Awesome AI for Infra › Full-Stack AIOps Platforms

Kkkirito-123/Mutil-Rag-Agent

⭐ 133 Python added to this list on 2026-07-27 repository created 2026-05-05

The Multi-Agent AIOps Platform V3 is designed for OnCall and SRE scenarios, transforming user-provided fault descriptions or Alertmanager alarms into structured tasks. It intelligently selects relevant 'skills' and leverages Retrieval-Augmented Generation (RAG) with Milvus vector search and custom MCP tools to gather evidence. The platform then generates traceable Markdown reports for incident diagnosis. Version 3 enhances the platform with 'fast' and 'deep' dual-mode diagnosis, a Redis Streams-based queue for background workers, a Postgres event-based fact database, an event center, and an approval and permission structure. It also includes an LLM Wiki, RAG evaluation, and concurrency testing capabilities. Key features include skill-first diagnosis, where the system first determines resource type (e.g., host, network, container) or general OnCall playbook before narrowing down tools. The 'fast' mode follows a Plan-Execute-Replan flow for quick diagnosis of single-cause issues, while the 'deep' mode uses specialized parallel agents (MetricAgent, LogAgent, InfraAgent, RunbookAgent) for cross-verification of evidence in complex incidents. The system supports various RAG techniques, including Parent-Child chunking, Milvus retrieval, BM25, RRF fusion, and optional re-ranking. It integrates with MCP tools for system, web search, Windows logs, network, and Docker operations, and incorporates permission boundaries, tool metadata, guardrails, and manual approval for enhanced control. The project is primarily for learning and entry-level reference rather than production deployment, with a focus on demonstrating Agent workflow, RAG implementation, and skill routing.

https://github.com/Kkkirito-123/Mutil-Rag-Agent

AIOpsRAGmulti-agentincident managementtroubleshootingLangGraphMilvusPrometheusSREOnCallIT operationsanomaly detectionroot cause analysisautomation