⚖️ Inscribe.AI : Document Search & Synthesis Tool

A highly efficient AI tool for legal professionals to rapidly search, sort, and synthesize knowledge from massive document sets. This project demonstrates an advanced Retrieval-Augmented Generation (RAG) architecture using Groq for real-time, low-latency performance combined together using Langchain.

---

✨ Core Features & Technical Highlights

This application moves beyond basic RAG by incorporating agentic principles and robust data management.

• 🧠 Agentic Retrieval (Multi-Query): Uses the Groq LLM to decompose complex user questions intelligently (Query Division) into multiple sub-queries, ensuring comprehensive context is retrieved from the VectorDB, leading to more accurate answers.
• 🎯 Contextual Compression: Implements an LLMChainExtractor (a form of Re-ranking) to filter out irrelevant information retrieved by the Multi-Query step, ensuring the final Groq model only sees the most pertinent chunks.
• **📄 This document parsing solution utilizes PyPDF and PDFMiner for efficient text extraction from digital PDFs, and intelligently falls back to Tesseract OCR for handling scanned or handwritten documents.
• 🔒 Persistent & Lifecycle Management: Data is stored securely and locally in ChromaDB. It includes an 7-Day Automatic Buffer to manage data lifecycle by deleting old documents in a separate, scheduled background process.
• ⚡ Low-Latency Synthesis: Leverages the Groq API (llama-3.1-8b-instant) for blazing-fast answer generation.

---

🛠️ Project Architecture

The application is structured into two main, independently running processes for maximum resilience:

File / Component	Purpose	Functionality
app.py	Frontend	Streamlit UI for file upload and querying.
ingest.py	Ingestion Pipeline	Handles file reading, encryption, chunking, embedding, and storage.
search.py	Agentic RAG Engine	Contains the Multi-Query Retriever, Contextual Compression, and the Groq LLM chain.
scheduler.py	Background Process	Runs continuously to automatically delete documents older than 7 days.
utils.py	Utilities	Contains file encryption (Fernet) and the robust Hybrid PDF Parser (PyMuPDF + pytesseract).

---

⚙️ Setup and Installation

Prerequisites

1. Python 3.10+
2. Tesseract OCR Engine: Must be installed separately on your operating system to enable the handwritten document feature.
3. Poppler (for Windows/Linux): Required for PyMuPDF image rendering if Tesseract is used.

Steps

1. Clone the repository:

   git clone [YOUR-REPO-URL]
   cd [YOUR-REPO-NAME]

2. Create and activate a virtual environment:

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

3. Install dependencies:

   pip install -r requirements.txt

4. Configure Environment Variables (.env file): Create a file named .env in the root directory and add your API key and a secret key:

   GROQ_API_KEY=your_groq_api_key_here
   FERNET_KEY=your_fernet_key_here
   # Optional: CHROMA_PERSIST_DIR=./chroma_db
   

---

🚀 How to Run the Application

You must run the frontend and the data lifecycle scheduler in separate terminals.

1. Start the UI

Run this command in your first terminal:

streamlit run app.py