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

id	13328
title	Active content chunks drift off ordinal-100 — portal lists wrong PR/issue numbers
state	Closed
labels	bugairegressionarchitecture
assignees	neo-opus-grace
createdAt	Jun 15, 2026, 12:02 PM
updatedAt	Jun 21, 2026, 2:17 AM
githubUrl	https://github.com/neomjs/neo/issues/13328
author	neo-opus-grace
commentsCount	1
parentIssue	null
subIssues	[]
subIssuesCompleted	0
subIssuesTotal	0
blockedBy	[]
blocking	[]
closedAt	Jun 15, 2026, 2:02 PM

Active content chunks drift off ordinal-100 — portal lists wrong PR/issue numbers

Closed v13.1.0/archive-v13-1-0-chunk-3 bugairegressionarchitecture

neo-opus-grace commented on Jun 15, 2026, 12:02 PM

Authored by Claude Opus 4.8 (Claude Code), @neo-opus-grace (Grace). Operator prio-0 (broken neomjs.com/news portal).

Context

neomjs.com/news/pulls (and tickets) list wrong PR/issue numbers. The chunk folders should be exactly 100 items (ADR 0004 Universal Ordinal-100; folder labels are position-ranges 1-100, 101-200…), but on-disk active chunks are 148/135/180/100/75 — non-uniform, with overlapping/gapped ranges (e.g. resources/content/pulls/chunk-1 holds 148 PRs spanning 9537-12693). This is in the neo repo content directly; the pages deploy just mirrors it.

Root Cause (V-B-A: ADR 0004 §2.2, `contentPath.mjs`, `PullRequestSyncer.mjs`, `IssueSyncer.mjs`, `pulls.mjs`)

Ordinal-100 chunking is position-dependent: a item's chunk = Math.floor(rank/100)+1 where rank is its position among all active items sorted ascending by id. But the syncers only ever rank/re-place the delta:

Stale-rank drift (pulls + issues): syncPullRequests/pullFromGitHub call #planBuckets(metadata, <delta-fetch>) and only relocate the delta items. When a release is cut and a batch is archived, the active set shrinks → every surviving active item's rank shifts → its correct chunk changes — but only delta items move; the rest stay in stale folders. reconcileClosed*Locations relocates terminal items to archive but nothing re-ranks the remaining active tier.
Pulls worse (asymmetry): PullRequestSyncer:492 does metadata.pulls = {} (delta-rebuilt, "not a full index" per its own JSDoc), vs IssueSyncer:601 issues: {...metadata.issues} (cumulative). So even delta pulls' itemIndex is computed over a partial set.

The index generators (pulls.mjs/tickets.mjs) faithfully mirror the on-disk folders (getSourceBucket → dirname); they're not independently buggy. contentPath.mjs's floor(itemIndex/100)+1 math is correct. The fix is in the placement.

The Fix

Add a full-corpus active-tier re-chunk pass: scan all active pr-*.md/issue-*.md/discussion-*.md, sort ascending by id, recompute chunk-N via contentPath, fs.rename any mis-located file, prune empty dirs. This is the active-tier sibling of reconcileClosed*Locations (which only archives). Idempotent (no-op when already ordinal-correct).
Wire it into the sync/rebuild pipeline (rebuildContentIndexesAndSeo.mjs before the index generators) so it runs every sync → no re-drift.
Fix the metadata.pulls = {} vs {...metadata.issues} asymmetry.
Regenerate the current content + indexes to un-break the portal now.

Acceptance Criteria

Every active chunk folder holds exactly 100 items except the last (per content type).
The re-chunk pass is idempotent and runs in the sync/rebuild pipeline.
The portal Latest list shows the newest PRs/issues first, with contiguous position labels (1-100, 101-200, …).
Applies to pulls, issues, and discussions.

Out of Scope

Archive-tier chunks (sealed per ADR 0004 §2.6 — immutable, do not re-chunk).

ADR 0004 (learn/agentos/decisions/0004-github-content-architecture.md), contentPath.mjs, the github-workflow syncers, buildScripts/docs/index/*.mjs.

Release classification: prio-0 operator-flagged (public portal correctness); boardless (no portal board).

Origin Session ID: 0f5d9f1d-0683-452d-aac1-f467297186ac