Developed for czela.net (a non-profit community ISP in the Czech Republic with ~400 active infrastructure devices and 2,000 member connections) this project bridges real-world physical network reality with NetBox SSoT.
Standard telemetry tools (SNMP, LLDP, FDB, OSPF) often fail due to incomplete data and unmanaged hardware cascades. Our hybrid agent combines standard sync code with a local quantized LLM. It generates practical
recommendations for volunteer network admins to fill documentation gaps, assists in local troubleshooting during outages, and preserves complete data privacy without cloud dependencies.
Rather than replacing existing monitoring software, our project introduces a **Hybrid Auto-Discovery & Reliability Agent**. The core logic is **hardware-agnostic** and designed to run local LLMs (currently prototyped with Ollama/mistral-nemo).
We are interested in evaluating **Axelera Metis** as a potential low-power edge accelerator for our network distribution nodes.
1. Interactive Admin Co-Pilot (Filling the Documentation Gap)
When traditional sync scripts hit incomplete data, the agent doesn't guess blindly. It generates concrete recommendations for volunteer admins (e.g., *"Detected 5 active CPEs on Site X, but no local switch is documented. Is an unmanaged switch present? Please verify or update monitoring data."*). Once admins update the source system, traditional scripts finish the sync cleanly.
2. Local Device Querying & Root Cause Analysis
The agent accesses local device APIs (read-only) to cross-reference live state against NetBox during network issues, suggesting potential root causes and spotting similar single-points-of-failure (SPOFs) across the graph.
3. Outage Resilience & Zero-Cloud Privacy (The Cherry on Top)
* **Data Privacy:** Network topology, IP schemes, MAC tables, and member details remain strictly within the local network.
* **Offline Operation:** Internet outages are when network engineers need local tools the most. Running LLMs locally ensures the auto-discovery and diagnostic agent remains **100% operational** even when WAN connections are down.
