⚡ DocInferX

Local AI Document Intelligence Engine

📘 Overview

DocInferX is an offline-ready RAG (Retrieval Augmented Generation) system that lets you upload PDFs or images, extracts the text automatically (OCR + PDF parsing), indexes it using FAISS vector search, and allows you to chat with your documents using a local LLM (Phi-2).
The project is built for privacy-focused document intelligence: fast, local, and completely offline.

---

⭐ Features

• Upload PDF or Images (PNG/JPG/JPEG)
• Automatic OCR using PaddleOCR
• Smart text chunking & cleaning
• FAISS Vector Search for fast recall
• Local LLM (Phi-2) for answering queries
• Matrix rain cyber UI
• Streamlit Frontend
• Docker support for easy setup
• Runs fully offline
• Document Library view

---

🛠 Technologies & Tools Used

• Python 3
• Streamlit — UI framework
• FAISS — vector database
• Sentence Transformers — embeddings
• Phi-2 / HuggingFace Transformers — LLM
• PaddleOCR — OCR engine
• PyPDF2 / pdfreader — PDF parsing
• Docker — for containerized deployment

---

🔧 Installation & Run Guide

1. Clone the repository

git clone https://github.com/shekh-2810/DocInferX.git
cd DocInferX

2. Create a virtual environment

python3 -m venv venv
source venv/bin/activate      # Linux/macOS
venv\Scripts\activate         # Windows

3. Install dependencies

pip install -r requirements.txt

4. Run the app

streamlit run streamlit_app.py

5. Open on web browser

http://localhost:8501

🐳 Docker Setup

1.Build the image:

docker build -t docinferx .

2.Run the container:

docker run -p 8501:8501 docinferx

🧪 Testing Instructions:

1.Open the application in your browser.

2.Upload any PDF or image.

3.Wait for OCR + indexing to complete.

4.Open the Chat tab.

5.Ask questions related to the uploaded document.

6.Compare answers with the source document to verify accuracy.

Screenshots

Upload Page

Completion of upload

Sidebar

Results

Library

👤 Maintainer

Developed by Shashank Shekhar Choudhary.

⭐ Support

If you find this project useful, consider starring the repo!

👉 https://github.com/shekh-2810/DocInferX