Question Answering Using, Langchain + ChromaDB + llama2 + Question Answering Evaluation

A functional Question Answering project developed using:

1. Langchain building RAG based question answering system using chromadb and llama2

2. ChromaDB as vector database for storing the documents.

3. llama2 llm for answering the question using the top matching answers retrieved from chromadb.

4. Evaluating llama2 model

Currently using this application to query documents related to cord19 dataset. Modify the configs if you want to use your own data.

)

Installation instructions:

1. Install Python Version 3.11.8
2. Install Dependencies by running pip install -r requirements.txt
3. Download Llama2 model from https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML/resolve/main/llama-2-7b-chat.ggmlv3.q5_0.bin?download=true

How to run it the tool:

1. Create a empty folder name chroma_db_covid_data in the project folder.
2. Updating configs- Note config files are in path config/config_yml
   • Update chromadb_path key in the chromadb_config.yml with the chroma_db_covid_data folder path you have created just now
   • Update model_path key in inference_config.yml with the model file path where you have loaded llama2 model.
3. Run fe_pipeline.py file, this will create chromadb
4. Run app_main.py file -> In the termial run python -m streamlit run app_main.py --client.showErrorDetails=false