Tasks runner
Run package.json scripts from the editor with output streamed into a panel.
- Desktop
The tasks runner lets you run repetitive build / lint / test commands from the editor without dropping into a terminal. Output streams into a docked panel and any diagnostics it emits feed into the Problems panel.
What gets discovered
Today the runner discovers tasks from a workspace’s top-level package.json scripts block. Each script becomes one task, labeled with the script name.
Glueprint picks the package manager from packageManager in package.json. If that field isn’t set, it falls back to npm. If the chosen binary isn’t on your PATH, the run surfaces a recoverable “Command not found” error so you can install it (or set packageManager) and retry.
Makefile, justfile, and a project-level task-config file are slated to slot in behind the same discovery API in a later release. Workspace packages are also future scope — only the root package.json is considered today.
Running a task
- Open the palette —
Cmd/Ctrl+Shift+B. Pick a task from the list. - Re-run the last task —
Cmd/Ctrl+Alt+B. Skips the palette. - Re-run from the panel — the output panel has a re-run control on every previous run.
The output panel
When a task runs, the Tasks panel docks at the bottom of the editor. It shows:
- Per-run tabs — each run gets its own tab so you can flip back to a previous run’s output.
- Streaming output — stdout and stderr appear live. The panel auto-follows the tail unless you scroll up; it snaps back when you reach the bottom again.
- Run status — a badge per tab: running, exited (with exit code), failed, cancelled.
- Re-run — restart the task with the same configuration.
- Stop — send a stop signal to the running process.
The panel is resizable. Close it with the x in its header; reopen it from any task action.
Problem matchers
Some tasks emit compiler / linter output that maps to source locations (e.g. tsc, eslint). Glueprint runs those through problem matchers and feeds the resulting diagnostics into the Problems panel. Click a problem to jump to the file and line.
preLaunchTask
The debugger can chain a task before launching. If a debug configuration declares preLaunchTask, the runner runs that task first and only starts the debugger if the task exits cleanly. See Debugger.