← Mini-Rook / Development diary

MINI-ROOK / BUILD LOG 032

The Raven Never Sleeps

Scheduled health checks, with recovery still awaiting a failure test

SCHEDULER PASS · RECOVERY UNVERIFIED. Adapted from the project development record BUILD_LOG_032.md. Results describe the recorded checkpoint; publishing this article adds no new runtime test.

TECHNICAL CHECKPOINT

What this build actually records.

Stack and method
PowerShell health monitor · five-minute Scheduled Task · listener checks
Recorded progress
Initial execution and task invocation show healthy core/gateway and LastTaskResult 0.
Still open at this checkpoint
No deliberate failure/reboot test; depends on user session, Python, Ollama and startup environment.

Software development on an existing workstation. This entry does not establish a new robot, custom hardware build or autonomous publishing capability.

TEST STATUS / SEPARATE FROM THE ORIGINAL RECORD

What later checks established.

Initial execution and task invocation show healthy core/gateway and LastTaskResult 0.

Still open: No deliberate failure/reboot test; depends on user session, Python, Ollama and startup environment.

01 / BUILD LOG 032

The original development record

Rook Keeper is an idempotent Windows PowerShell health and recovery monitor for the local Mini-Rook core and its private Secure Gateway. A Windows Scheduled Task named Rookweld-MiniRook-Keeper runs every five minutes under the logged-in user and checks listeners on ports 18765 and 18766. Where a listener is missing, the script attempts to launch the corresponding local Python script and records the status in runtime/keeper.log. The gateway's expected unauthenticated HTTP 401 is counted as a healthy response.

Verified on Fury: initial keeper execution reported core=True and gateway=True; task registration succeeded; manual Scheduled Task invocation completed with LastTaskResult 0 and logged another healthy result. No process was deliberately killed, so automatic restart after failure or reboot is NOT yet end-to-end verified. The creator explicitly accepted publication with this limitation rather than interrupting active work for a reboot; a future build log will document any failure or recovery defects honestly.

Caveats: the scheduled task depends on the Windows user session, configured Python launcher and local Ollama availability. It does not manage Tailscale itself. A fresh process will not automatically inherit an interactive terminal's temporary GitHub private-key environment variables; release scanning that requires such variables may need separate startup configuration. This is a supervised local pilot, not an unattended system service.

Status: READY TO PUBLISH — SCHEDULER ACCEPTANCE PASS; FAILURE RECOVERY AND REBOOT PERSISTENCE NOT YET VERIFIED. Chapter VIII — The Nest.

KEEP FOLLOWING THE BIRD

The notebook stays open.