AI Study Assistant

A multi-agent AI Study Assistant built with LangGraph as part of the SDAIA Academy – Building AI Agent Systems programme.

The system allows users to upload PDF study materials and interact with them through an AI-powered workflow that can answer questions, generate summaries, and create quizzes. It combines Retrieval-Augmented Generation (RAG), memory, Human-in-the-Loop interaction, and error handling to create a more reliable study experience.

Programme

• Author: Tala Alhadawi
• Programme: SDAIA Academy – Building AI Agent Systems
• Cohort: 26–30 July 2026

Key Features

• PDF document upload
• Multi-agent workflow using LangGraph
• FAISS-based Retrieval-Augmented Generation (RAG)
• Question Answering
• Automatic summarization
• Quiz generation
• Short-term and long-term memory
• Human-in-the-Loop quiz approval
• RetryPolicy and fallback error handling
• LangSmith tracing
• LangGraph Functional API

How It Works

The assistant follows a multi-agent workflow:

PDF → Document Processing → RAG Retrieval → Specialized Worker → Response

Depending on the user's request, the system routes the task to the appropriate worker:

• Question Answering Worker — answers questions using retrieved document content.
• Summary Worker — generates concise summaries from the study material.
• Quiz Worker — creates quizzes and requests human approval before proceeding.

The workflow also uses memory to maintain context and includes retry and fallback mechanisms to improve reliability.

How to Run

1. Open the notebook in Google Colab.
2. Add GROQ_API_KEY to Colab Secrets.
3. Add LANGSMITH_API_KEY to enable tracing.
4. Select Runtime → Restart session and run all cells.
5. Upload a text-based PDF.
6. Enter a study request.
7. Type yes or no when quiz approval is requested.

Reference

• SDAIA Academy GitHub

Technologies