aiyiyi121/sxdevops

What it solves

SxDevOps addresses the fragmentation of information and actions in traditional IT operations. Instead of manually switching between separate systems for alerts, logs, traces, and change records, it consolidates these capabilities into a single, AI-driven workflow. It transforms the operational process from "searching across systems" to a streamlined flow of observing status, gathering evidence, querying the system, and confirming actions.

How it works

The platform integrates four core layers:

1. Observability Fact Layer: Aggregates alerts, metrics, logs, and traces from sources like Prometheus, Grafana, and SkyWalking to provide evidence for the AI agent.
2. Event Center: Records execution results, critical write operations, and failure localization clues for post-mortem analysis.
3. Task Center: Manages host inspections, batch commands, script templates, and scheduled tasks.
4. AIOps Agent: Uses an LLM for understanding and planning. It employs a controlled orchestration chain: an Action Router identifies the task type, an Agent Mode (Direct, ReAct, or Plan + ReAct) determines the reasoning method, and a Preflight check ensures permissions and risk assessment before execution.

Crucially, the model only generates candidates and plans; all critical actions must pass through backend RBAC (Role-Based Access Control) and require manual human confirmation before execution.

Who it’s for

IT operations (Ops) engineers and SREs who need to reduce the cost of troubleshooting and consolidate fragmented operational tools into an AI-assisted diagnostic and execution platform.

Highlights

• Controlled AI Execution: Prevents LLM "hallucinations" or unauthorized actions by requiring human confirmation for all write/execute operations.
• Comprehensive Integration: Built-in support for Kubernetes, Docker, SSH, Prometheus, Grafana, and various tracing/logging tools.
• Comprehensive Audit: Every session, tool call, preflight check, and execution result is logged for auditing and learning.
• Structured Evidence: Uses "Skills" and SOPs to constrain the AI's output, ensuring conclusions are based on structured facts rather than free-form generation.