This project builds a pragmatic “AI analyst” on top of existing monitoring. It ingests events, clusters similar ones, and generates clean, structured summaries for humans.
Goals
- Cut alert noise so responders see patterns quickly
- Keep output auditable with consistent, structured mail formats
- Stay portable (Docker) and simple to operate
Highlights
- Event grouping with light‑weight embeddings and windowed correlation
- Summarization with guardrails and deterministic prompts
- Delivery via email with links to originals and run metadata
Tech
- n8n for orchestration and schedules
- Ollama for on‑box inference
- Python helpers for parsing and scoring