Lumbridge Compute

Governed compute for unified-memory AI hardware — the machines where CPU and
GPU share one pool and there is no separate VRAM allocation to bounce off.
Over-commit that pool and the box thrashes and wedges, SSH and ping included,
before the OOM killer gets a turn.

Compute does not run inference. It supervises the servers that do:

- Admission control. A model starts only if committed + requested + margin
  fits the budget. The refusal is the feature.
- A 1 Hz watchdog on MemAvailable that stops the newest model before thrash.
- Scenes: named sets of models activated as one transactional unit, with
  pre-flight validation and rollback to the previously active Scene on
  failure. Scenes reference model ids, never weight paths or commands, so a
  Scene obtained from elsewhere cannot introduce code.
- Process ownership bound to (boot_id, pid, start_time_ticks, pgid == pid),
  so a reused PID can never be group-killed.
- A protocol-transparent TCP gateway, so clients keep one address while model
  runtimes move behind it.
- An MCP server, so agents drive the node as tools rather than as a CLI.

One binary, six direct dependencies, no async runtime outside the MCP surface.

Published from the internal monorepo with a fresh history. The private
development tree keeps its own history; nothing here carries it.

systemd/lumbridge-compute-agent.service

@@ -0,0 +1,26 @@
+[Unit]
+Description=Lumbridge Compute resident Scene supervisor and model gateway
+After=network-online.target
+Wants=network-online.target
+
+[Service]
+Type=simple
+# A dedicated unprivileged service account. Create it with:
+#   sudo useradd --system --home-dir /var/lib/lumbridge --create-home lumbridge
+User=lumbridge
+Environment=HOME=/var/lib/lumbridge
+Environment=LUMBRIDGE_COMPUTE_ROOT=/var/lib/lumbridge/compute
+# The gateway binds loopback by design. It has no authentication and no TLS, so
+# listening on 0.0.0.0 would publish every model on this node to the whole
+# network. To reach it from elsewhere put a reverse proxy or an overlay network
+# in front of it, rather than widening this bind.
+ExecStart=/usr/local/bin/lumbridge-compute agent --listen 127.0.0.1:8011 --upstream 127.0.0.1:8001 --floor 3
+Restart=always
+RestartSec=2
+# Model process groups are lifecycle-owned by Compute and must survive an agent
+# binary restart; the agent will revalidate their boot/start identities on resume.
+KillMode=process
+TimeoutStopSec=15
+
+[Install]
+WantedBy=multi-user.target