",
    "central_plant_heater_terminal_units": "baseboards",
    "central_plant_heater_size": 293.0,
    "central_plant_heater_average_installation_year": 2005
  },
  ... [all other sections]
}

After submission, inform the operator that the survey is complete and carbon modeling will be re-run automatically.

Important Rules

Parsing Natural Language

• Be generous with interpretation: "Old gas boilers" → infer gas_boiler type
• Confirm assumptions: "I'm assuming these are standard gas boilers, not condensing. Correct?"
• Handle units gracefully: Convert MBH, BTU/hr, or kW to ton-equivalents for every heating/cooling/DHW size field — never submit kW or gallons/litres for these fields
• Accept estimates: Installation year ±2 years is fine, capacity ±10% is fine

Unit Conversions

Keep these conversions ready:

All heating, cooling, and DHW capacity sizes (central_plant_cooler_size, central_plant_heater_size, central_plant_heat_pump_size, terminal_cooler_size, terminal_heater_size, terminal_heater_cooler_size, heat_pump_size, domestic_hot_water_heater_size) must be submitted as ton-equivalents — never kW, never litres/gallons. Per Christopher (2026-07-16): "all heating cooling and DHW size must be converted to ton equivalents." A DHW heater's thermal/recovery rating converts to tons the same way as any other heating capacity, and is preferred whenever documented. Rule of thumb when only tank volume is known: tons ≈ gallons ÷ 40 (litres ÷ 151) — based on typical storage water heater sizing (~4.5 kW element per 40–50 gal tank ≈ 1 ton per ~40 gal). Flag when this fallback was used instead of a documented nameplate rating.

Enum Mapping

Read references/enum-mapping.md for common operator terms → schema enum mappings.

When operator says "gas boiler," infer:

• If they mention "high efficiency" or "condensing" → condensing_gas_boiler
• Otherwise → gas_boiler (standard)
• If unsure, ask: "Is this a condensing boiler (AFUE ≥90%)?"

Never Invent Data

If operator doesn't know something and can't estimate:

• Mark it in notes: "Chiller capacity unknown — requires site visit for nameplate"
• Set field to null in submission
• Inform user which fields are still missing

When to Use This Skill vs. Other Skills

Use equipment-gap-questionnaire when:

• Documents don't provide enough detail to complete survey
• User wants to interview building operator to fill gaps
• Existing survey is partially complete and needs finishing
• Interactive Q&A is preferred over manual form-filling

Use audette-equipment-survey when:

• Complete equipment data is already in documents (PCNA, CNA, etc.)
• No operator interview needed
• Extracting and submitting in one shot

Use system-details when:

• User wants a comprehensive overview (not submitting to Audette)
• Preparing for a meeting or review
• Synthesis-oriented, not submission-oriented

Additional Resources

Reference Files

For question templates and enum mapping:

• references/question-templates.md - Example questions for each equipment section, handling typical operator responses
• references/enum-mapping.md - Common operator terminology mapped to schema enum values

The equipment schema itself lives in the audette-equipment-survey skill and should be loaded from there: skills/audette-equipment-survey/references/equipment-schema.md