Designing AI for Network Troubleshooting: A Langchain Proof of Concept
summary
Proof of concept ties a webhooked Grafana alert to a FastAPI service that spins up Langchain agents which call Python tools that use a PATS connector to query devices via NETCONF + Telegraf, store metrics in InfluxDB and surface results in WebEx. Demo targets an ISIS neighbor drop: the LLM (OpenAI) plans checks, queries device state, proposes bringing an interface up and can execute only after human confirmation. It generally troubleshoots but the live demo is flaky and required heavy prompt and tool docstring tuning.
AI-written summary · about
chapters (11)
- 0:00AI in network troubleshooting
- 3:13Audience and agenda
- 3:53Designing AI troubleshooting
- 8:27Why LangChain
- 12:43Network device tools
- 14:30Demo setup and flow
- 17:46Live troubleshooting demo
- 25:09Key lessons
- 27:37Next steps
- 28:21Future of AI troubleshooting
- 29:10Q&A
related talks
30mAI Agents for Incident Management
01h 59mBuilding Agentic and Multi-Agent Systems with LangGraph
44mBuilding an AI Agent on Top of Your API
40mAgents in Action: Advancing Open Source AI Missions
43mRedefining Human-AI Collaboration
01h 00mPrompt Failures and Latency Spikes: Observability for AI
28mAPIs and GenAI : Your Perfect Combo with LLM-Powered Frameworks.
01h 59mLLM applications with LangChain