[MatrixAdventurer]

由 LW 授权的 ChatGPT Personal 智能体提交。

What type of PR is this?

• fix (bug fix)
• perf (performance improvement)
• test (regression coverage)

Which issue(s) this PR fixes

Related to #354 and the performance suggestion on merged #352. No automatic issue closure requested.

What this PR does / why we need it

Automatic OpenClaw snapshot discovery needs names and timestamps, but its requests default to historical memory-count resolution on the server. Each uncached snapshot may trigger a historical COUNT and cache write whose result the plugin discards. With the default page budget this can touch up to 400 snapshots per scan.

Pass the existing resolve_memory_counts=false on every automatic page. Keep explicit transport limit/offset requests unchanged. The formatter/client and stats do not expose per-snapshot memory counts, so the public output stays unchanged. This is verified unnecessary database work, not a measured production latency improvement.

Verification

• Base main c3bb4b5; no dev exists, so main is used as explicitly authorized.
• Published head 2a613d0; full tree 5563356d9e8e31c320876ebb12139f9e0afff5ca equals tested local commit 6659385bc0b067fdd9c722e2ceed1a0b50fc416e.
• RED: npm test -- openclaw/tests/snapshot-pagination.test.ts against unchanged transport: 1 failed / 4 passed; all six automatic requests omitted the flag.
• GREEN: npm test: 234/234 tests across 15 files; npm run test:pack: build and verification passed, 6 packed files. Repeated full suite and packaging on Node24.19.0 and Node22.23.3.
• Real native loopback HTTP regression asserts three discovery and three stats pages request false, while explicit page selection retains the absent parameter. Independent review agent Approve and independent regression 5/5; remote head independently fetched and full-tree equality confirmed.
• git diff --check passed. Complete scope: 2 files, +8/-2. Generated dist is build-only and excluded by the repository; no dependencies, schema, server or workflow changes.
• Upstream OpenClaw CI 39639072851 SUCCESS; PR Title Check SUCCESS. Rust/MatrixOne CI 39639072838 SUCCESS, including 252 unit tests and DB Tests with restore/cleanup regressions.
• Local Rust/MatrixOne gates were not executed; actual upstream required CI has now completed successfully. PR is ready for review; XuPeng-SH assignment and review request were read back.
• Authenticated login, own fork upstream relationship and push permission verified. Only MatrixAdventurer/memoria:fix/openclaw-snapshot-counts-20261022 created/updated, using the GitHub connector and expected-SHA lease; no default branch written. Branch/commit publication and read-only verification performed by the signed execution subject. Issue [Bug]: OpenClaw snapshot discovery backfills historical counts it discards #354 assigned to actual login XuPeng-SH; PR is assigned to XuPeng-SH and the review request was read back after required gates succeeded.

由 LW 授权的 ChatGPT Personal 智能体提交。

[MatrixAdventurer]

由 LW 授权的 ChatGPT Personal 智能体提交。

固定发布与独立复审证据：review agent 独立 fetch 远程 2a613d0，完整 tree 5563356d9e8e31c320876ebb12139f9e0afff5ca 与已实测本地 6659385bc0b067fdd9c722e2ceed1a0b50fc416e 全树 diff 为空，parent 与指定基线一致，确认署名、diff --check 并明确 Approve。两文件 +8/-2，无后续代码修订。

已实际完成 Node24.19.0 和 Node22.23.3 两套环境的插件全量234/234和打包6文件验证；独立review另执行真实HTTP回归5/5。修复前同一回归1 failed/4 passed，6个自动请求均缺少计数禁用参数。固定本地及远程提交、创建/更新本人fork独立分支、只读核验均由以上署名主体执行。Issue #354 的 XuPeng-SH assignee 已在创建响应读回；PR暂保留draft，等待完整规定CI结果后再指派并请求该负责人审阅。

[MatrixAdventurer]

由 LW 授权的 ChatGPT Personal 智能体提交。

固定head 2a613d0 的 CI 续接证据：OpenClaw #39639072851 与 PR Title Check SUCCESS；Rust/MatrixOne #39639072838 中 Check & Clippy SUCCESS，Unit Tests SUCCESS。已读取单测原始日志，116+22+45+69 = 252 passed，0 failed。DB Tests job119358340694 仍在 All tests 步骤，后续恢复/清理回归尚未执行，不能声称完整CI通过。无新review、inline thread或requested changes；未改代码、未重跑CI，继续保留draft，完整门禁通过后再执行负责人审阅交付。只读核验执行主体同上署名。

[MatrixAdventurer]

由 LW 授权的 MOI 智能体提交。

完整 CI 与交付续接证据（固定 head 2a613d0，本轮无代码修改、未重跑 CI）：

• 继上一条记录（17:12:50Z，DB Tests 运行中）之后，本轮回访确认完整规定 CI 已全部完成且无失败：Rust/MatrixOne workflow run 39639072838 中 Check & Clippy、Unit Tests（252 passed）与 DB Tests（job 119358340694，含此前未完成的恢复/清理回归步骤）均 SUCCESS（DB Tests 完成于 17:18:21Z）；OpenClaw workflow run 39639072851 与 PR Title Check 维持 SUCCESS；Mergify Merge Queue 状态为 neutral/completed。当前无失败、运行中或缺失的必需检查。
• 自 17:12Z 以来无新增 review、inline thread 或 requested changes，无需修改。
• 全部门禁满足后已执行负责人交付：18:03:40Z 向 XuPeng-SH 发出 review 请求（工具返回成功）；18:04:38Z 将 PR 从 draft 标记为 ready（API 读回 isDraft=false）；18:05:13Z 将 PR assign 给 XuPeng-SH（API 读回 assignees 含 XuPeng-SH）。Issue [Bug]: OpenClaw snapshot discovery backfills historical counts it discards #354 此前已 assign 并读回由 XuPeng-SH 负责。
• 此前全部证据保持有效：RED/GREEN 回归（修复前同一回归 1 failed/4 passed，修复后 234/234 与打包 6 文件，双 Node 版本）、真实 HTTP 回归 5/5、独立 review agent 明确 Approve、两文件 +8/-2、diff --check 通过。本地测试不能替代完整 CI，本轮以上游实际 run 结果为准。
• 不自动合并、不自动关闭 Issue，等待负责人审阅。分支与提交仅在本人 fork（MatrixAdventurer/memoria:fix/openclaw-snapshot-counts-20261022），未写上游任何默认分支。以上只读核验与写操作均由署名主体通过 GitHub 连接器执行。

[loveRhythm1990]

LGTM. Reviewed both changed files at 2a613d0be81932a1b5fee791ef53e0df4a9abafa and traced the client/server call paths.

The issue is real: snapshot listing defaults to resolving missing historical memory counts, while OpenClaw snapshot discovery and stats().snapshotCount only consume snapshot metadata and the number of snapshots. Passing resolve_memory_counts=false on every automatically fetched page skips that unused work without changing pagination or the displayed results. Explicit limit/offset requests retain their existing behavior.

Validation:

• All 234 OpenClaw tests passed locally; package build and verification passed (6 packed files).
• With the base implementation restored, the new regression assertion fails as expected because all six discovery/stats requests omit the flag.
• Verified the server accepts the flag and bypasses historical count resolution when it is false. Rust unit/DB tests and OpenClaw CI passed on this head; database tests were not rerun locally.

No correctness or regression issues found. This confirms removal of unnecessary work; no latency benchmark was performed.