Skip to content

DEMO_PREP_CHECKLIST.md

Latest commit

 

History

History
278 lines (203 loc) · 6.84 KB

DEMO_PREP_CHECKLIST.md

File metadata and controls

278 lines (203 loc) · 6.84 KB

Quick Action Checklist - Grey Area Improvements

Priority: Focus on what matters most for your M.Tech presentation

✅ Immediate Fixes (Do This Week)

1. Fix LLM Timeouts (30 mins)

# Edit .env.local
LLM_TIMEOUT_COLD=120000   # Increase from 90000 to 120000
LLM_TIMEOUT_WARM=90000    # Increase from 45000 to 90000

# Restart services
./start_local.sh restart

Why: Prevents random timeout errors users are seeing

2. Add Basic Tests (2-3 hours)

# Install pytest
pip install pytest pytest-asyncio pytest-mock httpx

# Create test directory
mkdir -p backend/tests

# Write 3 critical tests:
# 1. Test health endpoint works
# 2. Test query endpoint returns response
# 3. Test embedding generation works

# Run tests
pytest backend/tests/ -v

Why: Demonstrates code quality, catches bugs before demo

3. Fix CORS for Production (10 mins)

# backend/main.py
ALLOWED_ORIGINS = [
    "http://localhost:3000",
    "https://yourdomain.com",  # Add your production URL
]

app.add_middleware(
    CORSMiddleware,
    allow_origins=ALLOWED_ORIGINS,  # Change from ["*"]
    allow_credentials=True,
    allow_methods=["GET", "POST", "DELETE"],
    allow_headers=["Content-Type", "Authorization"],
)

Why: Security best practice, shows production awareness

4. Add Metrics Endpoint (1 hour)

# Install Prometheus client
pip install prometheus-client

# Add to backend/main.py
from prometheus_client import Counter, Histogram, generate_latest
from fastapi.responses import PlainTextResponse

query_counter = Counter('rag_queries_total', 'Total queries')
query_duration = Histogram('rag_query_duration_seconds', 'Query duration')

@app.get("/metrics", response_class=PlainTextResponse)
async def metrics():
    return generate_latest()

# In ask endpoint, add:
query_counter.inc()
with query_duration.time():
    # ... existing query logic

Why: Shows observability knowledge, impress evaluators

🎯 For Demo Day (If Time Permits)

5. Add Document Management UI (3-4 hours)

  • Show list of uploaded documents
  • Allow deleting documents
  • Display document metadata (chunk count, upload date)

Why: Makes system more impressive and usable

6. Remove Console Errors from Frontend (1 hour)

// Replace all console.log/error with proper logging
// Or just wrap in if (process.env.NODE_ENV === 'development')

if (process.env.NODE_ENV === 'development') {
  console.log('Debug info');
}

Why: Cleaner demo, looks more professional

7. Add Error Boundary to Frontend (30 mins)

// frontend/src/components/ErrorBoundary.js
// Wrap <App /> in ErrorBoundary
// Shows "Oops!" message instead of crash

Why: Graceful failure handling impresses judges

📊 For Presentation (No Code Required)

8. Create System Architecture Diagram

  • Draw how components connect
  • Show data flow: User → Frontend → Backend → LLM/Milvus → Response
  • Include in presentation slides

Why: Demonstrates system understanding

9. Prepare Performance Metrics

# Measure and document:
# - Query response time (yours: ~8-10s vs Docker: 60-90s)
# - Memory usage (before/after)
# - Throughput (queries per minute)
# - Accuracy metrics (if possible)

Why: Data-driven results impress evaluators

10. Document Challenges & Solutions

  • List problems faced (Docker timeouts, Node.js localStorage, etc.)
  • Explain solutions implemented
  • Show learning and problem-solving skills

Why: Demonstrates technical depth and resilience

🚫 Skip These (Not Worth Time Now)

  • ❌ Full authentication system → Demo with API keys is fine
  • ❌ Complete Confluence API integration → Local mode works
  • ❌ Streaming responses → Cool but complex
  • ❌ Analytics dashboard → Nice but not critical
  • ❌ Multi-language support → Out of scope
  • ❌ A/B testing → Over-engineering for demo

📅 Suggested Timeline

Today (2-3 hours)

  • Fix LLM timeouts
  • Add metrics endpoint
  • Remove console errors

Tomorrow (3-4 hours)

  • Write 5 basic tests
  • Fix CORS config
  • Add error boundary

Before Demo (1-2 hours)

  • Create architecture diagram
  • Collect performance metrics
  • Practice demo flow

🎤 Demo Script Suggestion

  1. Introduction (1 min)

    • "Enterprise RAG system for knowledge management"
    • "Runs locally on Mac M4 Pro with 8-10x performance improvement"

  2. Architecture Overview (2 min)

    • Show diagram
    • Explain: Frontend → Backend → Ollama (LLM) + Milvus (Vector DB)

  3. Live Demo (3-4 min)

    • Show frontend UI
    • Ask a question: "What is the agile workflow?"
    • Highlight fast response time (~8-10s)
    • Show source citations with relevance scores
    • Open metrics endpoint (show Prometheus metrics)
    • Check health endpoint (show all services green)

  4. Technical Highlights (2 min)

    • Metal GPU acceleration (explain ARM64 optimization)
    • Vector similarity search (explain embeddings)
    • Document ingestion (show 14 sample docs loaded)

  5. Challenges & Solutions (2 min)

    • Docker performance issues → Local deployment
    • LLM timeout errors → Increased timeouts, retry logic
    • Node.js localStorage → Environment variable fix
    • Show test coverage (if implemented)

  6. Q&A (2-3 min)

    • Be prepared to explain:
      • Why Milvus over other vector DBs
      • Why Mistral model
      • How RAG differs from plain LLM
      • Scalability considerations

🎯 Success Criteria for Demo

  • ✅ System starts without errors
  • ✅ Query returns in <15 seconds
  • ✅ Source citations displayed
  • ✅ Health checks pass
  • ✅ No visible console errors
  • ✅ Can answer at least 3 different questions
  • ✅ Metrics endpoint shows data
  • ✅ Architecture diagram is clear

💡 Pro Tips

  1. Have backup demo recording in case live demo fails
  2. Test your demo flow 3+ times before presentation
  3. Prepare for "What would you improve?" question (reference IMPROVEMENT_AREAS.md)
  4. Know your metrics cold (response time, accuracy, document count)
  5. Practice explaining RAG to non-technical audience

📞 If Something Breaks During Demo

Backend won't start:

./start_local.sh stop
./start_local.sh clean
./start_local.sh start
# Wait 2 minutes

Query times out:

  • "This query is complex, let me try a simpler one"
  • Have 2-3 pre-tested queries ready

Frontend crashes:

Nothing works:

  • Show architecture diagram
  • Walk through code structure
  • Explain design decisions

Bottom Line: Focus on fixes 1-4 (LLM timeouts, tests, CORS, metrics) + demo preparation. Skip everything else for now. You can always mention "future improvements" in Q&A.

Time Budget: ~8-10 hours total for a solid demo-ready system.