How heytarshitgupta/ai-khata-book works
AI Khata Book turns spoken Hindi transaction details into persisted ledger records: the user speaks through the browser, the server asks Gemini to extract transaction fields, then saves them in Supabase and returns a response for the interface. The README also describes a dashboard for viewing and managing transactions. The frontend source is not included in the sampled files, so its exact request and dashboard data wiring cannot be verified; those links are omitted.
Main entry points: server.js Voice Endpoint · index.html Speech Recognition
Components 7
Ledger Interface
- Speech Recognition
index.html - Transaction Dashboard
index.html
Transaction Service
- Voice Endpoint
server.js - Field Extraction
server.js
External Integrations
- Gemini
- Supabase Ledger
Other components
- Ledger User
Connections 7
- Ledger User → Speech Recognition: speaks Hindi
- Speech Recognition → Voice Endpoint: submits text
- Voice Endpoint → Gemini: requests extraction
- Gemini → Field Extraction: returns fields
- Voice Endpoint → Supabase Ledger: inserts transaction
- Voice Endpoint → Transaction Dashboard: returns result
- Ledger User → Transaction Dashboard: views and manages