mgoodman60 Project Data Intelligence (v3.0) — SKILL Documentation
Overview
Comprehensive extraction and intelligence system for construction project documents.
1. Philosophy: "No Data Left Behind"
The project-data-intel skill is the single source of truth for all extracted construction project data. It operates on a universal principle: every piece of information captured during project setup, document processing, and field operations should be searchable, viewable, and actionable from a single, role-aware interface.
Key design principles:
- Template-based architecture: Static HTML template ships with the plugin, dynamic data is injected
- Project-agnostic: Works for any building type (senior care, office, industrial, healthcare, etc.)
- Incremental updates: Only changed data is re-serialized; unchanged sections remain cached
- Role-aware rendering: Superintendent, PM, Owner, and Architect each see customized views
- AI-native: Chatbot and rendering system leverage Anthropic API for context-aware insights
- Offline-capable: Keyword search fallback when API unavailable
2. Architecture: Template + Data Injection
The v3.0 dashboard uses a separation of concerns model:
File Structure
AI - Project Brain/
├── {PROJECT_CODE}_Data_Intel.html ← Static template (10,000+ lines, generated once)
├── {PROJECT_CODE}_data.js ← Dynamic data payload (regenerated each run)
├── references/
│ └── data-intel-template.html ← Master template in plugin (never regenerated)
└── logs/
└── data-intel-generation.log ← Audit trail
Template Lifecycle
- First run: Copy
data-intel-template.htmlfrom plugin →{PROJECT_CODE}_Data_Intel.html - Subsequent runs: Never touch the HTML file; only regenerate
{PROJECT_CODE}_data.js - Rationale: Template is stable (~10,000 lines), data evolves rapidly. Decoupled updates reduce file size and generation time.
Data Payload Structure (data.js)
const PROJECT_DATA = {
// Core project metadata and configuration
config: {
projectCode, projectName, address, buildingType, occupancy,
company, startDate, endDate, architect, pm, super, status
},
// Spatial & geometric data
spatial: {
gridLines: { x: [...], y: [...] },
buildingAreas: { footprint, perimeter, floors, zones },
rooms: [{id, name, area, type, occupancy, finishes, doors, equipment}],
floorPlans: [{floor, gridRef, area, roomCount}]
},
// Building specifications and systems
specs: {
building: {structure, foundation, roof, exterior, interior},
pemb: {supplier, model, dimensions, frames, reactions, design_loads},
cfs: {profile, gauge, spacing, height},
concrete: {bearing, sog_interior, sog_exterior, rebar, mixes, anchor_bolts, testing},
mep: {
equipment: [
{
id: "RTU-1", tag: "RTU-1", type: "Rooftop Unit",
discipline: "mechanical", // mechanical|plumbing|electrical|fire_protection
system: "hvac", // hvac|exhaust|plumbing|electrical_power|lighting|fire
description: "10-Ton Rooftop Unit",
location: { grid: "C-D/3-4", room: null, mounting: "Roof" },
capacity: { cooling_tons: 10, heating_mbh: 250, airflow_cfm: 4000 },
electrical: { voltage: 208, phase: 3, mca: 48, mocp: 60 },
served_rooms: ["101","102","103"],
manufacturer: null, model: null,
spec_section: "23 81 26", source_sheet: "M-301"
}
],
panels: [{
panel: "LP-1", location: "Electrical Room 110",
voltage: "208/120V", phase: 3, wires: 4,
main_breaker_amps: 225, bus_rating_amps: 225,
fed_from: "MDP", mounting: "surface", aic_rating: 22000,
circuits: [{number: 1, breaker_amps: 20, poles: 1, description: "Lighting 101-103", va: 1800, phase: "A"}],
totals: {connected_va: 45000, demand_va: 31500, spare_breakers: 6, space_slots: 4}
}],
single_line: {
utility_service: {voltage: 208, phase: 3, service_amps: 800},
main_switchboard: {rating_amps: 800, main_breaker_amps: 800},
transformers: [], distribution_tree: [],
generator: null, ats: null, ups: null
},
fixtures: {
lighting: [{mark: "A", description: "2x4 LED Troffer", wattage: 40, lumens: 5000, cct_k: 4000, mounting: "recessed", quantity: 48}],
plumbing: [{tag: "WC-1", type: "water_closet", manufacturer: null, model: null, mounting: "wall-hung", flush_gpf: 1.28, ada: true, quantity: 12}]
},
distribution: {
duct_mains: [{size: "24x12", type: "supply", from_equipment: "RTU-1", material: "galvanized"}],
pipe_mains: [{system: "domestic_cold", size: "2\"", material: "copper"}]
},
fire_protection: {
system_type: "wet", design_standard: "NFPA 13", hazard_class: "Light",
riser: {location: "Mech Room 110", size: "6\""}, fdc: {location: "East exterior", type: "Siamese"},
heads: [{type: "concealed_pendant", temp_rating: "155F", k_factor: 5.6, coverage_sqft: 225}],
fire_pump: null
},
device_counts: [{room: "101", duplex: 6, gfci: 2, dedicated: 1, data_telecom: 3}],
conflicts: [],
extraction_coverage: {
sheets_processed: [],
total_equipment_count: 0,
completeness_pct: 0
}
},
finishes: {flooring, wall_finishes, ceiling, paint, acoustical},
doors: [{opening, type, material, hardware, schedule, supplier}],
windows: [{opening, type, material, schedule, supplier}],
casework: [{id, location, type, material, supplier}]
},
// Schedules and timeline
schedule: {
baseline: [{task, duration, start, finish, preds, slack, critical}],
milestones: [{event, plannedDate, actualDate, status}],
critical_path: [],
float_analysis: {},
procurement_gantt: [{item, supplier, orderDate, deliveryDate, duration}]
},
// Project directory & contacts
directory: {
owner: [{name, role, phone, email, address}],
gc: [{name, company, phone, email}],
architect: [{name, company, phone, email}],
engineers: [{discipline, name, company, phone, email}],
subs: [{trade, company, name, phone, email, contract}],
suppliers: [{category, name, contact, phone, email, terms}],
inspectors: [{agency, contact, phone, email}],
equipment_operators: [{equipment, name, phone, license}]
},
// Procurement & change control
submittals: {
status_summary: {pending, review, approved, rejected, deferred},
pipeline: [{id, item, supplier, submitted, days_in_review, status, approver, notes}]
},
procurement: {
po_log: [{po_number, supplier, item, quantity, unit_price, total, date, status}],
material_tracking: [{item, po, location, delivery_date, received, notes}],
vendor_quotes: [{item, vendors: [{name, price, delivery, lead_time}]}],
alerts: [{priority, item, issue, action_required, due_date}]
},
// Hold points & inspections
inspections: {
hold_points: [{hp_id, description, trigger, responsible, status, date_completed}],
test_results: [{test_id, type, location, result, pass_fail, engineer, date}],
third_party: [{agency, type, scope, inspection_date, result, notes}],
photo_log: [{id, location, date, category, photographer, notes}]
},
// Daily operations
dailyReports: [{date, superintendent, weather, temp, crew_size, work_summary, issues, safety_events}],
// Request for Information
rfis: {
log: [{rfi_id, date_submitted, question, submitted_by, assigned_to, days_open, response, date_resolved}],
categories: {design, constructability, materials, coordination, other}
},
// Meetings & coordination
meetings: [{date, type, attendees, agenda, decisions, action_items}],
// Change management
changeOrders: [{co_id, description, cost_impact, schedule_impact, status, date, approvals}],
// Delays & impacts
delays: [{id, event, cause, trade, duration_days, impact, mitigation, status}],
// Pay applications & financials
payApps: {
summary: {total_contract, total_completed, percent_complete, total_invoiced, retainage},
history: [{period, contractor, amount, completed_value, retainage, approved_date}]
},
// Punch list & closeout
punchList: [{id, location, item, trade, priority, assigned_to, status, due_date}],
// Visual context (optional)
visualContext: {
site_photos: [{date, location, category, photographer, file_reference}],
progress_photos: [{date, phase, photographer, file_reference}],
as_built_markup: [{drawing, date, marked_up_by, file_reference}]
},
// AI rendering requests (optional)
renderings: [{prompt, type, model, generated_date, ai_image_reference}]
};
const API_KEYS = {
anthropic: process.env.ANTHROPIC_API_KEY || '',
gemini: process.env.GEMINI_API_KEY || '',
flux2: process.env.FLUX2_API_KEY || ''
};
const ACTIVE_ROLE = 'superintendent'; // Changed dynamically in UI
const DATA_MANIFEST = {
generated: '2026-02-19T14:30:00Z',
template_version: '3.0.0',
project_code: 'MOSC-825021',
files: {
'config.json': {md5: 'abc123', rows: 1, last_update: '...'},
'spatial.json': {md5: 'def456', rows: 42, last_update: '...'},
'specs.json': {md5: 'ghi789', rows: 156, last_update: '...'},
// ... 14 more files
},
section_visibility: {
mep_equipment: true,
mep_panels: true,
mep_single_line: true,
mep_fixtures: true,
mep_fire_protection: true,
mep_data_gaps: true,
doors_hardware: true,
concrete_mix_designs: true,
field_tolerances: true,
// ... auto-populated based on data content
}
};
3. Incremental Update System
The dashboard uses smart regeneration to minimize file I/O and improve performance:
Update Algorithm
function updateProjectData(projectPath, sourceFiles) {
// Step 1: Check if data.js exists
if (NOT EXISTS data.js) {
// Full generation path
generateAllJSON(sourceFiles) → data.js
copyTemplate(plugin/references/data-intel-template.html) → {PROJECT_CODE}_Data_Intel.html
return "NEW_DASHBOARD_CREATED"
}
// Step 2: Load existing manifest
manifest = parseManifest(data.js)
// Step 3: Compute MD5 for each source
for each sourceFile in [config, spatial, specs, schedule, directory, ...]:
currentHash = MD5(sourceFile)
storedHash = manifest.files[sourceFile.name].md5
if (currentHash !== storedHash) {
CHANGED_FILES.push(sourceFile)
}
// Step 4: Selective serialization
if (CHANGED_FILES.length === 0) {
return "NO_CHANGES_DETECTED"
}
// Step 5: Regenerate only changed sections
for each changedFile in CHANGED_FILES:
re_parse(changedFile) → update PROJECT_DATA[section]
update manifest.files[changedFile.name].md5
log("Updated: " + changedFile.name)
// Step 6: Write data.js
serializeToJS(PROJECT_DATA, API_KEYS, ACTIVE_ROLE, manifest) → data.js
updateGenerationLog()
return "UPDATED: " + CHANGED_FILES.length + " sections"
}
Fallback: Corrupted Manifest
If DATA_MANIFEST is invalid, the system falls back to full regeneration with validation:
- Parse all JSON source files fresh
- Validate schema against PROJECT_DATA structure
- Recompute all MD5 hashes
- Regenerate data.js
- Log warning in generation log
Extended reference: Detailed examples, templates, scoring rubrics, and best practices are in
references/skill-detail.md.