Assessment Agent

The Assessment Agent is an intelligent AI system that autonomously processes educational content and generates contextually relevant assessment questions. Using advanced natural language processing and machine learning capabilities, it understands content, maintains knowledge in a vector database, and adaptively generates questions based on context and requirements. The agent offers two interactive interfaces: a terminal-based interface and a web-based interface powered by Streamlit.

Key Capabilities

1. Intelligent Content Processing

• Autonomous text extraction from various file formats (PDF, DOCX, TXT, HTML, CSV)
• Smart text chunking with context preservation
• Vector embeddings generation for semantic understanding

2. Knowledge Management

• Vector database storage for efficient content retrieval
• Intelligent caching of processed content
• Metadata tracking for content updates

3. Adaptive Question Generation

• Context-aware question creation
• Multiple difficulty levels and question types
• Semantic search for relevant content
• Quality assurance through understanding-focused questions

4. Interactive Interfaces

• Real-time response to user inputs
• Session state management
• Progress tracking and feedback

Prerequisites

• Python 3.8 or higher
• A reliable internet connection (required for API calls and model downloads)

Setup

1. Environment Configuration

Create a .env file in the project root with the following keys:

• LLM_API_URL: The URL for the LLM API.
• LLM_MODEL_NAME: The model name used by the LLM API.

Example .env file:

LLM_API_URL=your_model_chat_api_url
LLM_MODEL_NAME=chat_llm_model_name

2. Dependency Installation

Install all required dependencies using pip:

pip install -r requirements.txt

3. Data Directory Setup

The project retrieves input files from the data directory. By default, a sample file (react_basics.docx) is available.

To use custom data: Remove react_basics.docx from the data folder and add your own files. Supported formats: PDF, DOCX, TXT, HTML, and CSV.

Running the Agent

You can run the Assessment Agent in two ways:

1. Terminal Interface

For a command-line experience, execute:

python assessment.py

Features:

• Interactive command-line interface
• Process files from the data directory
• Generate questions with specified difficulty and type
• Option to generate multiple sets of questions

2. Web Interface

For a user-friendly web interface, execute:

streamlit run chat.py

Features:

• Modern web-based UI
• Drag-and-drop file upload
• Visual progress tracking
• Option to use existing processed data
• Download generated questions as text files
• Maintain history of generated question sets
• Download individual sets or all questions at once

Logging

The application logs events both to the console and to the assessment_agent.log file located in the project root. This log file provides detailed debugging information and runtime insights.

Project Structure

assessment_agent/
├── assessment.py         # Terminal interface implementation
├── chat.py              # Web interface implementation
├── config.py            # Configuration settings and environment variable loader
├── requirements.txt     # Project dependencies
├── .env                # Environment configuration file
├── data/               # Directory containing input files
│   └── react_basics.docx # Default sample file (remove to use your own data)
└── README.md           # This file

Additional Information

• Model Initialization: Uses Hugging Face's Transformers library with a default model (bert-base-uncased) for embeddings.
• Database Storage: Embedded vectors and related metadata are managed via LanceDB.
• Rate Limiter: In-built mechanism to manage and throttle requests to the LLM API, ensuring adherence to API limits.
• Web Interface: Built with Streamlit for a responsive and intuitive user experience.

Interface Comparison

Feature	Terminal Interface	Web Interface
File Processing	From data directory	Drag-and-drop upload
Question Generation	Interactive prompts	Form-based input
Question History	Single session	Persistent across runs
Download Options	Display only	Text file export
Progress Tracking	Text-based	Visual progress bar
Multiple Question Sets	Yes	Yes, with history

For further details on configuration and troubleshooting, consult the inline comments and documentation within the source code.