Skip to content

Stop notes part-way, and wait for Ollama after a reboot, and v0.33.0 - #52

Merged
nickcollinsdotco merged 1 commit into
mainfrom
stop-summary
Oct 14, 2026
Merged

nickcollinsdotco merged 1 commit into
mainfrom
stop-summary

Conversation

@nickcollinsdotco

Copy link
Copy Markdown
Owner

Backlog resource-safety #10. Nothing could stop a slow summary or re-pass, and an Ollama still starting after a reboot failed the notes outright.

  • Stop on the note's progress line and in the status bar's job list. Works queued, during the re-pass (checked per chunk), in the memory wait, and part-way through a summary part. Outcome::Stopped — not a failure, no toast.
  • The summary request is streamed over a plain socket (synthesis/stream.rs) read in 200 ms slices, so it can hang up at any moment; ureq could not be interrupted while Ollama loads. Shutting a cloned socket does not wake a blocked read on Windows (tried), hence the slices.
  • Measured with new examples/stop_check.rs on a 721-segment session: 1 ms to stop mid-part with Ollama logging cancel task; stopped mid-load, Ollama logs "aborting load" and nothing stays loaded. Full uninterrupted run still writes cited notes.
  • Notes wait up to a minute for an installed Ollama that is not answering yet.
  • Gallery: note-stopping, note-stopped; note-regenerating gains the button.

Not verified live: the Ollama wait (needs Ollama quit and restarted); Stop clicked in the real app (wiring covered by unit tests and the fake backend).

/verify green; hang-up test mutation-checked.

🤖 Generated with Claude Code

Nothing could stop a slow summary: each part was one blocking request of
up to ten minutes. The summary request is now streamed over a plain
socket read in short slices, so Stop hangs up at once, and Ollama
abandons a load or a generation when its client goes (measured with
examples/stop_check.rs: 1 ms mid-part, and "aborting load" mid-load with
nothing left in memory). The re-pass checks between chunks; the queue and
the memory wait check too. A stopped job is its own outcome, not a
failure. Notes wait up to a minute for an Ollama installed but not yet
answering, instead of failing on the two-second probe.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@nickcollinsdotco
nickcollinsdotco merged commit ae2094f into main Oct 14, 2026
2 checks passed
@nickcollinsdotco
nickcollinsdotco deleted the stop-summary branch October 14, 2026 21:35
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant