PR #539 (6/10 rebased): feat: add opt-in LCM task suggestion tools by 100yenadmin · Pull Request #27 · electricsheephq/lossless-claw-test

100yenadmin · 2026-05-03T23:11:19Z

Rebased upstream Martian-Engineering/lossless-claw#539 onto our test fork's main, which already carries the 8 hotfix PRs (Martian-Engineering#572-Martian-Engineering#579) plus PR Martian-Engineering#516.

Resolution strategy: inverse merge

pr-539 was branched from an older main that predates the hotfix stack, so a literal "take pr-539" would delete hotfix code. Started from main and layered pr-539's net-new code on top.

Per-file resolution

File	Strategy
`src/engine.ts`	3-way UNION. Preserved hotfixes (`normalizeSummaryOverlapText`, `messageContentCoveredBySummary`, `deps.clock.now()`, `TableColumnInfo` cast). Added pr-539's `ObservedWorkStore` + `TaskBridgeSuggestionStore` fields, inits, and getters.
`src/db/migration.ts`	Took main; injected pr-539's `ensureObservedWorkTables`, `ensureObservedWorkIndexes`, `ensureTaskBridgeSuggestionTables`, `ensureTaskBridgeSuggestionIndexes` migration steps after `ensureRollupViews`. Also fixed `SummaryColumnInfo` -> `TableColumnInfo` in `addColumnIfMissing` (pr-539 already had the right type).
`src/plugin/index.ts`	Took main; added imports + `registerTool` calls for `createLcmWorkDensityTool` plus the `taskBridgeToolsEnabled`-gated `createLcmTaskSuggestionsTool` / `createLcmTaskSuggestionReviewTool`.
`openclaw.plugin.json`	Union; appended `lcm_task_suggestion_review`, `lcm_task_suggestions`, `lcm_work_density` to `contracts.tools`. Added schema + uiHints for `taskBridgeToolsEnabled` alongside main's rollup token-budget keys.
`src/db/config.ts`	Union; kept rollup token-budget knobs and added `taskBridgeToolsEnabled` config field with `LCM_TASK_BRIDGE_TOOLS_ENABLED` env wiring.
`src/timezone-windows.ts`	Took pr-539's `startOfWeekDayString` export.
`README.md`, `docs/configuration.md`	Added `LCM_TASK_BRIDGE_TOOLS_ENABLED` / `taskBridgeToolsEnabled` documentation.
`test/engine.test.ts`, `test/rollup-store-builder.test.ts`, `src/tools/lcm-recent-tool.ts`, `src/rollup-builder.ts`, `src/store/rollup-store.ts`	Took main (hotfix-only files).
New pr-539 files	Taken cleanly: `observed-work-store.ts`, `task-bridge-suggestion-store.ts`, `lcm-task-suggestions-tool.ts`, `lcm-work-density-tool.ts`, plus tests, specs, changesets.

Validation

0 conflict markers across the worktree
npm run build clean (815.6kb bundle)
npm test: 988/988 passing (53/53 suites)

Net diff

19 files changed, 3161 insertions(+), 9 deletions(-)

Summary by CodeRabbit

New Features
- Added work density tool to report observed work summaries across time periods.
- Added event observation search functionality.
- Added opt-in task suggestion tools with user review capability (non-destructive, suggestion-only).
- Added taskBridgeToolsEnabled and observedWorkMaintenanceEnabled configuration options.
Documentation
- Added specifications for observed work density tracking.
- Added specifications for task bridge suggestion system.
- Updated configuration documentation with new settings.
Tests
- Added comprehensive test coverage for observed work and task bridge systems.

…suggestion tools Rebased upstream PR Martian-Engineering#539 (Martian-Engineering#539) onto main with the 8 hotfix PRs (Martian-Engineering#572-Martian-Engineering#579) + Martian-Engineering#516 already in place. Inverse-merge resolution per refined policy: - engine.ts: 3-way UNION; preserved hotfixes (normalizeSummaryOverlapText, messageContentCoveredBySummary, clock.now(), TableColumnInfo cast) AND added pr-539's ObservedWorkStore + TaskBridgeSuggestionStore fields, inits, and getters. - migration.ts: kept main's Martian-Engineering#569+Martian-Engineering#574 migrations; injected pr-539's new ensureObservedWorkTables, ensureObservedWorkIndexes, ensureTaskBridgeSuggestionTables, ensureTaskBridgeSuggestionIndexes steps after ensureRollupViews. Fixed SummaryColumnInfo->TableColumnInfo in addColumnIfMissing helper. - plugin/index.ts: kept main's tool wiring; added imports + registerTool calls for createLcmWorkDensityTool plus gated createLcmTaskSuggestionsTool / createLcmTaskSuggestionReviewTool. - openclaw.plugin.json: union; appended lcm_task_suggestion_review, lcm_task_suggestions, lcm_work_density to contracts.tools and added schema/uiHints for taskBridgeToolsEnabled while preserving rollup token-budget keys. - db/config.ts: union; both rollup token-budget knobs and taskBridgeToolsEnabled config field + env wiring. - timezone-windows.ts: took pr-539's startOfWeekDayString export. - README.md / docs/configuration.md: added LCM_TASK_BRIDGE_TOOLS_ENABLED and taskBridgeToolsEnabled docs alongside main hotfix doc lines. - test/engine.test.ts, test/rollup-store-builder.test.ts, src/tools/lcm-recent-tool.ts, src/rollup-builder.ts, src/store/rollup-store.ts: took main (hotfix code only). - New pr-539 files (observed-work-store, task-bridge-suggestion-store, lcm-task-suggestions-tool, lcm-work-density-tool, plus tests/specs/ changesets) taken cleanly. Validation: 0 conflict markers, npm run build clean, 988/988 tests pass.

chatgpt-codex-connector · 2026-05-03T23:11:22Z

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.

coderabbitai · 2026-05-03T23:11:26Z

Caution

Review failed

Pull request was closed or merged during review

📝 Walkthrough

Walkthrough

This PR introduces observed-work density tracking, event observations, and a task-bridge suggestion ledger for LCM. New database stores persist extracted work items and events from conversation summaries, enable incremental processing with cursor-based replay guards, and expose read-only density queries and suggestion management via opt-in tools.

Changes

Observed Work, Event Observations, and Task Bridge Suggestions

Layer / File(s)	Summary
Configuration & Schema `src/db/config.ts`, `src/db/migration.ts`, `openclaw.plugin.json`, `docs/configuration.md`, `README.md`	Config adds `observedWorkMaintenanceEnabled` and `taskBridgeToolsEnabled` boolean flags with env overrides. Plugin declares new tools and UI hints. Migrations create `lcm_observed_work_items`, `lcm_observed_work_sources`, `lcm_observed_work_state`, `lcm_event_observations`, and `lcm_task_bridge_suggestions` tables with supporting indexes.
Data Stores `src/store/observed-work-store.ts`, `src/store/event-observation-store.ts`, `src/store/task-bridge-suggestion-store.ts`	Three stores backed by SQLite: `ObservedWorkStore` manages work-item upserts with conflict-aware merging (preserves terminal status, ensures monotonic counters, clamps confidence), source deduplication, processing-cursor state, and density queries with optional source inclusion. `EventObservationStore` validates and normalizes timestamps to UTC ISO-8601, deduplicates source IDs, and executes time/kind/query-filtered observation lookups with bounded results. `TaskBridgeSuggestionStore` validates suggestion inputs, persists ledger entries with review-state transitions (pending → accepted/dismissed), and prevents concurrent races via transactional read-then-write; bulk upsert uses per-row SAVEPOINTs to demote partial failures while preserving batch order.
Extraction Logic `src/observed-work-extractor.ts`, `src/timezone-windows.ts`	`ObservedWorkExtractor` scans leaf summaries using a `(created_at, summary_id, rowid)` cursor to avoid same-second duplication. Applies regex-driven classification to extract work candidates (completion/unfinished/decision cues with negation-aware patterns) and event observations, derives stable hashes (`conversationId:kind:topicKey`) as work-item IDs, computes confidence bands, and upserts work/events with idempotency guards. On failure, marks state `pendingRebuild=true` to force full rescan. Returns aggregate scan/upsert counts. Utility `startOfWeekDayString` computes Monday-aligned dates for period queries.
Engine Wiring `src/engine.ts`	Engine constructor instantiates `EventObservationStore`, `ObservedWorkStore`, `ObservedWorkExtractor`, and `TaskBridgeSuggestionStore`, exposing them via public getters. The `maintain()` flow conditionally runs `observedWorkExtractor.processConversation()` when `config.observedWorkMaintenanceEnabled`, logs extraction metrics, and on failure sets `pendingRebuild=true` and clears the processing cursor.
Tools & Plugin Registration `src/tools/lcm-work-density-tool.ts`, `src/tools/lcm-event-search-tool.ts`, `src/tools/lcm-task-suggestions-tool.ts`, `src/plugin/index.ts`	`lcm_work_density` resolves timezone-aware period bounds (`today`/`week`/`month`/`date:YYYY-MM-DD`), queries density with optional statuses/kinds/topic/confidence filters, and applies output-budget trimming (iteratively removes per-item sources or whole items until JSON size fits `maxOutputTokens`). `lcm_event_search` filters observations by conversation/time/kind/query and returns bounded ranked results. `lcm_task_suggestions` generates deterministic suggestion IDs, derives candidates from observed-work density, and on `record` mode bulk-upserts into the suggestion store (preview mode disables writes). `lcm_task_suggestion_review` updates suggestion review status (accepted/dismissed) while validating ownership. Plugin conditionally registers these tools under their respective feature flags.
Tests & Documentation `test/observed-work-store.test.ts`, `test/task-bridge-suggestion-store.test.ts`, `specs/lcm-observed-work-density-option-b.md`, `specs/lcm-task-bridge-option-c-experimental.md`, `.changeset/*`	Comprehensive test suites verify extractor cursor semantics (no-duplicate same-second handling, rowid drift recovery, fallback-to-rowid when processed-summary ID is missing), evidence redaction, semantic kind preservation, density aggregation, source visibility toggling, review state transitions, and bulk upsert ordering/demotions. End-to-end tool tests exercise parameter validation, period parsing, scope enforcement, and output budget trimming with accounting. Spec documents define thesis, data models, tool contracts, behavioral rules, and implementation scope (inert store + tests; no default-on behavior or task mutation).

Sequence Diagram(s)

Skipped (configuration-heavy changes with independent store operations; no multi-component sequential flows requiring visualization).

Estimated code review effort

🎯 5 (Critical) | ⏱️ ~90 minutes

Rationale: Substantial changes spanning five store implementations with complex SQL (conflict handling, cursor semantics, budget-aware query trimming), regex-driven evidence extraction with replay guards, three tools with parameter validation and normalization, transactional concurrency semantics, and 1900+ lines of heterogeneous tests validating edge cases (rowid drift, same-second summaries, source deduplication, review transitions). Requires careful review of: (1) cursor idempotency & missing-summary-ID fallback logic, (2) conflict-resolution merge semantics (terminal status, monotonic counters, confidence clamping), (3) transactional boundaries in bulk operations, (4) regex classification correctness and negation handling, (5) timestamp canonicalization and timezone-aware period arithmetic, (6) output budget trimming and accounting accuracy.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 4.41% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and specifically describes the primary change: adding opt-in LCM task suggestion tools. It accurately reflects the main feature introduced across the changeset.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

Create stacked PR
Commit on current branch

🧪 Generate unit tests (beta)

Create PR with unit tests
Commit unit tests in branch rebase/539

_{Review rate limit: 0/5 reviews remaining, refill in 57 minutes and 13 seconds.}

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Copilot

Pull request overview

Adds an opt-in “observed work” read model and associated tools for summarizing work density, plus an inert task-suggestion ledger (preview/record + review) that explicitly does not mutate external tasks.

Changes:

Introduces SQLite tables + stores for observed work items/sources/state and task-bridge suggestions.
Adds lcm_work_density tool and opt-in lcm_task_suggestions / lcm_task_suggestion_review tools (gated by taskBridgeToolsEnabled).
Updates plugin wiring, config/env, manifest, docs, and adds comprehensive tests/specs/changesets.

Reviewed changes

Copilot reviewed 19 out of 19 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
`src/db/migration.ts`	Adds observed-work + task-bridge suggestion tables and indexes; small type alias cleanup.
`src/store/observed-work-store.ts`	Implements observed work persistence + density querying with optional source loading and limits.
`src/tools/lcm-work-density-tool.ts`	Exposes observed work density as a tool with period parsing and output-budget trimming.
`src/store/task-bridge-suggestion-store.ts`	Implements inert suggestion ledger (upsert pending, list, review).
`src/tools/lcm-task-suggestions-tool.ts`	Generates/records suggestions from observed work and provides a review tool.
`src/timezone-windows.ts`	Exports `startOfWeekDayString` for Monday-start week windows on `YYYY-MM-DD`.
`src/plugin/index.ts`	Registers `lcm_work_density` and conditionally registers task suggestion tools when enabled.
`src/engine.ts`	Wires new stores into `LcmContextEngine` and exposes getters.
`src/db/config.ts`	Adds `taskBridgeToolsEnabled` with `LCM_TASK_BRIDGE_TOOLS_ENABLED` wiring.
`openclaw.plugin.json`	Adds tool contracts + config schema/uiHints for `taskBridgeToolsEnabled`.
`test/observed-work-store.test.ts`	Tests migrations, density semantics, period handling, source redaction, and output budget trimming.
`test/task-bridge-suggestion-store.test.ts`	Tests suggestion ledger behavior and tool flows without any external task writes.
`docs/configuration.md` / `README.md`	Documents the new opt-in config/env var.
`specs/lcm-observed-work-density-option-b.md`	Design/spec doc for observed work density (Option B).
`specs/lcm-task-bridge-option-c-experimental.md`	Design/spec doc for experimental suggestion-ledger bridge (Option C).
`.changeset/*.md`	Release notes for observed work density and suggestion tooling/storage.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

…erministic outcome)

…GIN IMMEDIATE bypass); add bulk tests Wave 2 fix on PR #27 / rebase/539. Wave 1 (1df72b9) introduced a private runInTransaction helper that called raw BEGIN IMMEDIATE / COMMIT / ROLLBACK on the shared DatabaseSync handle, with a fallback nested-transaction path guarded by `(this.db as unknown as { inTransaction?: boolean }).inTransaction`. Two problems with that: 1. CRITICAL-1 — `node:sqlite`'s DatabaseSync does NOT expose `inTransaction` (that's a better-sqlite3 API). The fallback path was unreachable code. 2. CRITICAL-2 — the helper bypassed `withDatabaseTransaction` from src/transaction-mutex.ts, the per-database async mutex that serializes concurrent transaction entry points sharing the same DB handle. That's the exact race the mutex was built to fix (lossless-claw#260): two async paths racing BEGIN on a shared connection blow up with "cannot start a transaction within a transaction". Fix: replace `runInTransaction` with `withDatabaseTransaction(db, "BEGIN IMMEDIATE", ...)`. That's async, so `bulkUpsertSuggestions` and `upsertSuggestion` become async. The tool callsite (lcm-task-suggestions-tool record path) was already in an async execute() — now awaits the bulk call. Per-row failure semantics (MAJOR-1 in audit): pre-Wave-1, a bad row didn't poison the batch. Post-Wave-1, one FK-deletion race in 50 rows aborted the whole transaction. Restored with per-row SAVEPOINTs inside the outer transaction — a row that throws during INSERT (e.g. concurrent FK-delete of its work item or source evidence) is demoted to a new "skipped" outcome. The result array stays order-preserving so callers can correlate inputs ↔ results positionally. Input validation errors (bad confidence, blank ID, non-pending status, missing taskId for targeted kinds, no source IDs) still throw before the transaction opens — caller gets the same fast-fail. Empty-input early return preserved (no transaction opened for `[]`). Tests (MAJOR-2): added bulkUpsertSuggestions describe block with 4 cases — empty input, mixed insert+refresh, preserved_reviewed determinism with order preservation, mid-batch skipped row leaving good rows committed. Existing upsertSuggestion tests migrated to async/await; validation-error tests use `rejects.toThrow` since async functions convert sync throws to rejections. Total: 992 tests pass (988 baseline + 4 new). Build clean.

Copilot

Pull request overview

Copilot reviewed 19 out of 19 changed files in this pull request and generated 5 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

…ions on top PR #27 (rebase/539) carried older versions of observed-work-store, work-density-tool, and migration.ts that pre-dated PR #20/#21/#22/#25/#28 hardening on release/v0.9.4. This commit takes the release/v0.9.4 versions of those files as the base and layers PR #27's NEW work on top: - KEPT (PR #27 additions, layered on release): - bulkUpsertSuggestions API (uses release's mutex routing + per-row SAVEPOINT) - getSuggestionConversationId helper - taskBridgeToolsEnabled config flag - reviewSuggestion suggestionId trim - lcm_task_suggestions + lcm_task_suggestion_review tools (gated) - TaskBridgeSuggestionStore engine wiring + getter - PORT-FORWARDED from release/v0.9.4: - src/store/observed-work-store.ts (cursor state, getItem, listItems, etc.) - src/store/event-observation-store.ts (NEW from release; wired in engine) - src/observed-work-extractor.ts (NEW from release; NEGATED_COMPLETION_RE, fingerprintVersion, clearProcessingCursor) - src/tools/lcm-work-density-tool.ts (release hardening) - src/tools/lcm-event-search-tool.ts (NEW from release) - src/db/migration.ts (release index set) - src/db/config.ts (observedWorkMaintenanceEnabled flag preserved) - src/plugin/index.ts (release gating + PR #27 task-suggestions blocks) - src/engine.ts (release ObservedWorkExtractor + EventObservationStore wiring, pendingRebuild + clearProcessingCursor on extraction failure) - test/observed-work-store.test.ts (release version aligned with new store API) - MERGED (release base + PR #27 additions): - src/store/task-bridge-suggestion-store.ts: MAX_SUGGESTION_SOURCE_IDS=50 cap, chunked source-id IN(...) lookup, ORDER BY suggestion_id ASC tiebreaker, SQLInputValue typing — all from release; bulkUpsertSuggestions, getSuggestionConversationId, "skipped" outcome, savepoint counter — from PR #27. - src/db/config.ts: both observedWorkMaintenanceEnabled (from release) and taskBridgeToolsEnabled (from PR #27). - src/plugin/index.ts: gates work-density + event-search behind observedWorkMaintenanceEnabled; gates task-suggestions tools behind taskBridgeToolsEnabled. - ADDED to openclaw.plugin.json: - lcm_event_search to contracts.tools (was missing — manifest drift guard caught it) - observedWorkMaintenanceEnabled in uiHints + configSchema Tests: 1006/1006 pass (was 992 before; release hardening added more coverage). Net new tsc errors: 0 (332 vs 342 baseline = 10 fewer).

Conflict resolution: took theirs (rebase/539 port-forwarded release + layered PR #27 on top). New tools: lcm_task_suggestions + lcm_task_suggestion_review (gated behind taskBridgeToolsEnabled flag, default off). New store API: bulkUpsertSuggestions (per-row SAVEPOINT, skipped outcome), getSuggestionConversationId (cross-conversation review prevention). Verified all PR #20-#28 hardening preserved.

100yenadmin · 2026-05-04T11:02:32Z

Merged into release/v0.9.4 at 65a0aaa after Wave 1+2+3 adversarial review and comment resolution. Tracked branch: rebase/539. See release/v0.9.4 for the integrated artifact.

Copilot AI review requested due to automatic review settings May 3, 2026 23:11

Copilot started reviewing on behalf of 100yenadmin May 3, 2026 23:11 View session

Copilot AI reviewed May 3, 2026

View reviewed changes

Comment thread src/tools/lcm-task-suggestions-tool.ts

Comment thread src/tools/lcm-task-suggestions-tool.ts

Comment thread src/store/task-bridge-suggestion-store.ts

Comment thread src/tools/lcm-work-density-tool.ts Outdated

Comment thread src/tools/lcm-work-density-tool.ts

Eva added 2 commits May 4, 2026 06:23

fix(task-suggestions): wrap upserts in transaction (single fsync, det…

1df72b9

…erministic outcome)

100yenadmin requested a review from Copilot May 4, 2026 08:33

Copilot started reviewing on behalf of 100yenadmin May 4, 2026 08:34 View session

Copilot AI reviewed May 4, 2026

View reviewed changes

Comment thread src/tools/lcm-task-suggestions-tool.ts

Comment thread src/plugin/index.ts Outdated

Comment thread src/tools/lcm-work-density-tool.ts Outdated

Comment thread src/tools/lcm-task-suggestions-tool.ts

Comment thread src/tools/lcm-task-suggestions-tool.ts Outdated

review: address comments on PR #27

8f2e186

coderabbitai Bot approved these changes May 4, 2026

View reviewed changes

100yenadmin closed this May 4, 2026

Conversation

100yenadmin commented May 3, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Resolution strategy: inverse merge

Per-file resolution

Validation

Net diff

Summary by CodeRabbit

Uh oh!

chatgpt-codex-connector Bot commented May 3, 2026

Uh oh!

coderabbitai Bot commented May 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review failed

Walkthrough

Changes

Sequence Diagram(s)

Estimated code review effort

❌ Failed checks (1 warning)

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

100yenadmin commented May 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

100yenadmin commented May 3, 2026 •

edited by coderabbitai Bot

Loading

coderabbitai Bot commented May 3, 2026 •

edited

Loading