AGH Regulations Assistant

Image generated by AI

Overview

AGH Regulations Assistant is a tool designed to help users quickly find and interpret regulations and policies at AGH University. It leverages AI to provide accurate, context-aware answers to queries about university rules, procedures, and documentation.

Features

• Natural language understanding of regulation queries
• Fast and relevant responses
• Maintains message history for context-aware conversations
• Detects toxic messages and responds appropriately
• Recognizes queries about Jagiellonian University (UJ) and responds appropriately by clarifying that it is an assistant for AGH University
• Each response is evaluated and receives a rating

Installation

Note: The project was developed using Python 3.12.

1. Clone the repository:

   git clone https://github.com/michalszc/agh-regulations-assistant.git
   cd agh-regulations-assistant

2. (Optional) Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

Usage

Run the assistant with:

python src/main.py [OPTIONS]

Parameters

• --index_path <path>: Path to the FAISS index (default: faiss_index)
• --model_name <name>: Name of the LLM model to use (default: gemini-1.5-flash-8b)
• --model_provider <provider>: Provider for the LLM model (default: google_genai)
• --k <int>: Number of top documents to retrieve before filtering (default: 5)
• --threshold <float>: Minimum relevance score for document retrieval (default: 0.5)
• --delta <float>: Maximum score drop-off allowed from the top score (default: 0.05)
• --token-budget <int>: Maximum total tokens allowed, including the prompt (default: 1000)
• --max_history_tokens <int>: Maximum number of tokens to keep in the message history (default: 500)
• --show-graph: Whether to show the graph of the RAG pipeline (flag; no value needed)

Configuration

Before running the assistant, create a .env file in the project root directory with your Google API key:

GOOGLE_API_KEY=your_google_api_key_here

This key is required for accessing the LLM model via the Google GenAI provider. Alternatively, you can use a different LLM model and provider by specifying the --model_name and --model_provider options when running the assistant.

Example

python src/main.py --index_path faiss_index --model_name gemini-1.5-flash-8b --model_provider google_genai

License