Raise default budgets to high-but-bounded; release 0.13.1#7
Merged
Conversation
The previous defaults were too conservative for the long-running workloads autosentry exists to babysit (ML training, multi-hour data pipelines): a healer that needed ~15 attempts to land a kept fix would hit `max_restarts=10` and give up, while a stuck-loop healer could still spam the API hourly under `max_attempts_per_detector_per_hour=5`. Both knobs are now wide enough that productive workloads never feel capped, narrow enough that broken healing eventually stops. | knob | was | now | |-------------------------------------------------|-------|--------| | process.restart_policy.max_restarts | 10 | 50 | | healing.budget.max_attempts_per_detector_per_hour | 5 | 60 | | healing.budget.max_wall_seconds_per_incident | 600 | 7200 | Also reframes `max_restarts`: it's not a *budget* — `state.restarts` zeros on every kept fix, so this only trips when the healer can't land a kept fix for N restarts in a row. Documented as such. `healing.escalate_to_claude_after` decouples from `max_restarts // 5` to a literal `2`. Previously, raising max_restarts pushed Claude escalation later (exactly backwards: rules monopolized more attempts when the budget grew). Now rules get two cheap shots regardless of cap, and Claude takes over. New helpers in `autosentry.state`: - `budget_exhausted(restarts, max_restarts)` — single source of truth for the kill-switch check; honors the `0 = unlimited` sentinel. - `format_budget(max_restarts)` — renders `∞` for unlimited, integer otherwise. Used by Monitor, status, TUI, doctor, and the incident report so every surface displays consistently. Existing configs keep their explicit values; only fresh `init` and `init --upgrade` pick up the new defaults (upgrade prompts per-key). 326 tests passing (3 new helper tests, 2 retargeted to literal-2). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
📝 WalkthroughWalkthroughThis PR refactors restart and healing budget handling across autosentry by introducing centralized helper functions, increasing default budget limits, and decoupling escalation thresholds. Default ChangesRestart/Healing Budget Refactoring
🎯 3 (Moderate) | ⏱️ ~25 minutes
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@CHANGELOG.md`:
- Around line 9-10: The changelog added a new release heading "## [0.13.1] —
2026-06-04" but the reference-link footer still maps "Unreleased" to "v0.8.4"
and doesn't define links for "0.13.1" and "0.13.0", so update the CHANGELOG.md
reference-link footer: change the Unreleased reference to point to the correct
comparison (or remove if not needed), and add explicit reference-link entries
for [0.13.1] and [0.13.0] with their corresponding tags/compare URLs (matching
the repo's tag naming like v0.13.1 and v0.13.0) so the headings
"0.13.1"/"0.13.0" resolve properly.
In `@src/autosentry/monitor.py`:
- Around line 141-143: Update the outdated comments that claim an "unlimited"
restart budget to reflect the current capped default of 50; locate the comment
blocks around the monitor flow mentioning max_restarts (the one referencing
"Decoupled from ``max_restarts``" and the later block around lines ~378-380) and
change their wording to indicate the restart budget is capped at 50 and that
rules get two cheap shots before the kill-switch at 50 restarts applies,
preserving the original intent about transient retries but matching the actual
capped default semantics.
In `@src/autosentry/state.py`:
- Around line 175-176: Update the stale docstring in the budget_exhausted
function/doc (budget_exhausted in src/autosentry/state.py) to stop claiming that
"max_restarts <= 0 is the sentinel for 'unlimited' — the default"; instead state
that max_restarts <= 0 remains the sentinel for unlimited but the current
configured default restart cap is 50, so the doc should clarify that the default
behavior is a 50 restart cap unless explicitly set to <= 0 for unlimited.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 76a05c29-bac9-432f-87c4-6eb519c04d7a
⛔ Files ignored due to path filters (1)
uv.lockis excluded by!**/*.lock
📒 Files selected for processing (14)
CHANGELOG.mdpyproject.tomlsrc/autosentry/cli/commands/doctor.pysrc/autosentry/cli/commands/status.pysrc/autosentry/config.pysrc/autosentry/incidents/report.pysrc/autosentry/monitor.pysrc/autosentry/state.pysrc/autosentry/templates/autosentry.yaml.tmplsrc/autosentry/tui.pytests/test_cli.pytests/test_pipeline.pytests/test_reset.pytests/test_restart_budget.py
Comment on lines
+9
to
+10
| ## [0.13.1] — 2026-06-04 | ||
|
|
There was a problem hiding this comment.
Update changelog reference links for the new release.
The new 0.13.1 section is added, but the reference-link footer still points Unreleased to v0.8.4 and doesn’t define links for 0.13.1/0.13.0, so release headings won’t link correctly.
📝 Proposed fix
-[Unreleased]: https://github.com/ulmentflam/autosentry/compare/v0.8.4...HEAD
+[Unreleased]: https://github.com/ulmentflam/autosentry/compare/v0.13.1...HEAD
+[0.13.1]: https://github.com/ulmentflam/autosentry/releases/tag/v0.13.1
+[0.13.0]: https://github.com/ulmentflam/autosentry/releases/tag/v0.13.0🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@CHANGELOG.md` around lines 9 - 10, The changelog added a new release heading
"## [0.13.1] — 2026-06-04" but the reference-link footer still maps "Unreleased"
to "v0.8.4" and doesn't define links for "0.13.1" and "0.13.0", so update the
CHANGELOG.md reference-link footer: change the Unreleased reference to point to
the correct comparison (or remove if not needed), and add explicit
reference-link entries for [0.13.1] and [0.13.0] with their corresponding
tags/compare URLs (matching the repo's tag naming like v0.13.1 and v0.13.0) so
the headings "0.13.1"/"0.13.0" resolve properly.
Comment on lines
+141
to
+143
| # Decoupled from ``max_restarts`` so the unlimited-budget | ||
| # default doesn't push Claude escalation off to infinity — | ||
| # rules get two cheap shots at known transients, then the |
There was a problem hiding this comment.
Fix outdated “unlimited default” comments in monitor flow.
Line 141 and Line 378 describe uncapped/unlimited restart budget as the default, but the default is now capped (50). Please update both comments to match current kill-switch semantics.
Also applies to: 378-380
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/autosentry/monitor.py` around lines 141 - 143, Update the outdated
comments that claim an "unlimited" restart budget to reflect the current capped
default of 50; locate the comment blocks around the monitor flow mentioning
max_restarts (the one referencing "Decoupled from ``max_restarts``" and the
later block around lines ~378-380) and change their wording to indicate the
restart budget is capped at 50 and that rules get two cheap shots before the
kill-switch at 50 restarts applies, preserving the original intent about
transient retries but matching the actual capped default semantics.
Comment on lines
+175
to
+176
| ``max_restarts <= 0`` is the sentinel for "unlimited" — the default. | ||
| Centralized so every caller (Monitor, doctor, vault) agrees on the |
There was a problem hiding this comment.
Update stale default wording in budget_exhausted docstring.
Line 175 says the unlimited sentinel is “the default,” but the current default restart cap is 50. Please align this wording to avoid operator confusion.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/autosentry/state.py` around lines 175 - 176, Update the stale docstring
in the budget_exhausted function/doc (budget_exhausted in
src/autosentry/state.py) to stop claiming that "max_restarts <= 0 is the
sentinel for 'unlimited' — the default"; instead state that max_restarts <= 0
remains the sentinel for unlimited but the current configured default restart
cap is 50, so the doc should clarify that the default behavior is a 50 restart
cap unless explicitly set to <= 0 for unlimited.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
The previous defaults gave up too quickly for the long-running workloads autosentry exists to babysit (ML training, multi-hour data pipelines), while still letting a stuck-loop healer spam the API. Both knobs are now wide enough that productive workloads never feel capped, narrow enough that broken healing eventually stops.
process.restart_policy.max_restarts1050healing.budget.max_attempts_per_detector_per_hour560healing.budget.max_wall_seconds_per_incident6007200max_restartsis reframed as a kill-switch, not a budget:state.restartszeros on every kept fix, so this only trips when the healer can't land a kept fix for N restarts in a row. A productive healer runs the supervisor indefinitely.healing.escalate_to_claude_afterdecouples frommax_restarts // 5to a literal2. The old formula was backwards — raisingmax_restartspushed Claude escalation later, letting rules monopolize a bigger slice of the budget. Now rules get two cheap shots regardless of cap.New helpers in
autosentry.state:budget_exhausted(restarts, max_restarts)— single source of truth, honors the0 = unlimitedsentinel.format_budget(max_restarts)— renders∞for unlimited.Existing configs keep their explicit values; only fresh
initandinit --upgradepick up the new defaults (upgrade prompts per-key, so users can decline).Test plan
test_reset.py(budget_exhausted, format_budget)test_restart_budget.py(literal-2 escalation)test_cli.py(template default → 50)test_pipeline.py(stage inherits new default)ruff check+ruff format+pyreflyclean🤖 Generated with Claude Code
Summary by CodeRabbit
New Features
Configuration Updates