[MatrixAdventurer]

由 LW 授权的 ChatGPT Personal 智能体提交。

What type of PR is this?

• fix (bug fix)
• test (regression coverage)
• docs (documentation)

Which issue(s) this PR fixes

Related to #351. No automatic issue closure is requested.

What this PR does / why we need it

OpenClaw advertises “List all known memory snapshots”, but its public client requests only the first 20 REST rows and discards has_more. Older recovery points disappear from discovery, and stats reports the first page's length as the snapshot count.

Follow the existing offset/has_more pages for public discovery, bounded by maxListPages. Report an error on a truncated, empty-but-has-more or failed scan; the existing stats error path then reports an unavailable count. Explicit transport-level limit/offset calls and older unpaginated response shapes retain their existing behavior.

Verification

• Base: main 659a7779c1f3be787c4dba1d67cd12b998dd0b32; no exact upstream dev branch exists, so main is used as authorized.
• Published head: ad2096906ecff33f53bf75d066f60471754ce3e3.
• RED: new real-loopback-HTTP regression against the original transport: 4 failed / 1 passed. A 45-snapshot fixture exposes only release_0 through release_19 before the fix.
• GREEN: npm test -- openclaw/__tests__/snapshot-pagination.test.ts: 5/5; npm test: 220/220 across 14 files; npm run test:pack: build and package validation succeed, 6 packed files. Executed on Node 24.19.0 and repeated full suite/package on CI's Node 22.23.3 series.
• Independent review: Approve on published head ad2096906ecff33f53bf75d066f60471754ce3e3. The reviewer independently fetched the remote head and verified complete tree e60274553ccd19ff7e87b8efe791062590a38bc4 equals reviewed/tested local c0aecfba77ae2b259e9273b99c4c5988c27f1168, with an empty full-tree diff and independently run 5/5 real-HTTP regression.
• git diff --check passes. Scope: 3 files, +142/-4.
• Upstream OpenClaw CI 39590364118 SUCCESS: Node 22.23.3, regression 5/5, full suite 220/220 across 14 files, package build/6 packed files verified. PR Title Check 39590364230 SUCCESS. Rust/MatrixOne CI 39590364194: Check & Clippy SUCCESS, Unit Tests SUCCESS (116+22+45+69 = 252 passed); DB Tests SUCCESS (completed 2026-10-21T11:38:07Z); the complete Rust/MatrixOne workflow is SUCCESS. PR is ready for review, assigned to XuPeng-SH and review requested (read back). No live Memoria or local Rust/DB integration run is claimed.
• Git HTTPS push lacked credentials; publication used the authenticated GitHub connector on the verified user's MatrixAdventurer/memoria:fix/openclaw-snapshot-pagination-20261021 branch. Destination main was not written. Branch creation/file commits and read-only verification were performed by the signed execution subject.
• Generated dist artifacts are build-only and remain outside the source diff. No schema, dependencies or server behavior changes.

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记录：head ad2096906ecff33f53bf75d066f60471754ce3e3。独立 review agent 获取远程提交后确认完整 tree 为 e60274553ccd19ff7e87b8efe791062590a38bc4，与已实测本地 c0aecfba77ae2b259e9273b99c4c5988c27f1168 完全相同，全树 diff 为空，明确 Approve。完整修复 3 文件 +142/-4，全部 commit 保留指定署名。

RED（原始 transport + 新真实 HTTP 回归）：4 failed / 1 passed。GREEN：回归5/5，插件全量220/220，打包校验6文件；Node24.19.0和CI对应Node22.23.3均执行全量与打包成功。上游 Rust/MatrixOne CI 仍待实际结果，不以局部测试代替完整 CI。Issue #351 已指派 XuPeng-SH；PR 转 ready、指派与请求审阅将在所需 CI 通过后执行。分支创建、连接器文件提交与只读核验主体均为上述智能体。

[MatrixAdventurer]

由 LW 授权的 ChatGPT Personal 智能体提交。

本轮 CI 续接证据（head ad2096906ecff33f53bf75d066f60471754ce3e3 未变化）：

• OpenClaw #39590364118 SUCCESS，已读取原始日志确认 Node22.23.3，新增真实HTTP回归5/5、全量220/220、打包6文件通过。
• PR Title Check SUCCESS。
• Rust/MatrixOne #39590364194：Check & Clippy SUCCESS，Unit Tests SUCCESS，原始日志四组116+22+45+69共252通过；DB Tests（job119190198812）仍在 All tests 步骤运行，没有失败结论，本轮未取消或重跑。
• 最终独立复审已 Approve，完整树一致性证据见前一条评论和正文。
• Issue [Bug]: OpenClaw snapshot discovery silently stops at 20 snapshots #351 的 assignee XuPeng-SH 已回读确认。PR 仍 draft；DB及余下规定步骤完成并通过后再转 ready、指派PR并请求该负责人审阅，不把局部通过说成完整CI。

只读核验及本条交付记录均由上述智能体执行。

[MatrixAdventurer]

由 LW 授权的 MOI 智能体提交。

本轮续接完成 PR #352 的收尾（此前记录：等待 DB Tests 及余下规定步骤通过后再转 ready、指派并请求审阅）：

• CI 已全部结束且通过：Rust/MatrixOne 39590364194 的 DB Tests（job 119190198812）于 2026-10-21T11:38:07Z 完成，conclusion=success；同 run 的 Check & Clippy、Unit Tests（252 passed）此前已成功；OpenClaw 39590364118、PR Title Check 均 SUCCESS。head 仍为 ad2096906ecff33f53bf75d066f60471754ce3e3，本轮无新 push。
• Mergify Merge Queue 报 failure 属于合入队列在获得审阅前的预期门禁结果，不是测试失败，未因此改动代码。
• 本轮读取确认：该 PR 无任何人工 review/requested changes，讨论仅含此前两轮交付记录；Issue [Bug]: OpenClaw snapshot discovery silently stops at 20 snapshots #351 的 assignee 已在前轮回读确认为 XuPeng-SH。
• 已执行并回读确认：PR 由 draft 转为 ready（回读 isDraft=false）；PR assignee 置为 XuPeng-SH（update 返回 assignees=["XuPeng-SH"]）；已向 XuPeng-SH 发起 review 请求。
• 目标分支记录不变：上游本仓库无 dev 分支，基线为 main 659a7779c1f3be787c4dba1d67cd12b998dd0b32（已在正文记录阻塞原因）。

等待负责人审阅结论；本智能体不合并、不关闭 Issue、不推进部署。

[loveRhythm1990]

Code review: LGTM on ad2096906ecff33f53bf75d066f60471754ce3e3, with a non-blocking performance suggestion below.

The reported first-page truncation is real. The server's snapshot endpoint computes has_more from the offset and returned row count, so advancing by items.length is correct. The bounded loop terminates correctly at an exact page boundary, rejects incomplete scans, and lets the existing stats error path report snapshotCount: null. Explicit page requests and legacy first-response formats keep their prior behavior.

Validation: all 220 existing OpenClaw tests and package build/verification passed locally. The unchanged baseline transport fails four of the five new regression cases. I also ran 10 additional checks through the actual registered memory_snapshots/memory_stats tools with native loopback HTTP: empty lists, 20/40/45 records, exact page-budget boundaries, cap errors, malformed later pages, short-page offset progression, and legacy array/items/text responses all passed. Rust/MatrixOne CI, including the explicit restore/cleanup regressions, passed on this head. No local database integration run is claimed.

Non-blocking performance suggestion — plugins/openclaw/openclaw/http-client.ts, automatic-pagination request at line 302: pass resolve_memory_counts=false for discovery. ListSnapshotsQuery defaults this flag to true, and snapshot_summary_value runs a historical count_active_memories_at_snapshot for each uncached snapshot count, persisting registered results. The OpenClaw formatter/client discard memory_count, and stats only counts the snapshots themselves. This PR expands the work from the first 20 records to as many as 400 with the default page budget; when those counts are uncached, it can trigger hundreds of unnecessary historical counts. The existing server flag avoids that work without losing any fields this client uses. This is a code-path performance concern, not a measured production slowdown. Keep explicit low-level page behavior unchanged and add a request-query assertion for the automatic scan if taking this suggestion.

CI clarification: the recorded Mergify failure is The current Mergify configuration is invalid at queue_rules[0].autoqueue, not an expected failure for missing reviews. It is separate from this pagination change and the passing application tests.

Coverage: 3/3 changed files reviewed (100%), 0 skipped. No blocking correctness issues found.