RT-GIDS v1.0.0

First tagged release of RT-GIDS: a real-time network intrusion detection system with an XGBoost detector, a FastAPI backend and a Next.js threat dashboard.

Highlights

• Real-time detection and streaming over WebSocket, with a live threat dashboard
• SHAP-based model explainability and model insights pages
• JWT-protected API with rate limiting and Redis ACL support
• Runs in CPU-only mode with in-memory state, or against a real Redis
• CI on every push: 33 backend tests, frontend typecheck and production build, Docker image build

Fixed since the last CI run

• Frontend: 19 TypeScript errors that failed the typecheck (missing Button xs size, loosened API types, a ?? vs || bug in log heartbeat detection)
• Backend: feature-distribution percentile keys are now strings, matching the schema under pydantic 2.13
• Backend: POST /api/model/shap now returns the documented 503 with message and fallback when the model artefact is missing
• Backend: the AI summary job read pydantic models as dicts, so it always returned summary_text: null; it now generates the summary
• Added the MIT LICENSE file and removed placeholder links from the README

See the README for setup, architecture and the model metrics.