Skip to content

feat: dreaming engine with precise config (new PR) do not review/adjust.#798

Closed
jlin53882 wants to merge 1 commit into
CortexReach:masterfrom
jlin53882:fix/dreaming-clean
Closed

feat: dreaming engine with precise config (new PR) do not review/adjust.#798
jlin53882 wants to merge 1 commit into
CortexReach:masterfrom
jlin53882:fix/dreaming-clean

Conversation

@jlin53882
Copy link
Copy Markdown
Contributor

這是另一個 PR,使用精確編輯而非區塊替換。內容包含:\n- openclaw.plugin.json: +43 行精確編輯\n- index.ts: 原本所有修改\n- src + test: 新增檔案\n\n不要關閉原本的 PR 752,等確認沒問題後再合并。

Copy link
Copy Markdown

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

/rozhodli jsme|budeme pouvat/i,

P2 Badge Restore localized capture trigger literals

For Czech/Chinese inputs, these regex literals were mojibaked in this commit, so phrases that previously triggered capture such as radši no longer match the user text. This directly regresses auto-capture/category detection for non-ASCII memory commands and should be restored to the original Unicode literals.


if (emptyPages >= MAX_EMPTY_PAGES) {
debugLog(`paginate [${scope}]: stopping after ${MAX_EMPTY_PAGES} consecutive pages with no exact-scope matches`);
break;

P2 Badge Do not stop pagination after three empty pages

When a non-global scope has more than 3 * pageSize newer legacy/global rows ahead of its own rows, store.list([scope]) still returns those global/null rows first, so this cutoff exits before the offset reaches the target-scope memories. In that scenario the dreaming phases skip the scoped memories entirely despite the pagination being intended to avoid this starvation.


api.logger.info("memory-lancedb-pro: AccessTracker destroyed");

P2 Badge Flush access updates before destroying the tracker

On plugin stop, this calls destroy() without first awaiting accessTracker.flush(). Because recordAccess() buffers manual recalls for up to 5 seconds and destroy() clears pending entries before its best-effort async flush, any recalls in that debounce window lose their access_count/last_accessed_at updates, which also undermines the new dreaming promotion logic that relies on those counters.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread openclaw.plugin.json
Comment on lines +125 to +128
"properties": {
"lookbackDays": {"type": "number", "minimum": 1, "default": 7},
"limit": {"type": "number", "minimum": 1, "default": 20}
}
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Expose REM pattern threshold in the manifest

The engine and mergeDreamingConfig() support phases.rem.minPatternStrength, but the manifest sets additionalProperties: false for rem and only allows lookbackDays and limit. Users therefore cannot configure the REM sensitivity without failing plugin config validation, even though the runtime code reads that setting.

Useful? React with 👍 / 👎.

@jlin53882 jlin53882 changed the title feat: dreaming engine with precise config (new PR) feat: dreaming engine with precise config (new PR) do not review/adjust. May 11, 2026
@jlin53882 jlin53882 force-pushed the fix/dreaming-clean branch from 3cdfd3f to 4a1c163 Compare May 11, 2026 17:59
@jlin53882 jlin53882 closed this May 12, 2026
Copy link
Copy Markdown

Closing this PR for now.

This is not a standalone reviewable change: it is tightly coupled to the larger dreaming-engine work in #752, the title explicitly says "do not review/adjust", and the current head still has failing CI/review findings. Please keep the dreaming-engine discussion consolidated in #752 or open a fresh clean follow-up after that PR is ready.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants