HAL9000/cleveragents-core
HAL9000/cleveragents-core
0
0
Fork 2
forked from cleveragents/cleveragents-core
Code Pull requests 3 Activity

[sentinel #11238] fix(acms): harden hot/warm/cold tier service reliability #2

Open
HAL9000 wants to merge 2 commits from tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability into master
pull from: tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability
merge into: HAL9000:master
Conversation 0 Commits 2 Files changed 1 +41 -18
HAL9000 commented 2026-05-21 10:32:48 +00:00
Owner
Copy link

This is an auto-generated sentinel duplicate of upstream PR cleveragents/cleveragents-core#11238 for pipeline testing. It targets the fork's master and is isolated from the canonical pipeline.

Summary

Harden the ContextTierService for production reliability under concurrent plan execution and high-throughput scenario.

Changes

  1. Remove dead conflicting defaults: Deleted _DEFAULT_MAX_TOKENS_HOT, _DEFAULT_MAX_DECISIONS_WARM, _DEFAULT_MAX_DECISIONS_COLD which were never used (all budget flows through budget_from_settings() in context_tier_settings.py) and contradicted the canonical values.

  2. Warm-tier capacity enforcement: Added _enforce_warm_capacity() — count-based LRU eviction of the warm tier when max_decisions_warm is exceeded. Triggered after:

    • Cold → warm promotion (in promote())
    • Hot-budget-fallback restore to warm (when promoted fragment could't fit in hot)

  3. Immutable snapshot returns: get_all_fragments() and get_hot_fragments() now return model_copy() instances instead of live mutable references, preventing state corruption from external mutation while the RLock is held.

  4. Naming consistency: _COLD_SUMMARY_MAX_CHARS renamed to _default_summarisation_max_chars following snake_case convention used throughout the module.

Impact

  • No breaking API changes — return types and method signatures unchanged (copies still implement TieredFragment).
  • Warmer tiers no longer silently exceed max_decisions_warm capacity after promotions.
  • Thread-safety improved for all public reader methods.

Original upstream PR: cleveragents/cleveragents-core#11238

_This is an auto-generated sentinel duplicate of upstream PR `cleveragents/cleveragents-core#11238` for pipeline testing. It targets the fork's `master` and is **isolated** from the canonical pipeline._ --- ## Summary Harden the `ContextTierService` for production reliability under concurrent plan execution and high-throughput scenario. ### Changes 1. **Remove dead conflicting defaults**: Deleted `_DEFAULT_MAX_TOKENS_HOT`, `_DEFAULT_MAX_DECISIONS_WARM`, `_DEFAULT_MAX_DECISIONS_COLD` which were never used (all budget flows through `budget_from_settings()` in `context_tier_settings.py`) and contradicted the canonical values. 2. **Warm-tier capacity enforcement**: Added `_enforce_warm_capacity()` — count-based LRU eviction of the warm tier when `max_decisions_warm` is exceeded. Triggered after: - Cold → warm promotion (in `promote()`) - Hot-budget-fallback restore to warm (when promoted fragment could't fit in hot) 3. **Immutable snapshot returns**: `get_all_fragments()` and `get_hot_fragments()` now return `model_copy()` instances instead of live mutable references, preventing state corruption from external mutation while the RLock is held. 4. **Naming consistency**: `_COLD_SUMMARY_MAX_CHARS` renamed to `_default_summarisation_max_chars` following snake_case convention used throughout the module. ### Impact - No breaking API changes — return types and method signatures unchanged (copies still implement `TieredFragment`). - Warmer tiers no longer silently exceed `max_decisions_warm` capacity after promotions. - Thread-safety improved for all public reader methods. --- _Original upstream PR: cleveragents/cleveragents-core#11238_
HAL9000 self-assigned this 2026-05-21 10:32:48 +00:00 
- Remove unused conflicting _DEFAULT_* constants that conflicted with
  canonical defaults in context_tier_settings.py (which serves as the
  sole source of truth for budget/setting defaults).

- Add _enforce_warm_capacity() to enforce max_decisions_warm limit on
  warm tier after cold→warm promotion and hot-budget-fallback restore,
  preventing silent over-capacity data accumulation.

- Return deep copies (model_copy) from get_all_fragments() and
  get_hot_fragments() to prevent callers from mutating internal fragment
  state while the service holds its RLock under concurrent plan execution.

- Rename _COLD_SUMMARY_MAX_CHARS → _default_summarisation_max_chars for
  consistent snake_case naming throughout the module.

Fixes: PR #9663
This pull request can be merged automatically.
You are not authorized to merge this pull request.
View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.
git fetch -u origin tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability:tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability
git switch tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability

Merge

Merge the changes and update on Forgejo.

Warning: The "Autodetect manual merge" setting is not enabled for this repository, you will have to mark this pull request as manually merged afterwards.

git switch master
git merge --no-ff tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability
git switch tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability
git rebase master
git switch master
git merge --ff-only tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability
git switch tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability
git rebase master
git switch master
git merge --no-ff tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability
git switch master
git merge --squash tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability
git switch master
git merge --ff-only tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability
git switch master
git merge tests/sentinel-11238-pr-fix-9663-hot-warm-cold-tier-reliability
git push origin master
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
HAL9000
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
HAL9000/cleveragents-core!2
No description provided.