PyXie Proxmox Operations

Browser-based Proxmox VE (PVE) operations console: inventory, historical metrics, findings/recommendations, capacity/rightsizing, backup/protection visibility, and a real write-capable maintenance/migration/placement engine, gated by an explicit Safety Contract on every mutating action.

This is not a read-only product. VM live migration was the first write capability added; the app now includes a full staged maintenance/migration/placement engine covering guest lifecycle, resize, node reboot, host package updates, node/full maintenance runs, cluster rebalancing, per-workload NIC VLAN reassignment, and PBS backup-job membership -- and, since v0.37, creating, changing, cloning and deleting VMs and containers, snapshots, tags (with Proxmox's own colors), defaults and templates, resource pools, an ISO library and NFS storage setup (see docs/guest-management.md), and a 3-2-1 backup scorecard (docs/protection.md) and Proxmox Backup Server management: status, jobs and patching (docs/adding-a-pbs.md), restoring a VM or container from any restore point (docs/restore.md), and one backup plan with copies, namespaces, pruning, garbage collection and verification (docs/retention-policies.md, docs/backup-namespaces.md). See the Safety Contract section below for exactly how every one of those is gated. What PyXie does not do (yet) is listed plainly in docs/limits.md.

Installation

PyXie runs as one Docker stack (web, API, worker, PostgreSQL, Redis and Caddy for HTTPS) on a single Ubuntu 24.04 VM.

1. Install it: follow docs/docker-caddy.md. It covers what you need first (VM size, network, certificate choice) and a step-by-step first-time setup, from git clone to creating your admin account.
2. Connect your Proxmox cluster: on first sign-in PyXie shows a New here? banner on the Dashboard that opens Platform > Setup guide, which walks the setup in 9 steps with live status (the last-but-two, optional, connects a Proxmox Backup Server), with the reference material (accounts, names, privileges) folded underneath. Written up in docs/setup-guide.md; installing the host wrapper on each node is in docs/host-kit.md and docs/adding-a-host.md.
3. Keep it current: docs/updates.md (Settings > Updates) and docs/patching-a-node.md.

Bugs, requests and feedback: Platform > Feedback in the app, or docs/feedback.md.

The Tasks panel, Cancel and the top bar are explained in docs/tasks-panel.md; emptying, restarting and retiring a host in docs/host-lifecycle.md; host settings and network in docs/host-config.md; guest settings in docs/guest-config.md; stuck guest locks in docs/stale-locks.md; workers and limits in docs/workers.md. A root shell on a host, opened from PyXie, in docs/host-shell.md.

The older native systemd install (no Docker) is still supported but is the legacy path: docs/INSTALL.md. Pick one per host.

Screenshots

Captured from the running app against a fictional demo dataset (no real hostnames, VM names or accounts). The dataset and capture scripts are in ops/demo/, so they can be regenerated each release.

Dashboard -- cluster resource usage, host load, and what needs attention right now:

Cluster balance -- the balance gauge and per-node host load (outer arc memory, inner arc CPU), with a link into Balance Load:

Rightsizing -- per-workload observation status with evidence-backed sizing suggestions:

Maintenance -- node evacuation, guest lifecycle, host updates and full maintenance runs, every action previewed before approval:

Balance Load -- node load and a suggestion up front, then a preview with memory per node before and after the plan (with real GB figures and which guests leave and arrive), the planned moves, and a short, expandable list of moves that were considered but not made (no gain, blocked by a rule, redirected), plus the guests left alone on purpose (pinned, Do not move, moved recently). A per-cluster Automatic balancing card can prepare plans for you (Recommend only) or approve and run them one move at a time (Auto-approve, off by default, with a Revoke button), on a schedule you choose (every 5 minutes to daily). Every check and preview first reviews the affinity rules and host pins and plans the moves that fix them before it balances load; every check is kept in a History with what it looked at, and a Dashboard banner shows when automatic balancing is on. See docs/balance-load.md:

Affinity Rules -- where every guest sits now, shown as coloured pills that link to the guest, and each keep-together / keep-apart rule in plain words with its current status (a soft preference that is not met shows amber, a broken hard rule red); tag rules suggest the tags in use and show which guests they cover; enforced on migrations, evacuations and rebalancing:

This is the beginning of the README. Read the rest at the original. Read the full README ↗