This is a basic RAG chatbot and report generator made using LangChain, Streamlit, FAISS, Cohere's embed-english-v3.0 and Cohere's command-r

The project is deployed on streamlit. Visit and try from this link.

• You can upload multiple reports as PDF.
• Multiple indices can be added for better organization
• You can select more than one schemes and fields as input
• Generated report can be downloaded locally in CSV format for future references.
• For development purposes, you can see the chunks retrieved from the vector database for the specific query
• Additional Feature : There is a Chatbot as an option to generate your own personalized queries.

• Language : Python
• Libraries and Frameworks : LangChain, PyPdf, Tabula, Streamlit, Pandas
• Models: Cohere's embed-english-v3.0 and command-r
• Database: FAISS Vector Database

Worked with Python 3.11 anything above will probably work.

1. Clone the repo

git clone https://github.com/jojocoder28/Mutual_Fund_Chatbot

2. Create and activate virtual environment

cd Mutual_Fund_Chatbot
python -m venv .venv
.venv\Scripts\activate

3. Install Requirements

pip install -r requirements.txt

4. For the local machine you need to uncomment the import tabula and tabula.convert_into(uploaded_file[0], f"db/{index_name}/table.csv",pages='all', output_format='csv') in the 141st line of pages/Upload_Files.py

5. Create a .env file and put your Cohere API Key as COHERE_API_KEY and OpenAI API key as OPENAI_API_KEY

COHERE_API_KEY=[YOUR COHERE API KEY GOES HERE]

The chatbot uses Cohere's embed-english-v3.0 and command-r by default.

6. Run Chatbot.py

streamlit run .\Report_Generator.py

• Navigate to Upload Files in the sidebar to upload your own PDFs (make sure the PDFs are readable)

• Store the uploaded PDFs in a new or existing index.
• Navigate to Report Generator and select the desired index (An index for the year 2022 is already created).
• Select your scheme from the drop-down menu, or search in the search box

• Select the fields on which you want to generate a report on from the Field drop-down menu.
• Click on Generate. The report from your query will be generated in a tabular form.

• You can download the generated report in CSV format from the Download CSV File link.
• You can also use the Chatbot

• If you want to know the what chunks were sent to the llm to generate the report, click on the see chunks... drop down.

• Swarnadeep Das
• Sukanta Bala
• Areej Aasna