dannystocker/navidocs
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
navidocs/AGENT2_ACTION_PLAN.md

8.6 KiB
Raw Export PDF Permalink Blame History

Agent 2 - 4-Hour Countdown Action Plan

Mission: Get NaviDocs demo-ready for Riviera Plaisance presentation Deadline: ~4 hours from now Agent: Agent 2 - Current Functionality Tester

🎯 THE GOAL

Demonstrate working search functionality with real document data to show sticky engagement potential.

⚡ CRITICAL BLOCKER: Meilisearch Down

Status: Service running on StackCP only, not localhost Impact: Search feature broken (demo killer) Fix Time: 5 minutes

IMMEDIATE ACTION (START NOW)

# Option 1: Start Meilisearch locally (RECOMMENDED)
cd /tmp
meilisearch --http-addr 127.0.0.1:7700 \
  --master-key "5T66jrwQ8F8cOk4dUlFY0Vp59fMnCsIfi4O6JZl9wzU=" &

# Verify it started
sleep 2
curl http://localhost:7700/health
# Should return: {"status":"available"}

Expected Output:

{"status":"available"}

Time: 5 minutes (including startup delay)

📋 TIMELINE: What to Do When

PHASE 1: Setup (Next 10 min)

Goal: Get Meilisearch running + verify server connectivity

T+0:00  → Start Meilisearch (above command)
T+0:02  → Verify health endpoint
T+0:05  → CONFIRM: curl http://localhost:7700/health returns {"status":"available"}
T+0:10  → Meilisearch ready for indexing

Checklist:

  • Meilisearch started
  • Health endpoint responds
  • API server still running on :8001

PHASE 2: Test Data Creation (Next 20 min)

Goal: Create realistic test documents so demo has something to search

Problem: No test data exists currently

  • No sample PDFs
  • No sample images
  • No test entities in database

Action 1: Create Test Directory

mkdir -p /home/setup/navidocs/test-data/documents
mkdir -p /home/setup/navidocs/test-data/receipts

Action 2: Create Sample Test Files

Option A: Quick Workaround (5 min)

  • Use any PDF you can find locally
  • Or create a simple text document and convert to PDF
  • Focus on having SOME data to search

Option B: Proper Sample Documents (15 min)

  • Download real boat manual samples from YachtWorld
  • Create "Prestige 40 User Manual" (PDF)
  • Create "Engine Maintenance Log" (text → PDF)
  • Create 2-3 maintenance receipt images (JPG/PNG)

Sample Files Needed for Demo:

  1. prestige-40-owners-manual.pdf - Main document
  2. engine-maintenance-log.pdf - Searchable document
  3. receipt-engine-service-2025.jpg - OCR test image

Time: 15-20 minutes total

PHASE 3: Index Test Data (Next 15 min)

Goal: Populate Meilisearch index so search has real results

Step 1: Create Test Entities in Database

// Add to /home/setup/navidocs/test-data/seed-entities.js

const db = require('better-sqlite3')('./server/db/navidocs.db');

const orgId = 'org-demo-001';
const boatId = 'boat-prestige-40';

// Create organization
db.prepare(`
  INSERT INTO organizations (id, name, type, created_at)
  VALUES (?, ?, ?, datetime('now'))
`).run(orgId, 'Riviera Plaisance Demo', 'boat-dealer', Date.now() / 1000);

// Create boat entity
db.prepare(`
  INSERT INTO entities (id, organization_id, name, entity_type, make, model, year)
  VALUES (?, ?, ?, ?, ?, ?, ?)
`).run(boatId, orgId, 'Demo Prestige 40', 'boat', 'Jeanneau', 'Prestige 40', 2023);

console.log('Test entities created');

Step 2: Upload Documents Use the /api/upload endpoint:

curl -X POST http://localhost:8001/api/upload \
  -F "document=@/home/setup/navidocs/test-data/documents/prestige-40-owners-manual.pdf" \
  -F "documentType=manual" \
  -F "entityId=boat-prestige-40"

Step 3: Trigger Indexing Documents automatically index through BullMQ queue:

  • Redis triggers processing
  • OCR extracts text
  • Meilisearch indexes pages

Expected Result:

  • Documents appear in search results
  • Search for "engine" returns relevant pages
  • OCR confidence visible in results

Time: 10-15 minutes

PHASE 4: Verify Search Works (Next 10 min)

Goal: Confirm end-to-end search pipeline functional

Test 1: Health Checks

# API health
curl http://localhost:8001/health

# Meilisearch health
curl http://localhost:7700/health

# Check Meilisearch index
curl http://localhost:7700/indexes

Test 2: Search Query

curl "http://localhost:8001/api/search?q=engine&limit=5"

Expected Response:

{
  "results": [
    {
      "id": "page_doc-001_p1",
      "title": "Engine Maintenance - Prestige 40",
      "text": "Engine model: Volvo Penta D6...",
      "pageNumber": 1,
      "ocrConfidence": 0.95
    }
  ],
  "totalResults": 3,
  "processingTime": "42ms"
}

Test 3: Filter by Entity

curl "http://localhost:8001/api/search?q=propeller&entityId=boat-prestige-40"

Time: 5-10 minutes

PHASE 5: Create Demo Script (Next 15 min)

Goal: Have a polished demo sequence ready

Script: demo-search-walkthrough.sh

#!/bin/bash

echo "=== NaviDocs Search Demo ==="
echo ""
echo "1. Searching for 'engine maintenance'..."
curl -s "http://localhost:8001/api/search?q=engine+maintenance&limit=3" | jq .

echo ""
echo "2. Searching for 'propeller' in Prestige 40..."
curl -s "http://localhost:8001/api/search?q=propeller&entityId=boat-prestige-40" | jq .

echo ""
echo "3. Document statistics..."
curl -s "http://localhost:8001/api/stats" | jq .

echo ""
echo "=== Demo Complete ==="

Create file:

cat > /home/setup/navidocs/demo-search-walkthrough.sh << 'EOF'
# ... content above ...
EOF
chmod +x /home/setup/navidocs/demo-search-walkthrough.sh

Time: 10 minutes

📊 TIMELINE SUMMARY

T+0:00  → Start Meilisearch (5 min)
T+0:05  → Create test data (20 min)
T+0:25  → Index documents (15 min)
T+0:40  → Verify search (10 min)
T+0:50  → Create demo script (15 min)
T+1:05  → DEMO READY ✅

Remaining buffer: ~2 hours 55 minutes

🚨 CONTINGENCY PLAN

If Meilisearch still fails:

  1. Show API architecture without live search
  2. Demo database schema and document structure
  3. Show OCR pipeline code
  4. Explain search integration (diagram)
  5. Promise to deploy next week

If test data creation takes too long:

  1. Use pre-existing data from database
  2. Skip OCR testing (too time-consuming)
  3. Focus on search feature only

If upload endpoint fails:

  1. Directly insert test documents into database
  2. Skip upload demo, focus on search/retrieval
  3. Explain the full pipeline conceptually

📝 Success Criteria for Demo

✅ MUST HAVE:

  1. Meilisearch running and responding to health check
  2. At least 3 searchable documents in index
  3. Search query returns relevant results (e.g., "engine" finds engine pages)
  4. OCR confidence scores visible in results

✅ NICE TO HAVE:

  1. Document upload working end-to-end
  2. Filter by entity working (entity-specific search)
  3. Multiple document types (manual, receipt, spec)
  4. Clean terminal output (no error logs)

❌ CAN SKIP:

  1. Full OCR processing (takes 5+ min per document)
  2. Advanced search syntax (filters, facets)
  3. Multi-user authentication demo
  4. StackCP remote deployment verification

🔧 Troubleshooting Quick Reference

Problem: Meilisearch won't start

# Check if port 7700 in use
lsof -i :7700

# Kill old process
pkill meilisearch

# Start fresh
meilisearch --http-addr 127.0.0.1:7700

Problem: Search returns no results

# Check if index exists
curl http://localhost:7700/indexes

# Check if documents indexed
curl http://localhost:7700/indexes/navidocs-pages/stats

# Manually reindex
node /home/setup/navidocs/server/scripts/reindex-all.js

Problem: API not responding

# Restart server
pkill -f "node.*navidocs.*index.js"
cd /home/setup/navidocs/server && npm start

# Check port
lsof -i :8001

📞 Quick Reference Commands

Verify Meilisearch:

curl http://localhost:7700/health

Test Search API:

curl "http://localhost:8001/api/search?q=test"

Check Database:

sqlite3 /home/setup/navidocs/server/db/navidocs.db "SELECT COUNT(*) FROM documents;"

View Recent Errors:

tail -50 /tmp/navidocs-server.log

✅ Final Checklist (Before Demo)

  • Meilisearch running (curl http://localhost:7700/health)
  • API server running (curl http://localhost:8001/health)
  • Test documents indexed (at least 3)
  • Search query works (curl "http://localhost:8001/api/search?q=engine")
  • Demo script ready and tested
  • Terminal clean (no error logs visible)
  • Time check: Showtime in ~X hours

Agent 2 Report Generated: 2025-11-13 Time Budget: ~1 hour to prepare + ~3 hours buffer before showtime Next Agent: Agent 3 (UX/Sales Pitch) - Read this report to understand what's working/not working