chore: add ai harness rules#1843
Open
CarltonXiang wants to merge 17 commits into
Open
Conversation
Co-authored-by: harvey_xiang <harvey_xiang22@163.com>
…1861) * fix: respect autoInstall=false by removing install_recommended bypass When skillEvolution.autoInstall is set to false, the install_recommended path in autoInstallIfNeeded() was still triggering automatic installation. This fix ensures autoInstall=false completely disables all auto-install behavior. Closes #1398 Co-authored-by: Cursor <cursoragent@cursor.com> * feat: add harness (#1836) Co-authored-by: harvey_xiang <harvey_xiang22@163.com> * chore: update gitignore * fix: respect autoInstall: false config in all auto-install paths - Remove install_recommended bypass that ignored autoInstall config - Consolidate logic: check autoInstall once at method entry - Add unit tests verifying autoInstall: false blocks all installations Fixes #1398 * fix: respect autoInstall=false config in skill evolution - Fix autoInstallIfNeeded to skip ALL auto-installation when autoInstall=false - Previously install_recommended path bypassed autoInstall config check - Add comprehensive test coverage for autoInstall behavior (false/true/default) - Closes #1398 --------- Co-authored-by: jiachengzhen <jiacz@memtensor.cn> Co-authored-by: Cursor <cursoragent@cursor.com> Co-authored-by: HarveyXiang <harvey_xiang@163.com> Co-authored-by: harvey_xiang <harvey_xiang22@163.com> Co-authored-by: MemOS AutoDev <autodev@memos.dev>
…in Viewer UI (#1865) * fix(memos-local-plugin): guard OpenClaw runtime startup * fix:The failed task was wrongly recorded as a "successful experience". * fix(openclaw): correct Alibaba Bailian endpoint default in Viewer UI Change incorrect endpoint from https://dashscope.aliyuncs.com/compatible-mode/v1 to correct endpoint https://coding.dashscope.aliyuncs.com/v1 in both packages/memos-core and apps/memos-local-openclaw viewer files. Fixes #1489 --------- Co-authored-by: jiang <fdjzy@qq.com> Co-authored-by: Jiang <33757498+hijzy@users.noreply.github.com> Co-authored-by: 黑布林 <11641432+heiheiyouyou@user.noreply.gitee.com> Co-authored-by: Dubberman <48425266+whipser030@users.noreply.github.com> Co-authored-by: AutoDev Agent <autodev@memtensor.com>
…started outside Python (#1866) * fix(memos-local-plugin): guard OpenClaw runtime startup * fix:The failed task was wrongly recorded as a "successful experience". * fix: add --home CLI flag for Docker deployment config path - Add --home parameter to bridge.cts for explicit config directory - Enhance config missing warning with remediation steps and docs link - Log config warnings during bootstrap - Add comprehensive Docker deployment documentation - Update README with configuration and troubleshooting guides Fixes #1520 --------- Co-authored-by: jiang <fdjzy@qq.com> Co-authored-by: Jiang <33757498+hijzy@users.noreply.github.com> Co-authored-by: 黑布林 <11641432+heiheiyouyou@user.noreply.gitee.com> Co-authored-by: Dubberman <48425266+whipser030@users.noreply.github.com> Co-authored-by: MemOS AutoDev <autodev@memos.dev>
…ort — skipped traces s (#1870) fix: exclude short traces from embedding maintenance stats - Add shouldTraceHaveEmbeddings() helper to filter traces with insufficient content - Skip traces where both user_text and agent_text are under 10 chars - Skip traces where total combined length is under 20 chars - Fixes misleading 'missing' count after bulk import (issue #1746) - Applies filter consistently to stats computation and repair operations Co-authored-by: AutoDev Agent <autodev@memtensor.com>
…tored as user chunks (#1879) fix: prevent system messages from being stored as user memory chunks - System messages (including internal review prompts) were incorrectly stored as LongTermMemory chunks, polluting user's retrievable memory - Modified SystemParser.parse_fast() to return empty list for non-tool-schema system messages - Tool schema handling remains unchanged (stored as ToolSchemaMemory) - Added comprehensive test coverage for system message handling Fixes #1518 Co-authored-by: MemOS AutoDev <autodev@memos.ai>
docs(memos-local-plugin): clarify install path and stale dir names (#1540) The README's 'Quick start' section told users to use install.sh instead of npm install, but the warning was buried and users still tried 'npm install -g @memtensor/memos-local-plugin' first. The reporter in #1540 encountered this on a Hermes deployment. This change: - Promotes the 'do not run npm install -g' notice to a prominent IMPORTANT callout explaining why global install is wrong (no agent-home deploy, no config.yaml, no bridge/viewer) and that the tarball intentionally ships built artifacts only. - Adds a Troubleshooting subsection covering the two specific symptoms in the bug report: the 'package not found' misread, and the stale web/ and site/ directory names (web/ is now viewer/, site/ was removed by commit 26e7e3d). - Mentions install.ps1 for Windows alongside install.sh. - CHANGELOG: record the docs fix and reference #1540. Documentation-only change; no code or runtime behavior touched. Co-authored-by: MemOS AutoDev <autodev@memtensor.ai>
#1877) fix: increase trace pagination limit from 500 to 10,000 Fixes #1593 - Web UI memory count stuck at 500 Root cause: Multiple layers enforced a 500-item hard cap on trace queries, preventing accurate metrics calculation when trace count exceeded 500. Changes: - apps/memos-local-plugin/core/storage/repos/_helpers.ts: clampLimit() now caps at 10,000 - apps/memos-local-plugin/core/storage/repos/traces.ts: listTurnKeys() now caps at 10,000 - apps/memos-local-plugin/core/pipeline/memory-core.ts: listTraces() now caps at 10,000 The metrics() function requests 10,000 traces to calculate sessions, embeddings, writesToday, and dailyWrites statistics. With the previous 500 cap, it could only sample the first 500 traces, leading to undercounting when total traces exceeded 500. Impact: - Default pagination limit (50) unchanged - High-limit requests (e.g., metrics calculation) can now fetch up to 10,000 items - Existing indexes on ts and (episode_id, turn_id) support efficient 10K queries - Backward compatible with all existing API clients Co-authored-by: AutoDev Bot <autodev@memtensor.ai>
…er called in QClaw des (#1880) fix(memos-local-openclaw): connect to Hub on plugin load (QClaw desktop) The QClaw desktop app loads this plugin without calling service.start(), which used to mean connectToHub() inside startServiceCore() never ran. The Hub Server saw zero incoming connections from QClaw users and team sharing was completely non-functional, even though local memory worked. Split the client-side Hub connection out of startServiceCore() into a small idempotent helper and fire it eagerly at the end of register(), independent of the host's service lifecycle. The gateway CLI path still calls it via service.start() and the guard prevents a duplicate attempt. Fixes #1612 Cherry-picked from commit a645ddd Co-authored-by: MemOS AutoDev <autodev@memtensor.local>
…id semantics (#1876) feat: expose cube creation in HTTP API + clarify mem_cube_id semantics - Add CreateCubeRequest/Response models for cube creation - Add RegisterCubeRequest/Response models for cube registration - Create CubeHandler with create_cube() and register_cube() methods - Add POST /product/create_cube endpoint - Add POST /product/register_cube endpoint - Update API documentation to clarify cube_id vs mem_cube_id equivalence - Add integration tests in tests/api/test_cube_endpoints.py - All code passes Ruff linting Closes #1681 Co-authored-by: MemOS AutoDev <autodev@memos.ai>
…gin on Node 25 (#1875) fix: upgrade better-sqlite3 to 12.10.0 for Node.js 25 support - Upgrade better-sqlite3 from 12.6.3 to 12.10.0 in memos-local-plugin - Upgrade better-sqlite3 from 12.6.3 to 12.10.0 in memos-local-openclaw - Remove Node.js <25.0.0 constraint from memos-local-openclaw engines better-sqlite3 12.10.0 includes prebuilt native binaries for Node.js 25, resolving ABI mismatch errors on the latest Node.js version. Fixes #1734 Co-authored-by: MemOS AutoDev <autodev@memtensor.com>
…pisodes never scored (#1872) fix: include abandoned episodes in reward pipeline and add periodic rescore - Add closeReason === 'abandoned' to episodeRewardIsDirty() check - Add 10-minute periodic timer in init() to rescore dirty episodes - Timer uses unref() to prevent blocking shutdown - Fixes #1782: 98% of closed episodes were skipped (219 of 224 had closeReason: 'abandoned') Before: Only 7 episodes scored on bootstrap (stale open episodes) After: All 219 abandoned episodes eligible for scoring Co-authored-by: AutoDev Bot <autodev@memtensor.com>
…00% CPU on databases > (#1871) fix(memos-local-plugin): avoid bootstrap hang on large databases (#1787) The `namespace-visibility` migration was issuing a blanket `UPDATE ${table} SET share_scope='private' WHERE share_scope IS NULL` against every owner-aware table — including the `traces` table, which on busy installs is the largest, fattest table in the database. On databases past ~500 MB, that UPDATE held the synchronous bootstrap transaction in CPU-bound row rewriting (re-validating the JSON CHECK constraints on every row) for many minutes and never reached `migrations.summary`, manifesting as the regression filed in #1787: bridge process burns 80–157 % CPU after `sqlite.open` and never becomes healthy. The read path already normalises NULL share_scope to 'private' via `normalizeShareScope` and `COALESCE(share_scope, 'private')` in `visibilityWhere`, and new rows pick up the column DEFAULT, so the bulk UPDATE was cosmetic. Dropping it removes the bootstrap-time row rewrite entirely. The same issue also showed up in `memory-core.init()`'s startup "dirty-closed-episode" scan, which called `getManyByIds(traceIds).some(tr => tr.ts > scoredAt)` — hydrating every column of every trace (embedding BLOBs, full `tool_calls_json`, agent text) into Node memory just to inspect a single number for up to 500 episodes. Replaced with a new `traces.hasAnyNewerThan(ids, ts)` helper that issues a single `SELECT 1 ... LIMIT 1` per chunk. Tests: - Added a regression test in `tests/unit/storage/migrator.test.ts` that pre-seeds rows with NULL `share_scope` and asserts they stay NULL after migration 007 (would flip back to 'private' if the bulk UPDATE returned). - Added coverage for `traces.hasAnyNewerThan` in `tests/unit/storage/repos.test.ts`. Fixes #1787 Co-authored-by: MemOS AutoDev <autodev@memtensor.cn>
… (false positive) (#1874) fix: openclaw doctor reports false positive DuplicateOpenClawRuntimeError Add diagnostic mode detection to skip runtime lock acquisition when openclaw doctor is running alongside the gateway. Changes: - Add skipLock parameter to AcquireOpenClawRuntimeLockOptions - Modify acquireOpenClawRuntimeLock to return no-op lock when skipLock=true - Add isDiagnosticMode() helper checking OPENCLAW_DIAGNOSTIC_MODE env var and process args - Update register() to detect diagnostic mode and skip lock acquisition - Add unit tests for diagnostic mode skipLock behavior - Add integration test verifying doctor can run alongside gateway Fixes #1873 Co-authored-by: AutoDev Bot <autodev@memtensor.ai>
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.
Description
Please include a summary of the change, the problem it solves, the implementation approach, and relevant context. List any dependencies required for this change.
Related Issue (Required): Fixes #issue_number
Type of change
Please delete options that are not relevant.
How Has This Been Tested?
Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration
Checklist
Reviewer Checklist