Devotd Github
NewAI agents

beeline-work/beeline — Daily log work withdrawn, only the collectors remain

The 2026-11-07 commit in beeline-work/beeline deletes the work that produced day-by-day log reports along with the systemd timer, leaving only the raw log collectors for three apps.

Repositorybeeline-work/beeline ↗

The 2026-11-07 commit in beeline-work/beeline deletes the work that produced day-by-day log reports along with the systemd timer, leaving only the raw log collectors for three apps.

stars
★ 3
Forks
0
Language
TypeScript
License
MIT
Latest release
v0.1.74
Read original source ↗View on GitHub ↗

The latest commit in beeline-work/beeline is a cleanup that strips out the pipeline that had been producing day-by-day log reports, leaving only the raw log collectors. The 2026-11-07 commit (616286a, PR #4084) adds 24 lines across 6 files and deletes 92 lines. The root README also gains a single line explaining that the collector handles the raw logs and the daily work has been withdrawn.

What was deleted is scripts/log-walk/host-walk.sh and systemd's beeline-log-walk.service and beeline-log-walk.timer. These were the path that swept 24 hours of logs on the squire host and produced report.md, report.json, and source-sha, and the install script install-host.sh changed from taking an output directory argument to a form with no arguments, installing and enabling only the collector unit beeline-log-collector@<app>.service. The remaining collectors take the flyctl logs streams of the three apps beeline-server, beeline-free-agents, and beeline-voice-livekit and reconnect every 10 minutes, and the raw logs pile up at ~/.local/share/beeline-log-walk/raw/<app>/<UTC date>.log and are kept for 8 days. systemd brings dead collectors back up, and streams that ignore SIGTERM are force-terminated. In the environment file only PATH and the raw log path remain, and the creation of the state directory and the pinning of the node path were dropped along with them.

One question remains. The README paragraph deleted along with it stated that call socket rejections are not recorded in the logs at all and that HTTP rejections do not reach Sentry either, so the only place where repeated rejections could be examined was this work, and the replacement for that signal was written to be the database count of the work results. From the material of this commit alone, it cannot be confirmed whether that replacement has landed. In other words, raw log collection is retained and only the daily reports are gone, so it is too early to judge from the commit alone what repeated rejections and similar signals should now be read from. The repository description, license, and release information are not in the provided material.