Infrapilot-AI

An AI-powered infrastructure orchestration agent that detects, investigates, and remediates hybrid-cloud network failures autonomously — with a human-in-the-loop approval gate before any corrective action is taken.

What it does InfraPilot AI monitors a hybrid cloud + on-premises topology and, when a fault is detected, runs a structured investigation loop powered by Qwen (qwen3.7-max). The agent:

Calls read-only diagnostic tools in sequence (app health → tunnel → BGP → firewall → link quality → service health) Identifies the root cause from evidence — not assumption Proposes a single corrective action and waits for human approval Executes the approved remediation and verifies recovery Records the full incident lifecycle to a durable audit trail All of this streams live to a React dashboard over SSE.

Architecture ┌─────────────────────────────────────────────────────────────┐ │ Cloud ECS (Alibaba Cloud, ap-southeast-1) │ │ ┌──────────────┐ ┌──────────────┐ ┌──────────────────┐ │ │ │ business-app │ │ orchestrator │ │ dashboard API │ │ │ │ FastAPI+PG │ │ qwen3.7-max │ │ FastAPI + SSE │ │ │ └──────────────┘ └──────────────┘ └──────────────────┘ │ │ WireGuard tunnel (wg0) │ └─────────────────────────────────────────────────────────────┘ │ ┌─────────────────────────────────────────────────────────────┐ │ On-prem ECS (Containerlab host) │ │ ┌──────────────────────────────────────────────────────┐ │ │ │ node-a (FRR + WireGuard) ──eth1── node-b (Postgres)│ │ │ └──────────────────────────────────────────────────────┘ │ └─────────────────────────────────────────────────────────────┘ Three connectivity layers must all be healthy for the app to work:

WireGuard VPN tunnel: cloud ECS ↔ node-a eBGP session over tunnel: cloud FRR ↔ node-a FRR (advertises 10.10.0.0/24) Data path: cloud app → node-a FORWARD chain → node-b Postgres :5432

Built With

• containerlab
• python
• shell
• terraform
• wireguard