AIOps Part 6: AI-Powered Operations

This is the sixth and final capstone of the AIOps learning path. Using the AIOPS2026 repository as the primary source, it builds a complete route from large-model capabilities to a production-style network operations platform. The course covers the OpenAI Responses API, Structured Outputs, Function Calling, File Search/RAG, the Agents SDK, and closed-source model boundaries. It then moves to open-source model selection and deployment with Ollama, Hugging Face, ModelScope, vLLM, SGLang, multimodal OCR, fine-tuning boundaries, and security models such as Qwen3Guard and ProtectAI.

The agent interoperability section introduces MCP servers and clients, TLS authentication, pyATS MCP services, and an A2A multi-agent case, followed by LangChain agents, Deep Agents, subagents, skills, MCP, NetClaw/OpenClaw, and Hermes. The final project implements an AI automation console for network operations: NetBox provides assets and topology, Supabase/InfraDB stores device facts and configuration snapshots, Airflow/Collector performs automated collection, Telegraf/InfluxDB/Grafana forms the metrics pipeline, Elasticsearch/LogAI provides logs and events, Qdrant supports RAG evidence retrieval, OpenClaw delegates work through multiple workspaces, skills, sub-agents, and MCP services, and Batfish snapshots provide deterministic, verifiable troubleshooting. The course concludes by validating six workflows: device inspection, metrics queries, log queries, troubleshooting, alert-event analysis, and configuration-change analysis.

Companion Git repository: https://git.qytang.com/qytadmin/AIOPS2026

Total preview duration: 1h 16m 36s (recorded as 77 minutes in the system).