Follow-up to #197 / PR #295.
The library now wires pause resume end to end (executeWithResume, PauseOrchestrator), but the CLI still exits when a run pauses and prints a --resume hint. Two gaps to close for the CLI story:
-
No durable log from the CLI. Nothing in cli.ts creates an executionLog, and replaying past a pause is log-based — so even --resume <execId> on a pause-suspended run re-executes from scratch and re-pauses. A --durable [dir] flag (default .reqon-data/log) creating a FileExecutionLog would make --resume actually work for pauses.
-
No stay-alive mode. With --webhook --durable, a paused run should keep the process alive and go through executeWithResume (or a PauseOrchestrator) so an inbound webhook or the deadline resumes it in-process — matching what the docs describe for the library.
Also worth considering: the scheduler/daemon mode could start a long-lived PauseOrchestrator over the shared log so paused scheduled missions resume across runs.
Follow-up to #197 / PR #295.
The library now wires pause resume end to end (
executeWithResume,PauseOrchestrator), but the CLI still exits when a run pauses and prints a--resumehint. Two gaps to close for the CLI story:No durable log from the CLI. Nothing in
cli.tscreates anexecutionLog, and replaying past a pause is log-based — so even--resume <execId>on a pause-suspended run re-executes from scratch and re-pauses. A--durable [dir]flag (default.reqon-data/log) creating aFileExecutionLogwould make--resumeactually work for pauses.No stay-alive mode. With
--webhook --durable, a paused run should keep the process alive and go throughexecuteWithResume(or aPauseOrchestrator) so an inbound webhook or the deadline resumes it in-process — matching what the docs describe for the library.Also worth considering: the scheduler/daemon mode could start a long-lived
PauseOrchestratorover the shared log so paused scheduled missions resume across runs.