Ai Compliance Advisor

An AI assistant that provides a list of regulations and guidelines to follow when developing artificial intelligence software, with a particular focus on computer vision. The system answers specific questions such as "What regulations must I comply with for a facial recognition system in Europe?" or "Are there guidelines for ethics in the use of computer vision?" It uses a multi-agent architecture: the Query Agent retrieves the most relevant documents from a vector database built with embeddings, while the Compliance Agent processes the retrieved content and generates a precise answer using a Gemini LLM.

Features

• Document retrieval using vector embeddings.
• AI-based response generation.

Installation

1. Clone the repository.
2. Install dependencies: pip install -r requirements.txt.
3. Add your API keys (Gemini API) to a .env file like this GEMINIKEY=
4. Run the app locally: streamlit run app.py.

Technical Stack

• Language Model: Gemini-1.5-flash-latest
• Frameworks: LangChain, Streamlit
• Database: ChromaDB (vector database for document embeddings)
• Embeddings: HuggingFace Sentence Transformers

Developed By

This project was developed by Keita Vigano and Sara Borello. We are a master degree student of Data science of Bicocca

Built With

• pytohn