LLM Inference: HuggingFace

Dedicated endpoints

Messaging: Redis

A2A queues + cache

Automation: Scrapli

Async SSH/NETCONF

UI: Streamlit

Dashboard (port 8501)

Orchestration: Docker

7 services

Network Lab: Azure EVE-NG

3 Cisco routers

Parsing: TextFSM + Genie

CLI extraction

Observability: Grafana

Metrics + traces

Deployment

• Docker on single host
• GPU for LLM inference
• SSH tunnel to Azure

Targets

• LLM: <3s (p95)
• MCP: <5s (p95)
• MTTR: <5min