Products / Cockpit AI

Cockpit AI — Intelligent Server Operations Console

Cockpit AI brings CPU, memory, disk, network, process and container state from every server onto one console, and when anomalies occur, feeds metrics and logs to a local LLM to produce root-cause hypotheses and remediation advice. Its predecessor already runs in production at cockpit.u-eic.com on fully real data — this product is its multi-node, productized evolution. One engineer can see, diagnose and act on an entire fleet.

Open ConsoleEarly Access · Live

Core Capabilities

Unified Multi-Server Console

Lightweight agent probes collect real CPU / memory / disk / network metrics from every node. One console, monospace gauges, second-level refresh.

AI Anomaly Diagnosis

On anomaly, relevant metric curves and log snippets are fed to a local LLM, which returns root-cause hypotheses, blast-radius assessment and remediation steps — not just a threshold alert.

Process & Container Management

PM2 / Docker / systemd workloads in one view: status, restart counts, resource usage, with in-console restart and log inspection.

One-Click Inspection Reports

Daily/weekly automated inspections (disk watermarks, certificate expiry, slow queries, zombie processes) produce structured reports ready to share or archive.

Operation Audit Trail

Every restart and config change in the console is written to an audit log — who did what, to which machine, when. Traceable and exportable.

On-Premise Deployment

The whole system, LLM inference included, deploys inside the customer network. Ops data never leaves. Offline installer and upgrade channel provided.

Architecture

01Collection
Agent probes — lightweight per-server collectors for metrics / logs / process state
02Aggregation
FastAPI — metric aggregation, alert rules, diagnosis task scheduling
03Data
PostgreSQL — time-series metrics plus audit storage
04Diagnosis
Ollama local LLM — root-cause hypotheses and remediation; data never leaves the network
05Presentation
Next.js dashboard — fleet console / inspection reports / audit views

Tech Stack

  • Agent probes (Python, lightweight)
  • FastAPI aggregation service
  • PostgreSQL (time-series + audit)
  • Ollama local LLM diagnosis
  • Next.js dashboard
  • PM2 / Docker / systemd integrations

Pricing Direction

Early Access

Free onboarding, up to 5 nodes, in exchange for real-world feedback

Subscription

Per-node subscription (placeholder), including alerting and AI diagnosis quota

On-Premise License

Full in-network delivery plus annual maintenance (placeholder)

Current Stage

Stage: Alpha — the single-node version runs in UEIC production (cockpit.u-eic.com, all real data). M0 ✅ single-node real-data console live; M1 multi-node aggregation + AI diagnosis loop; M2 first external paying customer.

Request Cockpit AI Early Access

Tell us how many servers you run and your biggest ops headache — we prioritize onboarding accordingly.

Request Early Access