renaissance2005/airm-guidelines

Guidelines for AI Risk Management

A Streamlit app that generates structured risk mitigation guidelines for AI systems based on contextual risk data stored in a Neo4j knowledge graph. It uses Groq’s hosted GPT-OSS 20B model (mixtral-8x7b-32768) for LLM-based narrative generation.

🚀 Features

• —Interactive tabbed interface for selecting application context, risks, and treatments.
• —Queries real-time data from Neo4j AuraDB.
• —Uses Groq API to synthesize plain-language implementation guidelines.
• —Dynamic integration of stakeholder roles and AI life cycle phases.

🛠️ Deployment Notes

• —SDK: Docker (this Space builds from Dockerfile)
• —Entry point: streamlit run app.py
• —Port: Binds to $PORT (automatically configured by Spaces)

🔐 Configuration

Set the following secrets under Settings → Variables and secrets:

• —GROQ_API_KEY: Your Groq API key for model access
• —NEO4J_URI: Your Neo4j AuraDB URI (e.g. neo4j+s://<id>.databases.neo4j.io)
• —NEO4J_USERNAME: Usually neo4j
• —NEO4J_PASSWORD: Your Neo4j database password

📷 Architecture Overview

[image]

This visualization connects stakeholders, AI life cycle phases, and risk mitigation actions in a unified graph.