Br0GPT

Progress :

Implemented

- question-answering bot without memory(emb. retiriever,LLM- )
- Feeded PDF docs, autoconversion to txt file
- used FAISS docs. store as Database
- used Flask for frontend
- input can be in text or audio form

Immediate Goals

- user feedback, based on which lesser accurate documents can be retreiverd
- uploading user documents i.e., extending our database
- Memory factor
- Local LLM 
-  API Calls  // already almost done

Research

- Studying different retrievers,readers,document stores and setting a standard for usage in differnet applications

Installation

Create conda environment (Optional)

Firstly, we will create a conda environment called transcription

conda create -n BROGPT 

Secondly, we will login to the transcription environment

conda activate BROGPT

upgrade the pip beforehand

pip install --upgrade pip

pip install libraries

pip install -r requirements.txt

Temporary running process

run the python files in order: 1.AutoPDFconversion.py - makes the supplied PDF in scrap directory into chunks 2.preprocessing.py - preprocesses the chunk files(cleaning , indexing , updating the embeddings) 3.answer_generator.py