Andrew0x7/incidentiq

# IncidentIQ — Multi-Agent Incident Response ## 🚩 Pain Point Production incidents cost enterprises **$5,600/minute** on average (Gartner). Engineers spend 60% of incident time manually correlating logs across services, identifying root cause, and coordinating response. A single misdiagnosis can extend MTTR by hours, costing millions. ## 🧠 Multi-Agent Architecture IncidentIQ deploys **5 specialized agents** orchestrated by MiMo V2.5: | Agent | Focus | Techniques | |-------|-------|-----------| | 📋 Log Parser | Multi-source log ingestion & correlation | Pattern matching, temporal correlation | | 🔭 Trace Analyzer | Distributed tracing & latency profiling | Span analysis, cascade detection | | 🧠 Root Cause | Causal inference & dependency mapping | Bayesian networks, symptom filtering | | 🔧 Fix Generator | Auto-remediation & rollback plans | Template matching, risk scoring | | 🚨 Escalation | Severity routing & stakeholder comms | Blast radius, SLA tracking | ### Response Pipeline 1. **Ingestion** — Normalize logs from Docker, K8s, CloudWatch, application traces 2. **Analysis** — All 5 agents analyze in parallel → correlate findings 3. **Resolution** — Generate fixes → route escalation → produce report ## 🔧 Tech Stack - **Engine**: MiMo V2.5 (orchestration & natural language analysis) - **Log Parsing**: Multi-format parser (JSON, syslog, structured, unstructured) - **Tracing**: OpenTelemetry-compatible span analysis - **Frontend**: Vanilla JS (zero dependencies, instant load) - **Deployment**: GitHub Pages (static) ## 📊 Metrics - **14+ incident patterns** detected across 5 categories - **Severity scoring**: P0 (Critical) → P3 (Low) - **Timeline reconstruction** from raw logs - **Auto-fix suggestions** with risk assessment ## 🚀 Usage 1. Paste error logs, stack traces, or incident reports 2. Click **⚡ Analyze** 3. Review findings across Analyze, Timeline, Agents, and Pipeline views 4. Follow recommended fixes per finding ## 📄 License

标签：后端开发