NLP processor based expense tracker

1. Add a expense of rs 250 for ola I took to work Processes Query and adds into Db.

   

2. What is my total expense this month? Processes query, find relavant data rows and returns a natural response with total.

   

3. Categorical Search --> How much did I spend on food this month? Processes query, find categorically correct rows and returns a breakdown in natural language with total.

   

4. Hinglish support

   

Methodology/Pipleline

1. Natural language query to from whatsapp to backend via twilio.
2. Convert Natural language query to relavant sql query using LLM.
3. Hit Postgres with the sql query.
4. Use recieved Data as context, along with user query to LLM.
5. Get Natural Language Response.

API Reference

Get Response Chat

  GET /

Parameter	Type	Description
None	None	302 code to redirect to this repo

  POST /whatsapp

Parameter	Type	Description
body	string	Singular CRUD API for expense tracking via whatsapp

Future changes

1. Securing API, open currently to test and ship fast.
2. Better text to SQL models.
3. Semantic search on db rows, getting higher accuracy of fetching data. example coffee and kapi are essentially the same thing.
4. Backend is inactive, redirects / to the repo, only /whatsapp is active currently.