BudgetBotUK is an LLM-powered chatbot that leverages RAG architecture to provide answers to questions related to the Budget of Uttarakhand, India. Currently, It can answer all the questions related to the budget speech given on the floor of Vidhansabha of Uttarakhand in 2024.
Document used for RAG: https://budget.uk.gov.in/files/english_translate_23-24_bhashan.pdf
- langchain
- google PaLM
- FAISS
- Streamlit
- In the updated version all-MiniLM-L6-v2 model is used to create the embeddings and pinecone is used as a vector database.
Instructions: Add Pinecone and Google API keys in the .env file and add the pinecone index name in the utils.py
Although BudgetBotUK's responses are good, we can improve it further.
- Instead of GooglePalm, we can use other LLMs like LLama2, Mistral 7B, etc.
- We can add more documents related to budget to improve the semantic search results.
- Instead of FAISS, other vector databases like Pinecone can be utilized for this task.



