Repository navigation
fix(openclaw): filter API memory lists before pagination - #357
loveRhythm1990 merged 1 commit into
Conversation
Forward the existing REST memory_type selector on every page and retain the requested matching-row limit. Preserve embedded MCP scan semantics. Add native HTTP regressions for page budgets and partial results. Related to matrixorigin#356. 由 LW 授权的 ChatGPT Personal 智能体提交。
|
由 LW 授权的 ChatGPT Personal 智能体提交。 操作与验证记录：已在用户本人 fork MatrixAdventurer/memoria 的独立分支提交 b144f5b，产品树 dc5b8b41eb11051e8603f90deb74b134487b82a4 与本地测试版本一致。PR 已指派 XuPeng-SH，GitHub 回读确认。上述分支创建、提交、PR 创建、指派及只读核验均由本署名主体执行。 OpenClaw CI 已成功：Node 22，236 项测试及 6 文件打包验证通过。Rust/DB CI 的 cargo check、cargo clippy -- -D warnings 和 252 项 Rust 单元测试通过；真实 MatrixOne 数据库全量测试正在运行，尚不宣称完整 CI 通过。独立 review agent 对最终 3 文件、+60/-4 修复明确 Approve。 PR 暂保持 draft；数据库门禁通过后转 ready 并请求 XuPeng-SH 审阅。不自动合并、部署或关闭 Issue #356。 |
aptend
left a comment
aptend
left a comment
There was a problem hiding this comment.
Reviewed the complete three-file diff at b144f5b and traced the selector through REST routing/service/storage: memory_type is applied before SQL LIMIT/keyset pagination, and the transport preserves it on every page. The matching-row limit, partial flag, cache generation guard, request timeout/error cleanup, and test server teardown remain consistent. No blocking issue found.
Validation: the focused pagination suite completed locally with 15/15 passing on Node 22.22.0. Audited CI job 120143456207: 236/236 tests and build/6-file package verification passed. Its synthetic merge commit and this head have the identical full tree dc5b8b41eb11051e8603f90deb74b134487b82a4. The new HTTP fixtures emulate short pages; this is not a claim of a local live-MatrixOne run.
|
This pull request does not currently match the merge queue conditions, so it cannot be queued from here. The box comes back if it matches again. |
由 LW 授权的 ChatGPT Personal 智能体提交。
What type of PR is this?
Which issue(s) this PR fixes
Related to #356.
What this PR does / why we need it
An API-backed
listMemories({ memoryType: "episodic" })currently spends its page budget scanning unrelated memory types and filters afterward. A one-page request can return no episodic memories even when they exist. It also reports an embedded Rust MCP filtering limitation on complete API results.Forward the existing REST
memory_typeselector on every cursor page and preserve the requested matching-row limit. Keep embedded MCP bounded scanning unchanged. Add two native loopback HTTP regressions for exhausted page budgets and complete multi-page results. The fixture deliberately caps pages at two rows; this is an emulated short-page contract, not a live MatrixOne measurement.Validation and independent review
Base:
e2aa9071f70b0f7a3b2f53d297ffa220015a3d0b.Head:
b144f5b66fc455085ea369ca0821eceae7435a21.Product tree:
dc5b8b41eb11051e8603f90deb74b134487b82a4, identical to the locally tested commit tree.From
plugins/openclaw:npm test -- openclaw/__tests__/list-pagination.test.ts: 2 failed / 13 passed; both new native HTTP regressions failed.npm test: 236 tests passed across 15 files (Node 24.19.0).npm run test:pack: build and packed-package verification passed (6 packed files).git diff --check: passed.Scope: 3 repository files, +60/-4. No dependency/schema/workflow changes or tracked generated files are required. Rust and live MatrixOne checks were run in the upstream PR CI (not locally). All required runs for the fixed head succeeded:
Authorized operation record
Authenticated fork owner
MatrixAdventurer, fork parentmatrixorigin/memoria, and fork push/admin permissions were verified. Upstream has nodevbranch, so this PR targetsmainunder LW's explicit instruction. Created only the independent fork branchfix/openclaw-list-type-filter-20261023and its signed commit; no upstream branch was changed. Issue #356 and PR #357 are assigned toXuPeng-SH(GitHub readback confirmed). After all three CI workflows succeeded for the fixed head, the signed agent marks this PR ready and requestsXuPeng-SHreview; the final operation receipt records readback. No automatic merge, deployment or Issue closure.由 LW 授权的 ChatGPT Personal 智能体提交。