Document Summarizer Llama7b

Meet Document Summarizer Llama7b—your new best friend for turning dense documents into digestible nuggets of wisdom. Born from a caffeine-fueled, 12-hour hackathon sprint by a dynamic team (Amit, Adith, Saai, and Mithileshan), this tool dives deep into PDFs, images, Excel sheets, and CSV files with the prowess of a linguistic llama on a mission.

Not only does it summarize your documents with style, but it also comes equipped with a conversational Q&A interface. And with its nifty web extension, you can even chat with websites—making the internet spill its secrets just for you!

Features

• Document Processing: Extracts text from PDFs (using pypdf), images (via pytesseract and PIL), and tabular data from Excel or CSV files (with pandas).

• Conversational AI:
  Summarizes documents and generates key questions using a robust language model (powered by the Groq API and models like gemma2-9b-it).

• Interactive Interface: Runs on Streamlit, ensuring a smooth and engaging user experience.

• Automatic Charts: When uploading CSV or Excel files, the app displays quick visualizations for numeric columns.

• Web Extension:
  Interrogate websites directly to extract and summarize online content.

Installation Guide

1. Clone repository: git clone https://github.com/AmitSubhash/Document_Summarizer_Llama7b.git cd Document_Summarizer_Llama7b

2. Install dependencies:

pip install -r requirements.txt

Usage

Launch the interface: streamlit run app.py Supported operations:

1. Upload document (PDF/image/Excel/CSV)
2. Get auto-generated summary
3. Ask follow-up questions via chat
4. View quick charts for numeric columns in CSV/Excel files

Running locally

1. Ensure Python 3.10+ is installed.
2. Install the required packages: pip install -r requirements.txt.
3. Start the application with streamlit run app.py.
4. Upload a PDF, image, Excel or CSV file and interact with the chat interface for analysis.

Development

Contribution Welcome
Open to feature requests, bug reports, and PRs through GitHub issues.

License
Open-source (free for educational/personal use)

Team
Amit Subhash, Adith, Saai, Mithileshan - Luddy Hackathon 2024