DEMO_CHECKLIST.md

EasyClaw Demo Day Checklist

30 Minutes Before Demo

• Run: bash scripts/prewarm.sh <your-railway-url>
• Open http://localhost:5175 (or Railway URL) — confirm landing page loads
• Click "Try Demo" → confirm guide loads in ~20 seconds
• Test mic: start an interview, speak a sentence, confirm transcript appears
• Confirm Cmd+Shift+D resets the app cleanly

Vapi Dashboard (do once)

• Open https://dashboard.vapi.ai
• Open your EasyClaw assistant → Edit → System Prompt
• Paste contents of backend/vapi_system_prompt.md
• Save and test with a short call

Railway Environment Variables (verify set)

• ANTHROPIC_API_KEY (or leave empty for Gemini fallback)
• GEMINI_API_KEY ✓
• VAPI_ASSISTANT_ID ✓
• VAPI_PUBLIC_KEY ✓
• GUIDE_OUTPUT_DIR=/data/guide_output ← must point to a Railway Volume mount
• GUIDE_STORE_PATH=/data/guide_store.json

Demo Flow (3 minutes)

1. "EasyClaw turns a 2-minute voice conversation into a complete, personalized AI setup guide."
2. Click Start Interview → talk for 60 seconds (name, business, pain point, tech level, hardware, channel)
3. End interview → watch 20-second demo golden path → guide appears
4. Scroll through guide sections → show quality score
5. "Setup consultants charge $500/hour. We do this in minutes."

CI Before Demo

• Run: make test-e2e-demo — all pass
• Run: make test-real-api (requires ANTHROPIC_API_KEY set)
• Confirm guide quality score ≥ 7/10 in test output

Guide History (show during demo)

• Visit /guides to see all generated guides — shows business names + timestamps
• Demo line: "We've already generated 47 setup guides in beta"

Fallback Plan

• If Vapi fails: click "Try Demo" → select a demo persona → golden path still works
• If backend is slow: the demo stream endpoint is instant, no real generation needed
• Nuclear option: demo-restaurant guide is always pre-cached and available offline