Issue Station
Sign In
Tools hereToolbox

openclaw - 💡(How to fix) Fix False LCM/plugin diagnostics in 2026.4.8, plus openai-codex/gpt-5.4 context mismatch [1 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
openclaw/openclaw#63331•Fetched 2026-04-09 07:55:13
View on GitHub
Comments
0
Participants
1
Timeline
0
Reactions
0
Author
drjontamdrjontam
Participants
drjontamdrjontam

After updating OpenClaw to 2026.4.8 and lossless-claw to 0.6.3, core LCM compaction works, but status / doctor / gateway diagnostics report multiple false warnings.

Root Cause

These appear incorrect because:

  • config contains valid compaction / summary model settings
  • plugins.allow is populated
  • plugin install provenance exists
  • LCM compaction actually runs successfully in practice
RAW_BUFFERClick to expand / collapse

Summary

After updating OpenClaw to 2026.4.8 and lossless-claw to 0.6.3, core LCM compaction works, but status / doctor / gateway diagnostics report multiple false warnings.

False diagnostics observed

  • [lcm] Compaction summarization model: (unconfigured)
  • plugins.allow is empty; discovered non-bundled plugins may auto-load...
  • lossless-claw: loaded without install/load-path provenance...

These appear incorrect because:

  • config contains valid compaction / summary model settings
  • plugins.allow is populated
  • plugin install provenance exists
  • LCM compaction actually runs successfully in practice

Environment

  • OpenClaw: 2026.4.8
  • lossless-claw: updated from 0.5.3 to 0.6.3
  • platform: macOS
  • install style: global OpenClaw + user extension install under ~/.openclaw/extensions/lossless-claw
  • active configured default model: openai-codex/gpt-5.4

Repro

  1. Configure lossless-claw with:
    • plugins.entries.lossless-claw.config.summaryModel = openai-codex/gpt-5.4
  2. Also configure:
    • agents.defaults.compaction.model = openai-codex/gpt-5.4
  3. Ensure plugins.allow includes lossless-claw
  4. Restart gateway
  5. Run:
    • openclaw status
    • openclaw status --deep
    • openclaw doctor --non-interactive

Observed

Diagnostics still report:

  • [lcm] Compaction summarization model: (unconfigured)
  • plugins.allow is empty; discovered non-bundled plugins may auto-load...
  • lossless-claw: loaded without install/load-path provenance...

Expected

Diagnostics should reflect actual state:

  • summary model should be shown as configured
  • allowlist should not be reported empty if populated
  • installed plugin should not be reported as lacking provenance when install metadata exists

Evidence that diagnostics are false

Config present

openclaw.json contains:

  • plugins.entries.lossless-claw.config.summaryModel = openai-codex/gpt-5.4
  • agents.defaults.compaction.model = openai-codex/gpt-5.4

Install provenance present

plugins.installs.lossless-claw contains a valid install record for 0.6.3

Compaction actually works

Gateway logs show successful LCM compaction passes, including lines like:

  • auto-compaction succeeded for openai-codex/gpt-5.4
  • repeated LCM compaction leaf pass (normal): ... -> ...

So this appears to be a diagnostics/reporting bug, not a real compaction failure.

Additional model-resolution inconsistency

There also appears to be a model-context inconsistency between catalog surfaces:

  • openclaw models list shows openai-codex/gpt-5.4 with ~1025k context
  • runtime model discovery / session context resolution for the active openai-codex/gpt-5.4 path resolves as 272000

Separately, discovered models show:

  • openai-codex/gpt-5.4 -> 272000
  • openai/gpt-5.4-pro -> 1050000

This suggests there may also be a mismatch between configured-model display, discovered runtime model metadata, and the context budget actually used by sessions.

Likely area

Possibly one of:

  • plugin config handoff mismatch
  • status / doctor using stale or different config view
  • provenance / allowlist reporting using stale registry state
  • model catalog display vs runtime model-resolution drift

extent analysis

TL;DR

The most likely fix for the false warnings in OpenClaw diagnostics is to verify the configuration handoff between plugins and the main application, ensuring that the latest settings are properly propagated and reflected in the diagnostics output.

Guidance

  1. Verify plugin configuration: Double-check that the plugins.entries.lossless-claw.config.summaryModel and agents.defaults.compaction.model settings are correctly configured and match the expected values.
  2. Check for stale configuration: Ensure that the configuration files, such as openclaw.json, are up-to-date and properly loaded by the application, and that there are no caching issues that might cause stale data to be used.
  3. Investigate model catalog inconsistencies: Look into the discrepancies between the model context sizes reported by openclaw models list and the runtime model discovery, as this might indicate a deeper issue with model resolution or configuration.
  4. Review plugin loading and provenance: Verify that the lossless-claw plugin is properly loaded and that its install provenance is correctly recorded and reflected in the diagnostics output.

Example

No specific code snippet is provided, as the issue seems to be related to configuration and diagnostics rather than code-level problems.

Notes

The issue might be related to a bug in the diagnostics reporting or a mismatch between the configured settings and the actual state of the application. Further investigation is needed to determine the root cause and apply a definitive fix.

Recommendation

Apply a workaround by verifying and updating the configuration files, ensuring that all settings are correctly propagated and reflected in the diagnostics output. This might involve manually checking and updating the openclaw.json file and restarting the application to ensure that the changes are applied.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#authentication issue#prompt issue#agent setup#task chaining#parallel task

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING