PDF-Question-Answering-Assistant

This project is a PDF Question Answering Assistant that utilizes OpenAI's API, Gradio, and a Postgres-backed Vector Database. It allows users to input a PDF URL and ask questions about its content, leveraging vector search to provide accurate answers.

Features

• Extracts knowledge from a PDF using its URL.
• Utilizes vector search with PgVector2 for efficient information retrieval.
• Stores chat history in a Postgres database.
• Provides a user-friendly web interface using Gradio.
• Integrates with OpenAI's API for natural language processing.

Setup

1. Run PgVector with Docker

docker run -d \
  -e POSTGRES_DB=ai \
  -e POSTGRES_USER=ai \
  -e POSTGRES_PASSWORD=ai \
  -e PGDATA=/var/lib/postgresql/data/pgdata \
  -v pgvolume:/var/lib/postgresql/data \
  -p 5532:5432 \
  --name pgvector \
  phidata/pgvector:16

This sets up a PostgreSQL instance with the pgvector extension on port 5532.

Installation

1. Clone the Repository:

   git clone <repository_url>
   cd <repository_folder>

2. Install Dependencies:

   pip install -r requirements.txt

3. Create a .env File:

   OPENAI_API_KEY=your_openai_api_key
   

---

Usage

1. Start the Gradio Interface

python main.py

2. Access the App

Open your browser and go to: http://127.0.0.1:7860

You can:

• Enter a PDF URL
• Ask questions about the PDF content

---

Key Features

• Extracts knowledge from a PDF URL using PDFUrlKnowledgeBase
• Stores and retrieves conversation history with PgAssistantStorage
• Uses PgVector for vector-based knowledge search
• Gradio interface for user interaction

---

Configuration

• Database URL:

  db_url = "postgresql+psycopg://ai:ai@localhost:5532/ai"

• Environment Variable:

  OPENAI_API_KEY=your_openai_api_key

---