feat(health-score): raise context thresholds and update critical coaching [LCO-44] - #42
Merged
DevanshuNEU merged 3 commits intoApr 19, 2026
Conversation
…hing [LCO-44] HEALTHY_CEIL 50->70, DEGRADING_CEIL 80->90. Critical coaching now reads "Context nearly full. Start a new chat, or use Claude Projects for ongoing work." Update all hardcoded threshold values in audit, unit, and integration tests to match new constants.
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (4)
✅ Files skipped from review due to trivial changes (1)
🚧 Files skipped from review as they are similar to previous changes (2)
📝 WalkthroughWalkthroughUpdated health-score context utilization thresholds: Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Poem
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 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://cdn.github.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (2)
tests/integration/agent-pipeline.test.ts (1)
203-206:⚠️ Potential issue | 🟡 MinorStale inline comment:
DEGRADING_CEILis 90, not 80.At
contextPct: 85, turnCount: 25, Rule 1 does not fire (85 < 90); the fixture is actually critical via Rule 2 (context ≥HEALTHY_CEILand turns >TURN_DEGRADING_CEIL). The assertion still passes, but the justification comment is wrong.Proposed fix
it('health agent returns critical', () => { - // contextPct(85) >= DEGRADING_CEIL(80) + // contextPct(85) >= HEALTHY_CEIL(70) && turnCount(25) > TURN_DEGRADING_CEIL(20) -> Rule 2 critical expect(health.level).toBe('critical'); });🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@tests/integration/agent-pipeline.test.ts` around lines 203 - 206, The inline comment in the test "health agent returns critical" is incorrect: DEGRADING_CEIL is 90 not 80, so the current fixture (contextPct: 85, turnCount: 25) triggers Rule 2 rather than Rule 1. Update the comment to reflect the correct logic by stating that contextPct (85) < DEGRADING_CEIL (90) and the agent is critical because contextPct >= HEALTHY_CEIL and turnCount > TURN_DEGRADING_CEIL; reference the test identifier ('health agent returns critical') and the constants DEGRADING_CEIL, HEALTHY_CEIL, and TURN_DEGRADING_CEIL when making the clarification.lib/health-score.ts (1)
50-73:⚠️ Potential issue | 🟡 MinorStale threshold references in doc comments.
Two comments still reference the old thresholds, which will confuse future readers now that the constants have moved:
- Line 54: "Critical from context alone (>80%)" — should be 90%.
- Line 72: "past 20 turns with >50% context" — Rule 2 now fires at
HEALTHY_CEIL = 70, not 50.Proposed fix
- * A conversation can be Critical from context alone (>80%) or from - * a combination of moderate context + high turn count (the "attention - * valley" effect from context rot research). + * A conversation can be Critical from context alone (>=90%) or from + * a combination of moderate context + high turn count (the "attention + * valley" effect from context rot research). @@ // Rule 2: High context + many turns = Critical. - // The "attention valley": past 20 turns with >50% context, Claude's attention - // to mid-conversation details degrades measurably. + // The "attention valley": past 20 turns with >=70% context, Claude's attention + // to mid-conversation details degrades measurably.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@lib/health-score.ts` around lines 50 - 73, Update the stale inline doc comments in computeHealthScore to match the current threshold constants: replace "Critical from context alone (>80%)" with the correct DEGRADING_CEIL value (90%) and update the "past 20 turns with >50% context" reference to reflect HEALTHY_CEIL (70%)/the current Rule 2 threshold; ensure the comments mention the DEGRADING_CEIL and HEALTHY_CEIL symbols so future readers link the prose to the actual constants used by the function.
🧹 Nitpick comments (1)
tests/unit/health-score.test.ts (1)
97-113: Test body does not match its description and contains a dead computation.This block was touched by the threshold update, so worth cleaning up while here:
hat line 98 is assigned but never asserted — it's Rule 1 critical and unused.- The test is titled "uses singular 'message' when only 1 message remaining" but the only assertion matches the plural
~0 messages, and even that runs onlyif (h2.level === 'degrading'), so the test can pass without asserting anything.Consider replacing with an input that deterministically produces
~1 messageunder Rule 4 and asserting unconditionally (e.g.contextPct: 89, growthRate: 11→remaining = round(11/11) = 1).🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@tests/unit/health-score.test.ts` around lines 97 - 113, The test assigns an unused variable h and its body contradicts the title: it purports to check singular "message" but asserts "~0 messages" conditionally. Replace the setup to deterministically hit Rule 4 (remaining ≈ 1) by removing the unused h, call computeHealthScore with an input that yields remaining = 1 (e.g. contextPct and growthRate adjusted so (100 - contextPct)/growthRate rounds to 1), and then unconditionally assert that computeHealthScore(...).coaching matches the singular "message" form; update references to the input helper and computeHealthScore (and remove the conditional if (h2.level === 'degrading')) so the test always verifies the singular wording.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@tests/audit/health-score-audit.test.ts`:
- Around line 17-19: Update the stale section header and describe label that
reference an 80% threshold to reflect the current 90% threshold: change the
comment line ("computeHealthScore: Rule 1 (context >= 80% = critical)") and the
describe title string ("computeHealthScore: Rule 1 (high context = critical)")
in tests/audit/health-score-audit.test.ts to mention 90% so they match the
updated assertions and avoid misleading docs.
---
Outside diff comments:
In `@lib/health-score.ts`:
- Around line 50-73: Update the stale inline doc comments in computeHealthScore
to match the current threshold constants: replace "Critical from context alone
(>80%)" with the correct DEGRADING_CEIL value (90%) and update the "past 20
turns with >50% context" reference to reflect HEALTHY_CEIL (70%)/the current
Rule 2 threshold; ensure the comments mention the DEGRADING_CEIL and
HEALTHY_CEIL symbols so future readers link the prose to the actual constants
used by the function.
In `@tests/integration/agent-pipeline.test.ts`:
- Around line 203-206: The inline comment in the test "health agent returns
critical" is incorrect: DEGRADING_CEIL is 90 not 80, so the current fixture
(contextPct: 85, turnCount: 25) triggers Rule 2 rather than Rule 1. Update the
comment to reflect the correct logic by stating that contextPct (85) <
DEGRADING_CEIL (90) and the agent is critical because contextPct >= HEALTHY_CEIL
and turnCount > TURN_DEGRADING_CEIL; reference the test identifier ('health
agent returns critical') and the constants DEGRADING_CEIL, HEALTHY_CEIL, and
TURN_DEGRADING_CEIL when making the clarification.
---
Nitpick comments:
In `@tests/unit/health-score.test.ts`:
- Around line 97-113: The test assigns an unused variable h and its body
contradicts the title: it purports to check singular "message" but asserts "~0
messages" conditionally. Replace the setup to deterministically hit Rule 4
(remaining ≈ 1) by removing the unused h, call computeHealthScore with an input
that yields remaining = 1 (e.g. contextPct and growthRate adjusted so (100 -
contextPct)/growthRate rounds to 1), and then unconditionally assert that
computeHealthScore(...).coaching matches the singular "message" form; update
references to the input helper and computeHealthScore (and remove the
conditional if (h2.level === 'degrading')) so the test always verifies the
singular wording.
🪄 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: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: fbd63ccf-f884-4bed-b959-5e7ec850a134
📒 Files selected for processing (4)
lib/health-score.tstests/audit/health-score-audit.test.tstests/integration/agent-pipeline.test.tstests/unit/health-score.test.ts
…essage test [LCO-44]
This was referenced Apr 28, 2026
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
Raises the health-score thresholds to better reflect real model degradation curves: degrading now starts at 70% context fill (was 50%) and critical at 90% (was 80%). Updates the critical coaching message to mention Claude Projects as an option alongside starting a new chat.
Type of Change
feat— New featureWhat Was Changed
lib/health-score.ts:HEALTHY_CEIL50 -> 70,DEGRADING_CEIL80 -> 90. Critical Rule 1 coaching updated to "Context nearly full. Start a new chat, or use Claude Projects for ongoing work."tests/unit/health-score.test.ts: Updated two hardcoded values that referenced old thresholds (contextPct: 55 -> 75 in rule-priority test; stale comment 80 -> 90).tests/audit/health-score-audit.test.ts: Updated all hardcoded threshold values and describe/test names to match new constants (80->90, 79.9->89.9, 50->70 in Rule 2/3 boundary tests).tests/integration/agent-pipeline.test.ts: Updated expectation for 55%-context fixture from degrading -> healthy (correct at new HEALTHY_CEIL=70). Added new degrading-boundary describe block at 75% context.How to Test
bun run compile— should exit cleanbun run test— 1452 tests, all passChecklist
bun run test)bun run compile)bun run build)feat:,fix:,refactor:,test:,chore:)Related Issues
Closes LCO-44
Notes for Reviewer
The old thresholds (50/80) were overly conservative and flagged healthy conversations as degrading. The new thresholds (70/90) match observed model attention degradation patterns more accurately. All tests that previously used hardcoded values were updated; tests that imported exported constants auto-adjusted.
Summary by CodeRabbit
Updates
Tests