fix(compaction): pass model through runtime for safeguard summaries by battman21 · Pull Request #17864 · openclaw/openclaw

battman21 · 2026-02-16T07:50:46Z

Summary

Fixes #3479 — compaction always failing with "Summary unavailable due to context limits"

This PR addresses the root cause: ctx.model is undefined during compaction because extensionRunner.initialize() is never called in the compact.ts workflow. The compaction safeguard immediately returns a fallback summary without attempting LLM summarization.

Changes

Pass model through runtime registry
- Add model field to CompactionSafeguardRuntimeValue type
- Store params.model when building embedded extension paths
- Fall back to runtime.model when ctx.model is undefined
- Add warning when both are missing (diagnostic for future issues)

Why This Approach

Following the established runtime registry pattern (same as maxHistoryShare and contextWindowTokens) rather than attempting to call extensionRunner.initialize() directly, which is SDK-internal and not meant for external access.

Testing

Lint passes
Build passes
Verified against issue reproduction steps

Added model field to CompactionSafeguardRuntimeValue type with clear documentation
Modified buildEmbeddedExtensionFactories to store params.model in runtime registry
Updated compaction-safeguard.ts to use ctx.model ?? runtime?.model fallback pattern
Added diagnostic warning when both models are missing (logged once per session via WeakSet)
Comprehensive test coverage including the specific compact.ts workflow scenario and edge cases

Confidence Score: 5/5

This PR is safe to merge with no identified risks
The fix follows established patterns in the codebase, includes comprehensive test coverage (including edge cases), and the changelog entry is properly formatted. The implementation is minimal, focused, and directly addresses the root cause without introducing new complexity or side effects.
No files require special attention

_{Last reviewed commit: d9b1f0e}

mudrii

Code Review

Verdict: APPROVE ✅

What it fixes

Compaction always fails with "Summary unavailable due to context limits" because ctx.model is undefined in the compact.ts workflow — extensionRunner.initialize() is never called there, so the model is never set on the context.

Analysis

Model passthrough via runtime registry — Well done. Adds model field to CompactionSafeguardRuntimeValue, stores params.model in buildEmbeddedExtensionPaths, and uses ctx.model ?? runtime?.model fallback in the safeguard. This follows the exact same pattern already used for maxHistoryShare and contextWindowTokens — consistent and low-risk.

BASE_CHUNK_RATIO: 0.4 → 0.25 — Reasonable and conservative. On 200K context, chunks go from 80K → 50K tokens. MIN_CHUNK_RATIO is 0.15, so there is room to shrink further if needed. The rationale is sound: when compaction triggers reactively at 90%+ usage, oversized chunks can exceed what the summarization model can handle after adding prompt overhead. Trade-off is more summarization calls (slightly higher cost/latency) but significantly more reliable compaction.

Code quality — The runtime variable was already fetched later in the function; the diff just moves its usage earlier for cleaner flow. Type is Model<Api> which is correct. The redundant if (!ctx.model && !runtime?.model) inside the existing if (!model) block is always true at that point (since model = ctx.model ?? runtime?.model), so the inner check is unnecessary but harmless as documentation.

Minor suggestion

No unit tests were added for the ctx.model undefined → runtime fallback path. Would be nice to have but the change surface is small and the pattern is proven.

Overall clean, well-scoped fix. Ship it. 🚀

battman21 · 2026-02-16T08:29:01Z

Updated the PR based on review feedback:

Removed:

❌ BASE_CHUNK_RATIO change (0.4 → 0.25) — this is a separate behavioral change affecting all compaction flows, not just the bug fix. Will create a separate PR with perf/quality evidence if needed.

Added:

✅ Once-per-session warning using WeakSet to prevent log spam when both ctx.model and runtime.model are undefined
✅ Regression test verifying runtime.model is stored/retrieved correctly (proves the fallback path works)

The PR now focuses solely on fixing the root cause: ctx.model being undefined in compact.ts workflow.

Add integration tests to verify the model fallback behavior: - Test runtime.model fallback when ctx.model is undefined (compact.ts workflow) - Test fallback summary when both ctx.model and runtime.model are undefined - Test contextWindowTokens runtime storage/retrieval - Test combined runtime values (maxHistoryShare + contextWindowTokens + model) These tests verify the fix for issue openclaw#3479 where compaction fails due to ctx.model being undefined in the compact.ts workflow. The runtime registry pattern allows model to be passed when extensionRunner.initialize() is not called, ensuring summarization works in all code paths. Related: PR openclaw#17864

battman21 · 2026-02-16T08:46:13Z

Test Coverage Added

Added comprehensive integration tests for the model fallback behavior:

New Tests (4 total)

Runtime registry tests (3 tests):
- ✅ Stores and retrieves model from runtime (verifies storage mechanism)
- ✅ Stores and retrieves contextWindowTokens from runtime
- ✅ Stores and retrieves combined runtime values (maxHistoryShare + contextWindowTokens + model)
Integration tests (2 tests):
- ✅ Uses runtime.model when ctx.model is undefined - Verifies the root-cause fix: when extensionRunner.initialize() is not called (as in compact.ts workflow), the extension falls back to runtime.model and returns gracefully
- ✅ Returns fallback summary when both models are undefined - Verifies graceful degradation when neither ctx.model nor runtime.model are available

Test Results

✓ src/agents/pi-extensions/compaction-safeguard.e2e.test.ts (22 tests) 6ms
  Test Files  1 passed (1)
       Tests  22 passed (22)

All tests pass, covering:

Runtime registry storage/retrieval ✓
Model fallback logic in extension ✓
Graceful degradation when models are missing ✓

Commit: 9ebb455

Add integration tests to verify the model fallback behavior: - Test runtime.model fallback when ctx.model is undefined (compact.ts workflow) - Test fallback summary when both ctx.model and runtime.model are undefined - Test contextWindowTokens runtime storage/retrieval - Test combined runtime values (maxHistoryShare + contextWindowTokens + model) These tests verify the fix for issue openclaw#3479 where compaction fails due to ctx.model being undefined in the compact.ts workflow. The runtime registry pattern allows model to be passed when extensionRunner.initialize() is not called, ensuring summarization works in all code paths. Related: PR openclaw#17864

battman21 · 2026-02-16T09:22:34Z

Test Improvements (addressing medium-priority feedback)

Concrete Evidence of Fallback Path

Added explicit assertions to prove the fallback expression ctx.model ?? runtime?.model is exercised:

Test 1: Model fallback path

// Proves handler used runtime.model by verifying getApiKey was called with it
expect(getApiKeyMock).toHaveBeenCalledWith(model);

Test 2: Early return path

// Proves handler returned early without trying to use a model
expect(getApiKeyMock).not.toHaveBeenCalled();

What This Proves

✅ Fallback path exercised: When ctx.model is undefined but runtime.model is present, the handler calls getApiKey with runtime.model
✅ Early return path works: When both models are undefined, handler returns fallback summary without calling getApiKey

The previous version only verified the output and registry state, which could be true even if the fallback path was never used. Now we have direct evidence the handler consumed runtime.model via the getApiKey spy.

Test Classification Note

Acknowledged: test file is named .e2e.test.ts but uses unit-style mocking of extension handlers rather than true end-to-end behavior. This is acceptable for the scope of this fix (testing the extension API boundary).

Commit: cf69df1

Add integration tests to verify the model fallback behavior: - Test runtime.model fallback when ctx.model is undefined (compact.ts workflow) - Test fallback summary when both ctx.model and runtime.model are undefined - Test contextWindowTokens runtime storage/retrieval - Test combined runtime values (maxHistoryShare + contextWindowTokens + model) These tests verify the fix for issue openclaw#3479 where compaction fails due to ctx.model being undefined in the compact.ts workflow. The runtime registry pattern allows model to be passed when extensionRunner.initialize() is not called, ensuring summarization works in all code paths. Related: PR openclaw#17864

Fixes openclaw#3479 Root cause: extensionRunner.initialize() is never called in compact.ts workflow, leaving ctx.model undefined. Compaction safeguard checks ctx.model and returns fallback summary immediately without attempting LLM summarization. Changes: 1. Pass model through compaction safeguard runtime registry (same pattern as maxHistoryShare) 2. Fall back to runtime.model when ctx.model is undefined 3. Add once-per-session warning when both models are missing (prevents log spam) 4. Add regression test for runtime.model fallback This follows the established runtime registry pattern rather than attempting to call extensionRunner.initialize() (which is SDK-internal and not meant for direct access). Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Add integration tests to verify the model fallback behavior: - Test runtime.model fallback when ctx.model is undefined (compact.ts workflow) - Test fallback summary when both ctx.model and runtime.model are undefined - Test contextWindowTokens runtime storage/retrieval - Test combined runtime values (maxHistoryShare + contextWindowTokens + model) These tests verify the fix for issue openclaw#3479 where compaction fails due to ctx.model being undefined in the compact.ts workflow. The runtime registry pattern allows model to be passed when extensionRunner.initialize() is not called, ensuring summarization works in all code paths. Related: PR openclaw#17864

- Add baseUrl to Model mock objects (now required by Model<Api>) - Add explicit Model<Api> annotation to prevent provider string widening - Cast modelRegistry mock through unknown (ModelRegistry expanded) - Use non-null assertion for compactionHandler (TypeScript strict) - Type compaction result explicitly Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

vincentkoc · 2026-02-23T14:41:59Z

Rebase + readiness update:

Rebases cleanly onto current main.
Added changelog entry under 2026.2.22 (Unreleased) -> Fixes with contributor credit chain.
Focused test pi-extensions/compaction-safeguard.test.ts passes after rebase.

Given overlap, #4223 has been closed as duplicate with attribution preserved. This should stay canonical for the ctx.model/runtime-model safeguard path.

vincentkoc · 2026-02-23T15:09:38Z

@greptile review

…penclaw#17864) * fix(compaction): pass model through runtime to fix ctx.model undefined Fixes openclaw#3479 Root cause: extensionRunner.initialize() is never called in compact.ts workflow, leaving ctx.model undefined. Compaction safeguard checks ctx.model and returns fallback summary immediately without attempting LLM summarization. Changes: 1. Pass model through compaction safeguard runtime registry (same pattern as maxHistoryShare) 2. Fall back to runtime.model when ctx.model is undefined 3. Add once-per-session warning when both models are missing (prevents log spam) 4. Add regression test for runtime.model fallback This follows the established runtime registry pattern rather than attempting to call extensionRunner.initialize() (which is SDK-internal and not meant for direct access). Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * test: add comprehensive tests for compaction-safeguard model fallback Add integration tests to verify the model fallback behavior: - Test runtime.model fallback when ctx.model is undefined (compact.ts workflow) - Test fallback summary when both ctx.model and runtime.model are undefined - Test contextWindowTokens runtime storage/retrieval - Test combined runtime values (maxHistoryShare + contextWindowTokens + model) These tests verify the fix for issue openclaw#3479 where compaction fails due to ctx.model being undefined in the compact.ts workflow. The runtime registry pattern allows model to be passed when extensionRunner.initialize() is not called, ensuring summarization works in all code paths. Related: PR openclaw#17864 * fix(test): adapt compaction-safeguard tests to upstream type changes - Add baseUrl to Model mock objects (now required by Model<Api>) - Add explicit Model<Api> annotation to prevent provider string widening - Cast modelRegistry mock through unknown (ModelRegistry expanded) - Use non-null assertion for compactionHandler (TypeScript strict) - Type compaction result explicitly Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * Compaction: add changelog credit for model fallback fix * Update CHANGELOG.md * Update CHANGELOG.md --------- Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com> Co-authored-by: Vincent Koc <vincentkoc@ieee.org>

…penclaw#17864) * fix(compaction): pass model through runtime to fix ctx.model undefined Fixes openclaw#3479 Root cause: extensionRunner.initialize() is never called in compact.ts workflow, leaving ctx.model undefined. Compaction safeguard checks ctx.model and returns fallback summary immediately without attempting LLM summarization. Changes: 1. Pass model through compaction safeguard runtime registry (same pattern as maxHistoryShare) 2. Fall back to runtime.model when ctx.model is undefined 3. Add once-per-session warning when both models are missing (prevents log spam) 4. Add regression test for runtime.model fallback This follows the established runtime registry pattern rather than attempting to call extensionRunner.initialize() (which is SDK-internal and not meant for direct access). * test: add comprehensive tests for compaction-safeguard model fallback Add integration tests to verify the model fallback behavior: - Test runtime.model fallback when ctx.model is undefined (compact.ts workflow) - Test fallback summary when both ctx.model and runtime.model are undefined - Test contextWindowTokens runtime storage/retrieval - Test combined runtime values (maxHistoryShare + contextWindowTokens + model) These tests verify the fix for issue openclaw#3479 where compaction fails due to ctx.model being undefined in the compact.ts workflow. The runtime registry pattern allows model to be passed when extensionRunner.initialize() is not called, ensuring summarization works in all code paths. Related: PR openclaw#17864 * fix(test): adapt compaction-safeguard tests to upstream type changes - Add baseUrl to Model mock objects (now required by Model<Api>) - Add explicit Model<Api> annotation to prevent provider string widening - Cast modelRegistry mock through unknown (ModelRegistry expanded) - Use non-null assertion for compactionHandler (TypeScript strict) - Type compaction result explicitly * Compaction: add changelog credit for model fallback fix * Update CHANGELOG.md * Update CHANGELOG.md --------- Co-authored-by: Vincent Koc <vincentkoc@ieee.org>

mudrii approved these changes Feb 16, 2026

View reviewed changes

battman21 force-pushed the fix/3479-compaction-model-init branch from ced5b48 to 20ebf07 Compare February 16, 2026 08:28

openclaw-barnacle bot added agents Agent runtime and tooling size: XS labels Feb 16, 2026

openclaw-barnacle bot added size: M and removed size: XS labels Feb 16, 2026

battman21 force-pushed the fix/3479-compaction-model-init branch from 9ebb455 to cf69df1 Compare February 16, 2026 09:22

battman21 force-pushed the fix/3479-compaction-model-init branch from cf69df1 to 15ced77 Compare February 16, 2026 09:52

battman21 force-pushed the fix/3479-compaction-model-init branch from 15ced77 to c3ca172 Compare February 16, 2026 10:17

steipete closed this Feb 16, 2026

steipete reopened this Feb 17, 2026

battman21 force-pushed the fix/3479-compaction-model-init branch 2 times, most recently from f63ea1d to 0488b97 Compare February 17, 2026 03:04

battman21 force-pushed the fix/3479-compaction-model-init branch from 0488b97 to cd8f8af Compare February 17, 2026 03:07

battman21 force-pushed the fix/3479-compaction-model-init branch from cd8f8af to f9b1fb5 Compare February 17, 2026 07:51

openclaw-barnacle bot added the channel: telegram Channel integration: telegram label Feb 17, 2026

battman21 force-pushed the fix/3479-compaction-model-init branch from 804dbfe to 9122ef6 Compare February 17, 2026 09:43

openclaw-barnacle bot removed the channel: telegram Channel integration: telegram label Feb 17, 2026

battman21 and others added 4 commits February 23, 2026 09:40

Compaction: add changelog credit for model fallback fix

5945769

vincentkoc force-pushed the fix/3479-compaction-model-init branch from 6520fcb to 5945769 Compare February 23, 2026 14:41

vincentkoc mentioned this pull request Feb 23, 2026

fix: compaction safeguard falls through when ctx.model is unavailable #4223

Closed

vincentkoc added the dedupe:parent Primary canonical item in dedupe cluster label Feb 23, 2026

Update CHANGELOG.md

79627bf

vincentkoc changed the title ~~fix(compaction): pass model through runtime + reduce chunk ratio to 0.25~~ fix(compaction): pass model through runtime for safeguard summaries Feb 23, 2026

vincentkoc added 2 commits February 23, 2026 10:09

Merge branch 'main' into fix/3479-compaction-model-init

d9b1f0e

Update CHANGELOG.md

0c25af3

vincentkoc merged commit 01380f4 into openclaw:main Feb 23, 2026
6 checks passed

vabole mentioned this pull request Feb 23, 2026

ops: decision needed for upstream sync (3 disputed files) vabole/openclaw#3

Merged

3 tasks

github-actions bot mentioned this pull request Feb 23, 2026

📡 Upstream Digest — 2026-02-23 16:45 UTC curtismercier/openclaw-mods#110

Open

github-actions bot mentioned this pull request Feb 24, 2026

上游更新: v2026.2.23, v2026.2.22 — 114 P0 + 122 P1 待合并 jiulingyun/openclaw-cn#328

Closed

jiulingyun mentioned this pull request Feb 24, 2026

upstream(compaction): 移植 1 个冲突 commit (P1) — v2026.2.21→v2026.2.23 jiulingyun/openclaw-cn#428

Closed

6 tasks

arjunaskykok mentioned this pull request Feb 26, 2026

fix/test a2ui bundle preflight #27345

Open

2 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(compaction): pass model through runtime for safeguard summaries#17864

fix(compaction): pass model through runtime for safeguard summaries#17864
vincentkoc merged 7 commits intoopenclaw:mainfrom
battman21:fix/3479-compaction-model-init

battman21 commented Feb 16, 2026 •

edited by greptile-apps bot

Loading

Uh oh!

mudrii left a comment

Uh oh!

battman21 commented Feb 16, 2026

Uh oh!

battman21 commented Feb 16, 2026

Uh oh!

battman21 commented Feb 16, 2026

Uh oh!

vincentkoc commented Feb 23, 2026

Uh oh!

vincentkoc commented Feb 23, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Uh oh!

Conversation

battman21 commented Feb 16, 2026 • edited by greptile-apps bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Changes

Why This Approach

Testing

Related

Greptile Summary

Confidence Score: 5/5

Uh oh!

mudrii left a comment

Choose a reason for hiding this comment

Code Review

What it fixes

Analysis

Minor suggestion

Uh oh!

battman21 commented Feb 16, 2026

Uh oh!

battman21 commented Feb 16, 2026

Test Coverage Added

New Tests (4 total)

Test Results

Uh oh!

battman21 commented Feb 16, 2026

Test Improvements (addressing medium-priority feedback)

Concrete Evidence of Fallback Path

What This Proves

Test Classification Note

Uh oh!

vincentkoc commented Feb 23, 2026

Uh oh!

vincentkoc commented Feb 23, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

battman21 commented Feb 16, 2026 •

edited by greptile-apps bot

Loading