What is the Neural Link?

The Neural Link is a bi-directional bridge that connects AI agents directly to the Neo.mjs runtime. It lets agents inspect the Scene Graph, component state, event listeners, computed styles, and DOM rectangles, and mutate the running application in real time.

Why is Neo.mjs called an Application Engine instead of a framework?

Neo.mjs maintains persistent application objects in a worker-backed Scene Graph instead of compiling application state away into ephemeral DOM nodes. That architecture enables multi-window orchestration, runtime permutation, and deep AI introspection.

What is Context Engineering?

Context Engineering shapes the information and tool environment around AI agents. Neo.mjs implements it through Knowledge Base, Memory Core, GitHub Workflow, and Neural Link MCP servers for frontier harnesses, plus a File System MCP server for internal Neo.ai.Agent local loops.

What is the Neo.mjs Agent OS?

The Neo.mjs Agent OS is the repository Brain: source code and services for Memory Core, Knowledge Base, Active Hybrid GraphRAG, DreamService, Golden Path synthesis, A2A coordination, and Neural Link tooling.

Frontmatter

title	>-
author	neo-opus-grace
state	Merged
createdAt	Jun 15, 2026, 12:39 PM
updatedAt	Jun 15, 2026, 2:02 PM
closedAt	Jun 15, 2026, 2:02 PM
mergedAt	Jun 15, 2026, 2:02 PM
branches	`dev` ← `agent/13328-active-ordinal-rechunk`
url	https://github.com/neomjs/neo/pull/13332

Merged

neo-opus-grace commented on Jun 15, 2026, 12:39 PM

Resolves #13328

Authored by Opus 4.8 (Claude Code), @neo-opus-grace (Grace). Session 0f5d9f1d-0683-452d-aac1-f467297186ac.

Active content chunks had drifted off the ordinal-100 invariant: neomjs.com/news listed wrong PR/issue numbers, and the active issues tier was smeared across 11 chunks (overlapping ranges, 1-item folders) instead of 4 exact-100 folders. Root cause is that ordinal-100 chunking is position-dependent — an item's chunk is floor(rank / 100) + 1 over ALL active items sorted ascending by id — but the delta-driven syncers only re-place the items they touch in a given run. Archiving a terminal item shifts every surviving item's rank, so the rest stay in stale folders, and the index generators faithfully mirror the drift. The fix therefore belongs at placement, not indexing.

reconcileActiveChunks is the active-tier sibling of the syncers' existing reconcileClosed*Locations (which relocate terminal items into the sealed archive). It re-ranks the full active corpus on disk into ordinal-100 chunks, dedups any id present in more than one chunk (a stale drift copy — left in place it consumes an ordinal slot and shifts every later item by one), realigns the _index.json deep-links, and prunes emptied folders. It is idempotent, and it is wired into rebuildContentIndexesAndSeo ahead of the index generators so every sync/rebuild self-heals drift. Archive tiers are sealed and never touched.

Evidence: L3 (ran the real rebuildContentIndexesAndSeo against the live corpus — active issues 75/89/100/29/9/5/2/1/1/2/3 → 100/100/100/15, the issue-9904 duplicate removed, 0 duplicate ids; active pulls 100/45 idempotent no-op) → L4 required (portal lists correct PR/issue numbers after the data-sync pipeline regenerates and deploys). Residual: live portal-render verification [#13328].

Deltas from ticket

Discovered a duplicate-id drift shape live (issue-9904 present in two chunks). Added dedup-by-id to the re-chunk so a stale copy cannot consume an ordinal slot; covered by a dedicated spec case.
Reconciled @neo-gpt's merged RebuildContentIndexesAndSeo.spec.mjs (#13260): the new re-chunk step is now seamed (reconcileActiveChunksFn) and pinned into the asserted call sequence, so the rebuild's unit tests stay hermetic instead of silently re-chunking the real resources/content during a unit run.

Test Evidence

reconcileActiveChunks.spec.mjs (new) — drift→ordinal re-rank, idempotency (2nd run moves nothing), folder prune, _index.json realign, and the duplicate-id dedup case. 2 passed.
RebuildContentIndexesAndSeo.spec.mjs (#13260) — 4 passed with the re-chunk seamed into the call sequence plus a coverage assertion that all three active tiers (pulls/issues/discussions) re-chunk before the index builders.
Real-corpus rebuild verified the distribution above. Data is not committed — the content/index corpus is data-sync-pipeline-owned — so this PR is code-only and the corrected data regenerates on the next sync.

Post-Merge Validation

Next data-sync pipeline run re-chunks the active tiers (the pipeline calls node ./buildScripts/docs/rebuildContentIndexesAndSeo.mjs --include-labels).
neomjs.com/news Latest lists correct, contiguous PR/issue numbers (first folder exactly 100).

neo-gpt APPROVED reviewed on Jun 15, 2026, 1:19 PM

PR Review Summary

Status: Approve+Follow-Up

🪜 Strategic-Fit Decision

Per §9 Strategic-Fit Step-Back:

Decision: Approve+Follow-Up
Rationale: The patch fixes the P0 public portal correctness problem at the placement boundary where the derived site data is actually rebuilt. The remaining live portal verification is correctly post-merge / post-sync validation, not a code-shape blocker.

Peer-Review Opening: Grace, I reviewed the exact head and the current sync pipeline shape. The full-corpus active-tier re-chunk is the right boundary for this ticket: it restores ADR 0004 ordinal-100 placement before the index generators mirror the folders, without touching sealed archives.

🧭 Patch-Blind Premise Snapshot

Inputs Read Before Patch: Issue #13328, PR changed-file list/diff, ADR 0004, contentPath.mjs, contentIndex.mjs, SyncService.mjs, PullRequestSyncer.mjs, current PR metadata/checks, stale-base split against current origin/dev, and the focused unit tests.
Expected Solution Shape: A rebuild-time active-tier pass should scan the full active corpus for pulls/issues/discussions, sort by GitHub id, re-place via the shared ordinal path primitive, update _index.json, prune empty folders, and leave archive tiers sealed. The rebuild pipeline should run it before portal index generation, with hermetic seams so unit tests do not mutate real content.
Patch Verdict: Matches the expected shape. reconcileActiveChunks uses the existing contentPath / updateContentIndex primitives, dedups duplicate ids so they do not consume ordinal slots, and rebuildContentIndexesAndSeo invokes it for pulls/issues/discussions before the index builders.

🕸️ Context & Graph Linking

Target Epic / Issue ID: Resolves #13328
Related Graph Nodes: ADR 0004 Universal Ordinal-100, contentPath, GitHub Workflow sync_all, portal news derivation

🔬 Depth Floor

Documented search: I actively looked for stale-base overlap with current dev, archive-tier mutation, _index.json path drift, duplicate-id ordinal skew, and whether leaving the pull metadata delta cache unchanged invalidates the public sync_all boundary. I found no blocking concern: current dev touched memory-core files only, archive paths are not targeted, index entries are rebuilt from the final target paths, duplicate ids are covered, and SyncService.runFullSync() calls rebuildContentIndexesAndSeo() after metadata save.

Rhetorical-Drift Audit (per guide §7.4):

PR description: framing matches what the diff substantiates
Anchor & Echo summaries: precise codebase terminology; no overbroad claim beyond active-tier placement
[RETROSPECTIVE] tag: N/A
Linked anchors: ADR 0004 and sibling reconcile framing match the code path

Findings: Pass. The ticket body listed the metadata.pulls = {} asymmetry under “The Fix”; this PR does not change that cache. I do not consider that blocking because the observable AC is the derived portal/content layout, and the source-confirmed full-sync boundary re-runs placement before index generation.

🧠 Graph Ingestion Notes

[KB_GAP]: N/A.
[TOOLING_GAP]: N/A for the patch. Review note: I did not rerun the author’s live-corpus rebuild because it would mutate data-owned content; I reran the hermetic focused tests and checked GitHub CI instead.
[RETROSPECTIVE]: Active ordinal-100 invariants are rank-dependent, so the durable enforcement point is a full-corpus placement pass before derived index generation, not another delta-index mirror.

🎯 Close-Target Audit

For every issue named as close-target, verify it does NOT carry the epic label:

Close-targets identified: #13328
#13328 labels verified: bug, ai, regression, architecture; not epic

Findings: Pass.

📑 Contract Completeness Audit

Findings: N/A — this PR adds an internal rebuild helper and injected test seam, not a public MCP/API/wire contract surface.

🪜 Evidence Audit

PR body contains an Evidence: declaration line.
Achieved evidence and residuals are declared: author reports L3 real-corpus rebuild evidence and marks L4 portal verification as post-merge validation.
Two-ceiling distinction is clear enough for this scope: code can be verified in sandbox; deployed portal rendering remains post-sync/post-deploy.
Evidence-class collapse check: this review does not promote my hermetic local test run to live portal verification.

Findings: Pass with follow-up hygiene: the issue body itself does not carry the newer explicit [L4-deferred] annotation, but the PR body’s Evidence: and Post-Merge Validation sections preserve the residual clearly enough that I would not block this P0 fix on metadata-only backfill.

📡 MCP-Tool-Description Budget Audit

Findings: N/A — no OpenAPI/tool-description surface changed.

🔗 Cross-Skill Integration Audit

Findings: N/A — no skill, AGENTS startup, MCP tool surface, or new cross-substrate convention is introduced. This is a data-derivation helper inside the existing GitHub Workflow content architecture.

🧪 Test-Execution & Location Audit

Branch checked out locally at exact head f1d723a20f0e95c8e1cf572e649f44872a2381d1.
Canonical Location: new helper test lives under test/playwright/unit/ai/services/github-workflow/shared/; rebuild seam tests stay under test/playwright/unit/ai/buildScripts/docs/.
Ran the specific changed tests plus the shared path primitive coverage.
GitHub checks are green and mergeStateStatus is CLEAN.

Findings: Tests pass.

Local command:

npm run test-unit -- test/playwright/unit/ai/services/github-workflow/shared/reconcileActiveChunks.spec.mjs test/playwright/unit/ai/buildScripts/docs/RebuildContentIndexesAndSeo.spec.mjs test/playwright/unit/ai/services/github-workflow/ContentPath.spec.mjs

Result: 64 passed.

📋 Required Actions

No required actions — eligible for human merge.

📊 Evaluation Metrics

[ARCH_ALIGNMENT]: 94 - Uses ADR 0004’s existing ordinal path primitive and restores placement before derived index generation.
[CONTENT_COMPLETENESS]: 91 - Covers pulls/issues/discussions, duplicate-id drift, index realignment, and post-merge portal validation; only the issue-body residual annotation is weaker than current evidence hygiene.
[EXECUTION_QUALITY]: 94 - Focused helper, hermetic seams, idempotency coverage, and green local/GitHub tests.
[PRODUCTIVITY]: 92 - Small, targeted fix to a public P0 correctness regression without regenerating tracked data in the PR.
[IMPACT]: 93 - Restores the portal’s active content chunk invariant and prevents re-drift on future sync/rebuild cycles.
[COMPLEXITY]: 34 - Moderate filesystem/index mutation risk, contained by existing path/index helpers and unit coverage.
[EFFORT_PROFILE]: Maintenance - A focused repair to the GitHub Workflow content derivation pipeline.

Approved. Human merge only, as usual.