Talk to your shopping assistant like never before! DealLens AI is a real-time conversational shopping companion powered by Gemini 2.0 Flash Live API.
Think "Alexa for shopping" but smarter – speak naturally about products you're considering, and get instant voice responses with visual deal comparisons. DealLens transforms how you discover better prices and make purchasing decisions.
🎤 Natural Voice Conversations: Low-latency, interruptible shopping discussions
🛒 Real-time Price Discovery: "I see AirPods for £249 at Apple Store" → Get instant alternatives
👁️ Visual Deal Cards: See price comparisons and savings at a glance
🔊 Seamless Audio: Crystal clear responses with no breaking or distortion
↩️ Smart Interruptions: Change your mind mid-conversation, just like talking to a human
📱 Clean UI: Icon-only interface with visual state indicators
When shopping in-store, users often wonder if they can find the same product cheaper elsewhere. Manually checking websites is time-consuming and breaks the shopping experience.
- Users speak to the agent:
"I see a PS5 for £500 at Sony store. Can you check for cheaper options?"
- The agent parses intent and searches deals.
- Agent responds via voice and visual cards:
"Amazon has it for £469 and Argos for £479. You could save £31."
Planned enhancements include camera-based product recognition.
Technologies: Python FastAPI, WebSocket, Gemini Live API, Web Audio API, Vanilla JS
Audio Processing: Unified 24kHz sample rate, seamless chunk scheduling, persistent microphone streams
Data Storage: JSON-based deal database for rapid prototyping
Category: Live Agents with multimodal Voice + Visual output
Complete system architecture showing real-time voice processing, Gemini Live API integration, and deal search capabilities
Recent Technical Achievements:
- 🎵 Voice Breaking Eliminated: Seamless audio scheduling prevents gaps between response chunks
- ⚡ 50% Faster Response: 3-second silence detection (down from 10s)
- 🎤 No Permission Re-requests: Persistent microphone streams improve UX
- 📊 Web Audio API Compliant: Power-of-2 buffer sizes (8192) for optimal performance
Production Backend: https://deallens-backend-553067044467.us-central1.run.app
Quick Start:
- Open the frontend locally: Download & serve frontend files
- Frontend automatically connects to cloud backend
- Click microphone, grant permissions, and start talking!
The production backend runs on Google Cloud Run with enterprise-grade reliability.
cd app/backend
pip install -r requirements.txt- Visit Google AI Studio
- Create a free API key
- Create
app/backend/.envfile:
GEMINI_API_KEY=your_key_herecd app/backend
python main.pyOpen: http://localhost:8000/static/index.html
Frontend automatically detects local vs cloud backend
- Check "Connected!" status appears
- Click the microphone button
- Grant microphone permissions when prompted
- Say: "I found AirPods for £249, can you find cheaper?"
The backend is deployed on Google Cloud Run for global accessibility:
- URL:
<> - Region:
us-central1(Iowa, USA) - Scaling: Auto-scaling 0-100 instances
- Security: API keys in Secret Manager
- Monitoring: Cloud Run metrics and logging
See Server Documentation for full deployment guide.
- Server Documentation: Backend architecture, Gemini Live API integration, local development, and Cloud Run deployment
- Frontend Documentation: Web Audio API implementation, WebSocket communication, conversation flow, and browser compatibility
- Product Architecture: Business overview and product requirements
- System Architecture: High-level system design and technical decisions
Price Comparison:
🗣️ "I found iPhone 15 Pro for £999 at Currys, can you find it cheaper?"
🤖 "Amazon has it for £949 and Very for £969. You could save £50 with Amazon!"
Product Discovery:
🗣️ "What's the best deal on gaming headsets under £100?"
🤖 "Great question! I found the SteelSeries Arctis 7 for £89 at Game, down from £159!"
Smart Interruptions:
🗣️ "Actually, I meant wireless headsets instead"
🤖 "Got it! For wireless, the Sony WH-1000XM4 is £279 at John Lewis..."
- User speaks into the microphone on the web page
- Gemini Live API handles transcription & intent parsing in real-time
- Deal Search Tools query the product database for price comparisons
- User receives synchronized voice response + visual deal cards
- No microphone access: Ensure HTTPS is used (or localhost). Check browser permissions in Settings.
- Voice breaking/distortion: Clear browser cache and reload. Try Chrome/Edge for best compatibility.
- Microphone not working: Test microphone with other apps. Check device isn't muted.
- "Disconnected" status: Ensure backend is running on
localhost:8000. Check terminal for errors. - WebSocket errors: Try refreshing the page. Check firewall isn't blocking port 8000.
- API errors: Verify Gemini API key is valid and has Live API access enabled.
- Recommended: Chrome, Edge (full Web Audio API support)
- Limitations: Safari may have microphone permission issues
- Mobile: Works best on mobile Chrome/Safari with HTTPS
"Microphone permission denied": Grant permissions in browser settings"Not connected to backend": Start the Python backend server"Audio processing failed": Check Web Audio API support in browser console
Phase 1 ✅ Voice-first shopping assistant with seamless audio
Phase 2 🔄 Vision-enabled product recognition via camera
Phase 3 📋 Enhanced deal database with real-time pricing APIs
Phase 4 🤝 Multi-retailer integrations and purchase capabilities
This project is licensed under the MIT License. See the LICENSE file for details.
DealLens AI is developed by Nakul Bageja to explore practical applications of conversational AI in shopping. Contributions, suggestions, and feedback are welcome via Issues or Pull Requests.
Disclaimer: Product prices and availability are for demonstration purposes using sample data. This is a proof-of-concept showcasing Gemini Live API capabilities, created during Gemini Live Agent hackathon
Ready to revolutionize your shopping experience? 🛍️ Start talking to DealLens AI today!