Transforming Financial Insights with AI

An AI-Powered Financial Document Analysis Tool

Overview

This is a powerful tool/approach designed to analyze and extract insights from company annual reports

Prerequisites

• Python 3.8 or higher
• Required Python libraries (specified in requirements.txt)
• Annual report files in PDF format

Clone Repository

git clone https://github.com/shoryasethia/FinChat.git
cd FinChat

Install Dependencies

pip install -r requirements.txt

How to Use

1. Add Annual Reports

Place the Annual Report (.pdf) of a company in the data/pdfs/ directory.

2. Run Scripts

a. Using Qwen Model

python app.py

b. Using Gemini Model

In .env, add your GOOGLE_API_KEY, if using gemini.py

python gemini.py

• Output: Results will be saved in the output/ directory.

3. Knowledge Graph Interaction

a. Graph Response with Streamlit Interface

This launches a Streamlit application for a user-friendly Q&A chatbot experience.

streamlit run main.py

NOTE: To switch off debug mode, set level = "info" in .streamlit/config.toml

b. Graph Response in Terminal

Run the following command for graph-based terminal responses. Ensure you edit the query and select the appropriate mode of search (local, naive, global, hybrid) as per your requirements.

python graph-rag.py

Project Structure

FinChat/
|-- .gitignore
|-- .streamlit
   |-- config.toml
   |-- secrets.toml
|-- app.py
|-- context_reduction.py
|-- csv-combine.ipynb
|-- csv2txt.py
|-- data
   |-- pdfs
      |-- 3M_2018_10K.pdf          # Replace this with your desired file
|-- gemini-mask.py
|-- gemini.py
|-- graph-rag.py
|-- images-response.py
|-- LICENSE
|-- lightrag
   |-- base.py
   |-- lightrag.py
   |-- llm.py
   |-- operate.py
   |-- prompt.py
   |-- storage.py
   |-- utils.py
   |-- __init__.py
   |-- __pycache__
|-- main.py
|-- model_loader.py
|-- output
   |-- extracted_tables
      |-- 3M_2018_10K
         |-- data.csv
         |-- table_page_8_box_1.png
         |-- table_page_13_box_2.png
         |-- ...
   |-- images
   |-- knowledge_graph
      |-- graph_chunk_entity_relation.graphml
      |-- kv_store_full_docs.json
      |-- kv_store_llm_response_cache.json
      |-- kv_store_text_chunks.json
      |-- lightrag.log
      |-- vdb_chunks.json
      |-- vdb_entities.json
      |-- vdb_relationships.json
   |-- parquet
      |-- extracted_texts_and_imgs.parquet
      |-- extracted_texts_and_imgs_with_responses.csv
   |-- txts
      |-- final_output.txt
|-- pdf_extractor.py
|-- README.md
|-- requirements.txt
|-- tables-response.py
|-- utils_logging.py
|-- __pycache__

Project Rationale

Check out Project Report for details.

Demo

Check out streamlit demo

To view results of to-html

These code files are to create an .html page of past 10 years annual reports for faster analysis and data excess. This html page shows the annual report and have download buttons to download tables in excel format.

Page Url : http://103.42.51.135/drereddy.html

License

This project is licensed under the MIT License. See the LICENSE file for details.

Support

For any issues or questions, please open an issue in the GitHub repository.