https://github.com/renaldiangsar/pdf-summarizer-qa
PDF Text Summarization and QA app built using FastAPI, Langchain, and Streamlit
https://github.com/renaldiangsar/pdf-summarizer-qa
fastapi langchain langchain-python large-language-models stramlit
Last synced: about 2 months ago
JSON representation
PDF Text Summarization and QA app built using FastAPI, Langchain, and Streamlit
- Host: GitHub
- URL: https://github.com/renaldiangsar/pdf-summarizer-qa
- Owner: renaldiangsar
- Created: 2025-02-27T07:02:11.000Z (over 1 year ago)
- Default Branch: master
- Last Pushed: 2025-02-27T07:51:34.000Z (over 1 year ago)
- Last Synced: 2025-02-27T10:43:09.314Z (over 1 year ago)
- Topics: fastapi, langchain, langchain-python, large-language-models, stramlit
- Language: Python
- Homepage:
- Size: 61.5 KB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# PDF Summarization & QA Web App
## 📌 Overview
This project is a **FastAPI and Streamlit-based web application** that allows users to:
- **Summarize PDF documents** using an LLM-powered summarization model.
- **Ask questions** about the content of a PDF and receive relevant answers.
## 🚀 Features
- Upload a **PDF document** (one-time upload for both summarization and QA).
- Generate **different summaries** every time you run summarization.
- Perform **detailed summarization** for more insightful results.
- Ask **questions related to the PDF** and get precise answers.
- Uses **LangChain**, **Hugging Face embeddings**, and **FAISS** for retrieval.
- Frontend built with **Streamlit** for a smooth user experience.
---
## 🛠️ Tech Stack
- **Backend:** FastAPI, LangChain, Groq API, FAISS, Hugging Face embeddings
- **Frontend:** Streamlit
- **PDF Processing:** PyPDFLoader
---
## 🏗️ Installation & Setup
### **Clone the Repository**
```sh
git clone https://github.com/renaldiangsar/PDF-Summarizer-QA.git
cd PDF-Summarizer-QA
```
### **Create a Virtual Environment & Install Dependencies**
```sh
# open command prompt and run
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
pip install -r requirements.txt
```
### **Run the Backend (FastAPI)**
```sh
# open command prompt and run
uvicorn serve:app --reload # or just
# python serve.py
```
> The FastAPI server will start at `http://127.0.0.1:8000`
### **Run the Frontend (Streamlit)**
```sh
# open command prompt and run
streamlit run client.py
```
> The Streamlit app will open in your browser at `http://localhost:8501`
### Don't forget to give your api in .env file
- open .env file an set your groq and huggingface api
---
## 🔄 Workflow (How it Works?)
1. **User uploads a PDF** (file is stored temporarily).
2. **User selects:**
- "Summarize" → Calls FastAPI `/summarize/` endpoint to generate a summary.
- "Ask a Question" → Calls `/ask/` endpoint with the query to get a response.
3. **FastAPI processes the request** using:
- LangChain for text processing
- FAISS for document retrieval (for QA)
- Groq / Hugging Face models for LLM responses
4. **Response is displayed** on the Streamlit UI.
---
## 🛠️ Customization & Improvements
- Modify the **summarization prompt** in `serve.py` to change summary length/detail. Because shorter summarization will run faster.
- Adjust the **chunk size** in `RecursiveCharacterTextSplitter` for better retrieval.
- Use a **different LLM model** (e.g., GPT-4, LLaMA, or local models) for customization.
- If you want to do a lot of use, you can use paid Openai API.
---
## 📝 Future Enhancements
- Add **multilingual support** for summarization & QA.
- Implement **document summarization history**.
- Support **multiple PDFs at once**.
- Looking better option to PDF processing, because PyPDFLoader not give a optimal results for unclean/irregular pdfs
---
## Visual
---
This is my first project in github, there are still many shortcomings. I hope i can do better in my next project. 🎉