connect-schedule

name	connect-schedule
description	Wire the brain's scripts to run on a schedule — pull, ingest, lint, backup — so it stays current on its own. Use after at least one source is connected and the user wants hands-off refresh.

Connect: Schedule (run the scripts on their own)

You're turning on the automation so the brain refreshes without the user lifting a finger. Do this once at least one source is connected (/connect). The user is not technical — explain each part in one line, and get an explicit yes before you schedule anything.

What the scripts do

The scripts/ folder is the machinery. The schedule runs two routines plus a backup loop:

• morning-routine.sh (8am) — pulls fresh data into raw/ (runs every pull-<source>.sh, including the pull-drive.sh / pull-github.sh etc. the connect playbooks set up).
• evening-routine.sh (9pm) — the full loop: pull, ingest new raw/ into the wiki (auto-ingest.sh, headless), lint, then commit and push a backup.
• auto-commit.sh — commits and pushes every few minutes even when Obsidian is closed, so nothing is ever lost.

1. Schedule them (macOS)

One command installs all three as launchd agents (Sun–Thu, 8am and 9pm; no sudo, no Full Disk Access):

./scripts/install-launchagents.sh

Say this before you run it: the installer also turns on auto-commit, which pushes once the moment it loads and then every 5 minutes. So make sure there's nothing sitting in the folder they don't want backed up to GitHub yet. Then get the yes.

On Linux, point cron at scripts/morning-routine.sh and scripts/evening-routine.sh instead.

2. Confirm they're loaded

launchctl list | grep businessbrain

Three entries (morning, evening, autocommit) means they're scheduled. Logs land in /tmp/kb-morning-routine.log and /tmp/kb-evening-routine.log.

3. Test without waiting for the clock

Kick one off now so they see it work:

./scripts/evening-routine.sh
# or via launchd:
launchctl kickstart -k gui/$UID/com.businessbrain.kb-evening

A note on overnight ingest

The evening routine ingests headless — it calls the agent (claude -p, or codex exec) to turn new raw/ into wiki pages while they sleep, and pushes the result straight to main. That's the point of a self-maintaining brain. It needs the agent logged in for non-interactive use; if it isn't, the step skips itself with a message and the rest of the routine (the backup) still runs. They can always run /daily by hand instead.

Close

Tell them in two lines what now runs on its own and when. The brain is now updating itself. They can re-run /connect any time to add another source.