Upgrading

Run package updates from the project and Python environment used by Greatminds. Inspect the available release first:

greatminds update --check

This queries PyPI without changing the package or project. Updates are operator initiated; the ACP daemon does not currently send periodic release notifications.

Moving to 3.0

The 3.0 release uses ACP for every managed executor. Define agent manifests and role bindings in coordination/execution.yaml, including the installed adapter and harness versions. Setup preserves task files and an existing execution configuration; it does not choose harnesses or credentials. Verify this contract before starting the daemon. Codex, Claude and Grok are the accepted harness scope.

The packaged web workspace starts with greatminds web --port 8765. It has no application authentication and binds to localhost by default.

Prepare active work

Inspect runs and stop new dispatch before changing the daemon's environment:

greatminds run status
greatminds run pause

Pause leaves existing work running. Let it finish or explicitly cancel selected runs before restarting. Uncertain command or deployment effects need the recovery procedures in the operations runbook. A daemon restart does not prove those effects safe to repeat.

Apply the update

greatminds update

The command selects the detected environment manager, upgrades the package, checks the installed version in a fresh Python process, and refreshes project bootstrap state. A major version increase requires --major. For an explicitly registered service project, use --project PROJECT_NAME.

Existing installed services have their units refreshed and receive systemctl --user try-restart. Inactive services remain inactive. Update skips missing service installations; install a service separately when you need one. For a uv tool installation, use uv tool upgrade greatminds. Restart the web server after upgrading. Independently managed background daemons also need an explicit greatminds daemon restart to load runtime changes; check active work first. Restart a foreground daemon manually using the updated environment. Harnesses and ACP adapters have their own installation and update procedures.

To repeat only the project and installed-service refresh after managing the package yourself:

greatminds update --post-pip

Verify and resume

greatminds daemon doctor --project-dir "$PWD" --json
greatminds project schema --check
greatminds run status
greatminds watchdog

Doctor performs static checks; exercise an actual ACP conversation to verify provider authentication and protocol behavior. Resolve reported holds before resuming automatic dispatch with greatminds run resume.

The effective installed schema governs runtime policy. .greatminds/schema.yaml is a diagnostic mirror, not a project policy override. Setup creates a missing mirror but preserves an existing one. If project schema --check reports drift, compare the mirror with greatminds project schema; after reviewing the change, replace the mirror with that command's output. Restart the daemon after changing the installed package or an explicit GREATMINDS_CANON_DIR. Existing runs retain their pinned contracts.