A Grounded Knowledge Assistant for Auditable Kernel Security & Neurodiverse Accessibility.
Sentinel-RAG is a specialized accessibility tool designed to reduce cognitive load for individuals—including those with ADHD, autism, and dyslexia—interacting with complex Linux Kernel telemetry. Built for the Microsoft Neurodiversity Challenge, it leverages Python-driven Machine Learning to transform dense, high-anxiety technical data into structured, calm, and actionable insights.
- Cognitive Load Reduction: Uses Python logic to parse raw eBPF logs into a structured, distraction-free TUI (Terminal User Interface).
- MLT-Driven Cognitive Compression: Our Python backend (
rai_audit.py) uses Machine Learning Text (MLT) logic to dynamically adjust information density. This prevents "Analysis Paralysis" by allowing users to choose the depth of information that matches their current cognitive energy. - Level 0 (Technical): Full breakdown of 7+ security targets (such as:
capabilities.7,bpf.2). - Level 2 (Summarized): Intelligent compression into a single "SYSTEM SUMMARY" to prevent information overload.
- Grounded Remediation Hints: Provides non-anxiety-inducing, step-by-step tasks to fix non-compliance issues.
- Visual Focus Zones: Dedicated spatial areas for Vision, Kernel Monitoring, and AI Coaching to prevent sensory overwhelm.
- Python (Core Intelligence): The heart of the system. Uses Python for the Responsible AI (RAI) backend to analyze kernel data and generate grounded responses.
- Microsoft RAI Toolbox: Integrated to ensure all AI outputs are safe, ethical, and supportive.
- Go (Golang): Provides the high-performance, accessible interface shell using Bubble Tea & Lipgloss.
- RAG Integration: Security hints are strictly grounded in authoritative kernel "Rulebooks" (
capabilities.7,bpf.2) to eliminate AI hallucinations.
Follow these steps to set up the Sentinel-RAG environment on your local machine.
- Python 3.10+
- Go 1.21+
- Linux/WSL2 Environment (Required for Kernel telemetry access)
- Clone the Repository
git clone [https://github.com/champbreed/sentinel-rag.git](https://github.com/champbreed/sentinel-rag.git)
cd sentinel-rag- Configure Python Environment
- Initialize the virtual environment to isolate Machine Learning dependencies.
python3 -m venv venv
source venv/bin/activate
pip install --upgrade pip
pip install pandas responsibleai- Inspect the RAI audit logic and groundness scoring (use nano, vim, or cat)
nano scripts/rai_audit.py- Verify MLT Compression Engine
- View detailed technical audit (Level 0)
python3 scripts/rai_audit.py --level 0- View AI-compressed system summary (Level 2)
python3 scripts/rai_audit.py --level 2- Launch the Accessibility TUI
- Start the interactive Go-based interface.
go run main.go🛡️ Safety & Ethics
Sentinel-RAG adheres to Microsoft's Responsible AI principles. It employs a Strict-Groundedness policy, ensuring the AI Coach only suggests security remediations verified against official Linux Kernel documentation, preventing harmful or misleading "hallucinations" in critical security contexts.
Privacy-First Design: Sentinel-RAG processes kernel telemetry locally via eBPF. Sensitive system data remains within the local environment, ensuring that security audits do not compromise system privacy.