An air-gapped operational intelligence stack that forecasts service degradation, computes blast radius using dependency graphs, and assists operators with AI-powered incident analysis.
- Built for secure internal platforms, this stack runs locally, without any cloud dependency
- This stack can be used locally in a k8s cluster, by installing the helm chart of this stack
- Time-series database: Prometheus
- Vector database: QDrant
- Incident storage: PostgreSQL
- Backend services: FastAPI
- Local SLM: phi-3.5-mini
- Queue system: Kafka
- Basic ML worker, Aggregator worker flow done
- Next: SLM service
- Model training and custom dataset creation pending
