Conversation
Signed-off-by: Gout999 <gout999@users.noreply.github.com> (cherry picked from commit d5f506282351af9cc04617ede23f898fd6a829bc)
Collaborator
|
Thank you so much for your attention and contribution! We will arrange an internal review for this PR shortly, and all feedback will be shared right here in the discussion. |
Gout999
marked this pull request as ready for review
August 23, 2026 13:49
VittorioCai
pushed a commit
to VittorioCai/TencentDB-Agent-Memory
that referenced
this pull request
Sep 12, 2026
…xtraction switch had drifted back on The expert review threw out my problem statement: with extraction off the task is NOT "never queued" — trigger-service.ts writes the archive (161), takes the tasks mutex (179) and enqueues (209), and the worker pool starts either way. The defect is that the response cannot be told apart from one whose task will actually be extracted. Reproduced live and recorded with all three pieces of evidence the review asked for (evaluation/upstream/skill-extraction-flag/): the response (code 0 + task_id, no hint), the agent's _tasks.json (task registered with enqueued_at_ms), and Core's log (pool started, then "standalone SkillExtractor unavailable" from every worker, repeating). An empty candidate pool is explicitly not used as evidence — a real extraction may also produce none. The branch itself is fix/skill-extraction-flag in a separate worktree cut from upstream 0468a2a: one field, extraction_enabled, on the three archive entry points' success responses; the configuration switch only, null rather than false when the config cannot be read; 26 lines of implementation, 6 tests, plus the API doc and both SDKs. README records that "under 100 lines" holds for the implementation but not for the whole submission, and leaves that trade-off to the user. Dedup ran over the repository's full PR history, not only September; TencentCloud#1117 is the worker side and does not overlap. Found while setting up the reproduction: skill.extraction.enabled was ON live. The 11:33Z image switch ran start-memory-core.sh, which regenerates the whole mounted config (line 55) with extraction on — the trap commit c88e955 recorded, not re-checked after switching images. So ruling ③ was not in effect between 11:33Z and 17:56Z. Restored with the §10 record, drift scope verified as that one switch, and the three gate assets read back unchanged. No delivered run is affected: there were no runs on 9/12, and prepare.sh:81 fails hard when the switch is not off.
VittorioCai
pushed a commit
to VittorioCai/TencentDB-Agent-Memory
that referenced
this pull request
Sep 12, 2026
…t run on this base Checking the submit sequence turned up two things about the upstream side: - .github/PULL_REQUEST_TEMPLATE.md exists (Description / Related Issue / Change Type / Self-test Checklist / Additional Notes, with checkboxes), and --body-file bypasses it. PR-BODY.md is rewritten onto that structure with the boxes ticked against what was actually done — Bug fix plus Documentation update, and both self-test lines carrying the command and its result. The substance is unchanged; the reproduction, the scope declarations, the TencentCloud#1117 comparison and the size breakdown moved under Additional Notes. - .github/workflows/pr-ci.yml triggers on `pull_request: branches: [main]`, so none of its four jobs run for a PR targeting feat/server_team. The body says so and offers to re-run anything on request. Locally, against the real base: check-skill-queue-isolation.sh PASS, and `npm pack --dry-run` does not carry the new test file (files has `!src/**/*.test.ts`), so the published package keeps its shape.
6 tasks
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
2 participants
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
Addresses #1009 with a privacy-preserving observability slice for conversation-add
skill extraction.
The worker now reports aggregate candidate outcomes (
created,updated,non_create, and action counts) in its logs and task completion metadata. The helperdoes not emit skill content or skill names.
This PR does not change the extractor's decision policy or add a new task-result API;
it makes the existing outcomes inspectable first.
Verification
npm test -- --run src/core/skill/conversation-add/candidate-outcome.test.ts(2 tests passed)git diff --checkSigned-off-by: Gout999 gout999@users.noreply.github.com