feat: implement LCM architecture follow-up (draft) by 100yenadmin · Pull Request #526 · Martian-Engineering/lossless-claw

100yenadmin · 2026-04-28T21:56:03Z

PR #526: Draft integration/oracle branch: reference only, not the merge vehicle

Maintainer Quick Read

Field	Value
Stack position	draft/reference
Logical parent for review	main
Current head	`3b4398007e96daf70203e74081b4789144c77e12`
Current PR state	Draft by design
Merge note	Keep open as a comparison/oracle branch; do not merge by default.
Last body refresh	2026-05-02

Why Maintainers Should Care

This branch shows the broad architecture in one place, which is useful for comparison, but it is not the recommended production merge path. The decomposed PRs above are smaller, better hardened, and easier to review.

Stack Map

flowchart LR
  P516["#516 Temporal spine"] --> P517["#517 Observed work read model"]
  P517 --> P518["#518 Suggestion ledger"]
  P518 --> P537["#537 Deterministic extraction"]
  P537 --> P538["#538 Event observations"]
  P518 --> P539["#539 Opt-in suggestion tools"]
  P538 --> P540["#540 Maintenance wiring"]
  P540 --> P530["#530 Tracker/open state"]
  P538 --> P531["#531 Event episodes"]
  P530 --> P532["#532 Topic/vocabulary"]
  P526["#526 Draft reference"] -. "not the merge path" .-> P516

How It Works

flowchart TD
  A["#526 broad integration branch"] --> B["Architecture/oracle reference"]
  B --> C["Compare against split PRs"]
  C --> D["#516 temporal"]
  C --> E["#517 observed work"]
  C --> F["#518 suggestion ledger"]
  C --> G["#537-#532 follow-up slices"]
  A -. "not default" .-> H["Do not merge unless maintainers choose mega-PR path"]

What Ships In This PR

a full-stack reference branch for reviewers
a comparison point for the decomposed PRs
historical context for the architecture split
a place to inspect the architecture as one combined change

What This PR Intentionally Does Not Do

not the production merge vehicle
not guaranteed to contain the newest split-PR hardening
not recommended for normal merge review

How An Agent Uses It

Use this PR to understand the whole architecture at once.
Use feat: add temporal lcm_recent rollups (1/10) #516/feat: add LCM observed work density (2/10) #517/feat: add experimental LCM task bridge suggestions (3/10) #518/feat: extract observed work from summaries (4/10) #537/feat: add LCM event observations (5/10) #538/feat: add opt-in LCM task suggestion tools (6/10) #539/feat: run LCM observed extraction in maintenance (7/10) #540/feat: track LCM observed work open state (8/10) #530/feat: group LCM event episodes (9/10) #531/feat: add bounded LCM topic filters (10/10) #532 to actually review and merge.
If a behavior differs, treat the split PR as the newer production source unless maintainers decide otherwise.

Engineering Boundaries

LCM is an evidence layer. It does not become Cortex, GBrain, OpenClaw Tasks, or a fuzzy semantic authority.
Recaps and observed states are entry points for investigation, not proof of exact commands, SHAs, paths, timestamps, root cause, or task completion.
Exact claims still require source drilldown through lcm_describe, lcm_expand, lcm_expand_query, raw messages, GitHub, logs, tickets, or the relevant authority.
Source identifiers stay hidden by default unless a tool has an explicit includeSources or debug affordance.
Reads stay read-only. Writes happen through explicit storage, maintenance, extraction, or review-ledger paths for the slice that owns them.

Primary Review Files

Split PRs are the review files; this draft spans the whole LCM architecture.

Size And Review Surface

These are intentionally large LCM changes, so the body separates the GitHub-visible diff from the intended logical review slice. Later PRs target main, which makes the GitHub Files tab include parent-stack code; reviewers should use the logical parent listed above when judging the slice.

View	Compare base	Files	Additions	Deletions
GitHub-visible PR page	`main`	43	12297	44
Intended reviewer slice	`main`	65	12373	2094

Logical slice breakdown:

Category	Files	Additions	Deletions
runtime code	24	6460	927
tests	14	4068	1107
docs/release notes	25	1823	52
metadata/package	2	22	8

Review Concerns Already Folded Into The Code

Fallback message timestamp scans keep mixed-format julianday() correctness and now use matching expression indexes for indexed range/order plans.
This stays draft so maintainers do not accidentally merge the broad integration branch.
The split PRs contain the latest hardening and should be treated as the production path.
The branch remains useful as an oracle for architecture comparison.
No additional comments are needed to understand its role; the body is explicit.

Validation Evidence

GitHub test check: success
Kept draft/reference rather than restacked into production merge path

Suggested Maintainer Review Path

Start with the Why Maintainers Should Care and How It Works sections.
Use Primary Review Files for the logical slice; the GitHub Files tab can look much larger because the public PRs target main while the series is logically stacked.
Check the non-goals before asking whether LCM should create tasks, close tasks, write Cortex memory, or make authoritative project claims.
Review and merge in the order shown in the stack diagram. feat: implement LCM architecture follow-up (draft) #526 stays draft/reference unless maintainers explicitly choose a single large integration branch.

100yenadmin · 2026-04-28T21:57:57Z

Implementation follow-up status:

Head: e428eab
Local validation from /Volumes/LEXAR/repos/lossless-claw-upstream-pr518-20260429-041648:
- npm test -- --run test/lcm-ultimate-architecture.test.ts passed
- npm test -- --run test/rollup-store-builder.test.ts test/lcm-tools.test.ts passed
- npm test passed (48 files, 839 tests)
- npm run build passed
- git diff --check passed
GitHub CI: test passed
GraphQL unresolved review threads: 0

Safety boundaries remain intact: observed-work extraction is deterministic, lcm_event_search is read-only, task suggestion tools are opt-in behind taskBridgeToolsEnabled / LCM_TASK_BRIDGE_TOOLS_ENABLED, and suggestion review only updates the LCM ledger with no external task writes.

Copilot

Pull request overview

Implements the “LCM ultimate architecture” follow-up atop the temporal rollups + observed-work + inert task-bridge stack by adding deterministic extraction during maintain(), new read tools for observed work and event cues, and opt-in inert task-suggestion tooling with default source redaction.

Changes:

Add observed-work extraction + event observation extraction from new leaf summaries during engine maintenance, with per-conversation processing state.
Add new tools: lcm_work_density, lcm_event_search, opt-in lcm_task_suggestions/lcm_task_suggestion_review, plus config-gated lcm_rollup_debug.
Extend migrations, config surfaces, specs/docs, and tests to cover the new LCM read models and tool gating/redaction behavior.

Reviewed changes

Copilot reviewed 36 out of 36 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
test/task-bridge-suggestion-store.test.ts	Adds suggestion-ledger store tests
test/plugin-prompt-hook.test.ts	Updates recall-policy prompt expectations
test/observed-work-store.test.ts	Adds observed-work store/tool tests
test/lcm-ultimate-architecture.test.ts	End-to-end extraction/tool behavior tests
src/tools/lcm-work-density-tool.ts	Implements `lcm_work_density` tool
src/tools/lcm-task-suggestions-tool.ts	Implements inert suggestion/review tools
src/tools/lcm-rollup-debug-tool.ts	Implements `lcm_rollup_debug` tool
src/tools/lcm-event-search-tool.ts	Implements `lcm_event_search` tool
src/timezone-windows.ts	Adds/extends timezone window helpers
src/store/task-bridge-suggestion-store.ts	Adds suggestion ledger store
src/store/rollup-store.ts	Adds rollup store helpers/queries
src/store/observed-work-store.ts	Adds observed-work read model store
src/store/index.ts	Exports new stores/types
src/store/event-observation-store.ts	Adds event observation store
src/rollup-builder.ts	Rollup build logic + state updates
src/plugin/index.ts	Registers new tools + prompt text
src/observed-work-extractor.ts	Deterministic leaf-summary extraction
src/engine.ts	Wires stores/builders into engine + maintain()
src/db/migration.ts	Adds rollup/observed-work/event/suggestion tables
src/db/config.ts	Adds config toggles + env overrides
specs/lcm-ultimate-architecture-plan.md	Adds canonical architecture spec
specs/lcm-temporal-memory-plan.md	Updates temporal-memory plan/map
specs/lcm-task-bridge-option-c-experimental.md	Updates task-bridge spec details
specs/lcm-observed-work-density-option-b.md	Updates observed-work spec details
skills/lossless-claw/references/recall-tools.md	Updates recall tool guidance
skills/lossless-claw/SKILL.md	Updates skill guidance + routing
openclaw.plugin.json	Adds config schema/labels for toggles
docs/configuration.md	Documents rollup debug toggle
README.md	Documents new env vars/toggles
.changeset/lcm-ultimate-architecture.md	Changeset for architecture follow-up
.changeset/lcm-temporal-skill-docs.md	Changeset for skill doc updates
.changeset/lcm-temporal-rollups.md	Changeset for temporal rollups/tools
.changeset/lcm-task-bridge-suggestions.md	Changeset for suggestion ledger
.changeset/lcm-observed-work-density.md	Changeset for observed-work density

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

100yenadmin · 2026-04-29T07:43:10Z

Completion / review-thread triage update for head 9e29aa056c12651bc9f8153d978369041f2f1166.

Implemented and verified in this slice:

Conservative observed-work extraction from leaf summaries, with rowid-based incremental cursoring so same-second summary inserts are not skipped.
Event observation scaffolding for primary/secondary/injected/echo/imported evidence.
Opt-in task suggestion ledger tools only; no OpenClaw task writes.
taskBridgeToolsEnabled documented and typed as a concrete boolean.
lcm_work_density topic schema wording now matches deterministic topic_key / title / rationale matching.
Suggestion re-records preserve reviewed status, review metadata, and original creator; regression test covers dismissed -> re-upsert -> still dismissed.
Bounded follow-up reads: lcm_work_density and task suggestions reject allConversations=true for now.
Rollup maintenance uses bounded daysBack based on last_rollup_check_at instead of fixed 30-day sweeps on every pending rebuild.
lcm_rollup_debug parses SQLite timestamps as UTC before formatting.

Review threads addressed and resolved: 6/6. No unresolved review threads remain.

Verification:

Local: npm test -> 48 files / 847 tests passed.
Local: npm run build -> passed.
Local: git diff --check -> passed.
GitHub CI: test passed.

Scope caveat: this remains LCM evidence/observation/suggestion infrastructure. It does not grant LCM task authority, does not write Cortex/OpenClaw tasks, and keeps exact proof routed through source drilldown.

100yenadmin · 2026-04-29T10:34:16Z

Decomposition status update: this PR is now intentionally a draft/reference branch, not the merge candidate.

The over-broad integration work has been split into reviewable follow-up slices:

feat: extract observed work from summaries (4/10) #537 feat: extract observed work from summaries — deterministic observed-work extraction, rowid/summary cursoring, retry-safe state, and extraction evidence semantics.
feat: add LCM event observations (5/10) #538 feat: add LCM event observations — event observation storage/search, source redaction, first-event ordering primitives.
feat: add opt-in LCM task suggestion tools (6/10) #539 feat: add opt-in LCM task suggestion tools — disabled-by-default preview/review tooling behind taskBridgeToolsEnabled; no external task writes.
feat: run LCM observed extraction in maintenance (7/10) #540 feat: run LCM observed extraction in maintenance — explicit maintenance/write-path wiring with read-vs-write-path tests.
feat: track LCM observed work open state (8/10) #530 recut on the extraction parent — observed open-state/tracker behavior.
feat: group LCM event episodes (9/10) #531 recut on the event-observation parent — event episode grouping.
feat: add bounded LCM topic filters (10/10) #532 recut on the topic/vocabulary parent — bounded topic filters and vocabulary improvements.

The current merge path remains #516 -> #517 -> #518, then the narrow follow-ups above. The old #526 branch stays useful as an integration/reference artifact but should not be merged as-is.

Latest cleanup applied across the split stack:

lcm_recent fallback confidence now reflects actual fallback coverage instead of hardcoding medium.
lcm_work_density now reuses the shared ISO timestamp parser.
unresolved GraphQL review threads are zero across feat: add temporal lcm_recent rollups (1/10) #516, feat: add LCM observed work density (2/10) #517, feat: add experimental LCM task bridge suggestions (3/10) #518, feat: implement LCM architecture follow-up (draft) #526, feat: track LCM observed work open state (8/10) #530, feat: group LCM event episodes (9/10) #531, feat: add bounded LCM topic filters (10/10) #532, feat: extract observed work from summaries (4/10) #537, feat: add LCM event observations (5/10) #538, feat: add opt-in LCM task suggestion tools (6/10) #539, and feat: run LCM observed extraction in maintenance (7/10) #540.

100yenadmin · 2026-04-30T04:07:35Z

Final adversarial pass - keep draft/reference

Score: mergeability 6.5/10 as a production PR, functionality 8.3/10 as an integration oracle, description 9.8/10.

Functionality audit: This branch is green and useful as an end-to-end architecture/reference branch. It now includes the same key hardening from the split stack: temporal aggregate fix, event-extraction atomicity, source caps, preserved-reviewed task suggestion accounting, deterministic date-window tests, and ultimate-architecture topic-key alignment.

Scenario coverage: Demonstrates the full intended architecture in one place: temporal recap, observed-work extraction, event search, task suggestions, topic matching, and maintenance wiring.

Final state: MERGEABLE, GitHub test pass, GraphQL review threads: 0 unresolved, but still draft by design. Local focused suite plus ultimate-architecture tests, full npm test, npm run build, and git diff --check passed on head c191929.

Recommendation: Do not merge #526 by default. Keep it as a draft/reference/oracle branch and land the decomposed production path: #516 -> #517 -> #518 -> #537 -> #538 -> #539 -> #540 -> #530 -> #531 -> #532.

100yenadmin · 2026-05-01T08:53:16Z

Final May 1 hardening/status update:

Merge order: draft/reference — Integration/oracle branch
Head: c191929
Review state: 0 unresolved GraphQL review threads after the latest push
GitHub state: MERGEABLE / CLEAN, test check green
Validation: GitHub CI green; not the merge vehicle
Score: Reference only, not scored as production merge path
Note: Keep draft. The decomposed PRs now carry newer hardening; use feat: implement LCM architecture follow-up (draft) #526 for comparison/architecture traceability, not default merge.

The PR description top note has been refreshed with current head SHA, visible-vs-logical LOC breakdown, scope, caveats, and validation status. LCM remains an evidence layer: recaps, observed work, events, episodes, and suggestions are entry points, not proof or task authority.

Eva added 6 commits April 28, 2026 20:57

feat: add temporal lcm recent rollups

de5ef5e

fix: restore temporal recall prompt parity

5f66495

fix: harden temporal rollup review concerns

dd25423

feat: add observed work density scaffold

da9eb44

feat: add inert task bridge suggestions

99e7440

feat: implement LCM ultimate architecture follow-up

e428eab

100yenadmin marked this pull request as ready for review April 29, 2026 06:46

Copilot AI review requested due to automatic review settings April 29, 2026 06:46

Copilot started reviewing on behalf of 100yenadmin April 29, 2026 06:47 View session

Copilot AI reviewed Apr 29, 2026

View reviewed changes

100yenadmin mentioned this pull request Apr 29, 2026

feat: track LCM observed work open state (8/10) #530

Closed

100yenadmin changed the title ~~feat: implement LCM ultimate architecture follow-up~~ feat: implement LCM architecture follow-up Apr 29, 2026

This was referenced Apr 29, 2026

feat: group LCM event episodes (9/10) #531

Closed

feat: add bounded LCM topic filters (10/10) #532

Closed

Eva added 5 commits April 29, 2026 14:11

fix: harden lcm architecture review paths

89c7dbf

fix: stabilize rollup sweep tests

bdb9d01

fix: bound lcm follow-up reads

b626373

fix: parse rollup debug timestamps as utc

801f75f

test: preserve reviewed task suggestions

9e29aa0

This was referenced Apr 29, 2026

feat: extract observed work from summaries (4/10) #537

Closed

feat: add LCM event observations (5/10) #538

Closed

feat: add opt-in LCM task suggestion tools (6/10) #539

Closed

feat: run LCM observed extraction in maintenance (7/10) #540

Closed

100yenadmin marked this pull request as draft April 29, 2026 10:34

Eva added 16 commits April 30, 2026 00:09

fix: include raw messages in recent fallback

7be6613

docs: preserve temporal handoff package

735fc0b

test: keep fallback footer inside budget

e572fbc

fix: harden task bridge suggestion ledger

4598a5c

fix: gate observed work maintenance writes

3a32dba

fix: harden temporal rollup degraded states

cd4367f

fix: avoid task-targeted suggestions without task ids

8f28e64

fix: keep task suggestions inert after review

3ca9236

fix: surface stale temporal coverage gaps

c23104c

fix: harden temporal aggregate rollups

bb0f01b

fix: report preserved task suggestions

de690eb

fix: count preserved task suggestions

1d77bf9

fix: make event extraction atomic

bb5ba19

fix: cap observed work source details

5aa7956

test: stabilize broad rollup draft checks

9e65b24

test: align ultimate draft topic filters

c191929

This was referenced May 1, 2026

feat: add temporal lcm_recent rollups (1/10) #516

Closed

feat: add LCM observed work density (2/10) #517

Closed

feat: add experimental LCM task bridge suggestions (3/10) #518

Closed

fix: index recent fallback timestamp scans

3b43980

100yenadmin changed the title ~~feat: implement LCM architecture follow-up~~ feat: implement LCM architecture follow-up (draft) May 3, 2026

100yenadmin closed this May 6, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: implement LCM architecture follow-up (draft)#526

feat: implement LCM architecture follow-up (draft)#526
100yenadmin wants to merge 28 commits into
Martian-Engineering:mainfrom
electricsheephq:feat/lcm-ultimate-architecture

100yenadmin commented Apr 28, 2026 •

edited

Loading

Uh oh!

100yenadmin commented Apr 28, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

100yenadmin commented Apr 29, 2026

Uh oh!

100yenadmin commented Apr 29, 2026

Uh oh!

100yenadmin commented Apr 30, 2026

Uh oh!

100yenadmin commented May 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

100yenadmin commented Apr 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR #526: Draft integration/oracle branch: reference only, not the merge vehicle

Maintainer Quick Read

Why Maintainers Should Care

Stack Map

How It Works

What Ships In This PR

What This PR Intentionally Does Not Do

How An Agent Uses It

Engineering Boundaries

Primary Review Files

Size And Review Surface

Review Concerns Already Folded Into The Code

Validation Evidence

Suggested Maintainer Review Path

Uh oh!

100yenadmin commented Apr 28, 2026

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!

Uh oh!

100yenadmin commented Apr 29, 2026

Uh oh!

100yenadmin commented Apr 29, 2026

Uh oh!

100yenadmin commented Apr 30, 2026

Final adversarial pass - keep draft/reference

Uh oh!

100yenadmin commented May 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

100yenadmin commented Apr 28, 2026 •

edited

Loading