• 

  Home Screen

• 

  Register Screen

• 

  Login Screen

• 

  Voice Assistant Screen

• 

  User Dashboard

• 

  Backend and ML infrastructure

Shop-Assistant

Talk to your inventory.

Inspiration

Inventory management in retail is often a pain — handwritten notes, scanning barcodes, and messy spreadsheets. We wanted to create something simple, natural, and intuitive. The inspiration for Shop-Assistant came from imagining a world where retailers could just talk to their inventory system like they talk to their staff.

"Two apples, five bananas, and three cartons of milk."
— That’s all it should take to update your stock.

What It Does

Shop-Assistant is a smart voice-powered inventory companion.

• Retail staff or customers speak naturally to add or update items.
  
• The system transcribes audio → extracts structured meaning → updates stock.
  
• Updates are idempotent, meaning no accidental duplicates even if requests are retried.
  
• Events are published to RabbitMQ queue for downstream services.
  
• Everything is secured with JWT.
  

How We Built It

1. Mobile Client:

   • Captures audio streams or uploads.
     
   • Stores requests offline and syncs when back online.
     

2. Transcription Layer:

   • Uses Google Cloud Speech-to-Text for accurate multilingual transcription.
     

3. NLU Service:

   • Calls gptoss-20b with a strict JSON schema prompt.
     
   • Returns validated JSON with items, quantities, units, and intent.

Built With

• axios
• docker
• expo.io
• fastapi
• google-cloud-speech-to-text
• gptoss-20b
• jwt
• nestjs
• postgresql
• python
• rabbitmq
• react-native
• typescript