A multi-agent system that ingests network telemetry (logs, SNMP, config diffs, topology) and produces cited, ranked root-cause hypotheses for network incidents — cutting NOC triage time from tens of minutes to seconds.
🚧 Early build — Milestone 1 (ingestion pipeline) in progress. See docs/architecture.md for the full design and ROADMAP.md for milestones.
NOC engineers spend a large share of incident-response time on triage: correlating alerts across tools, pulling device configs, checking change history, and writing the summary — before any actual fix happens. NetRCA automates that triage step, with every claim traceable back to source evidence.
See docs/architecture.md for the full system diagram, agent breakdown, and design rationale (ADRs in docs/ADRs).
Stack: FastAPI · LangGraph · PostgreSQL · Redis · Qdrant (RAG) · React/TypeScript · Docker · Kubernetes · GitHub Actions · OpenTelemetry/Langfuse
- Log/config/telemetry ingestion pipeline
- RAG retrieval over runbooks and past incidents
- Multi-agent RCA orchestration (LangGraph)
- Evaluation harness with golden dataset (accuracy, faithfulness, latency, cost)
- Streaming incident timeline UI
- Full observability (traces, cost dashboard)
- Kubernetes deployment + CI/CD
git clone <your-repo-url>
cd netrca
docker compose up(Docker Compose setup lands in Milestone 7 — see roadmap.)
| Layer | Tech |
|---|---|
| Orchestration | LangGraph, LangChain |
| API | FastAPI |
| Data | PostgreSQL, Redis, Qdrant |
| Frontend | React, TypeScript |
| Infra | Docker, Kubernetes, GitHub Actions |
| Observability | OpenTelemetry, Langfuse |
MIT — see LICENSE