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	14113
title	Dimension-audit fact-gatherer for the data-integrity diagnostics runner (#14109 input)
state	Closed
labels	enhancementaitestingarchitecture
assignees	neo-opus-ada
createdAt	Jun 26, 2026, 4:59 PM
updatedAt	Jun 26, 2026, 5:31 PM
githubUrl	https://github.com/neomjs/neo/issues/14113
author	neo-opus-ada
commentsCount	0
parentIssue	null
subIssues	[]
subIssuesCompleted	0
subIssuesTotal	0
contentTrust
projected
quarantined	0
signals	[]
blockedBy	[]
blocking	[]
closedAt	Jun 26, 2026, 5:31 PM

Dimension-audit fact-gatherer for the data-integrity diagnostics runner (#14109 input)

Closed v13.1.0/archive-v13-1-0-chunk-7 enhancementaitestingarchitecture

neo-opus-ada commented on Jun 26, 2026, 4:59 PM

Context

The #14109 data-integrity diagnostics runner gathers per-collection facts and feeds them to the pure detect-producers. The embedding-dimension producer (buildDimensionConsistencyDiagnosis, #14102/#14104) consumes per-collection {collection, expectedDimension, mismatchedVectorCount} samples — but no helper computes mismatchedVectorCount (auditChromaVectorCoverage checks index coverage, not per-vector dimension). This is the dimension fact-gatherer Vega (#14109 steward) confirmed as a separate slice owned by me (the producer author); the runner imports it.

The Fix

An exported auditCollectionVectorDimensions({collection, collectionName, expectedDimension, sampleSize}) in ai/scripts/maintenance/checkChromaIntegrity.mjs (sibling of auditChromaVectorCoverage / probeStoredEmbeddingExportability): sample up to sampleSize ids, read their stored embeddings (mirroring probeStoredEmbeddingExportability's collection.get({limit}) → collection.get({ids, include:['embeddings']})), and count present-but-wrong-dimension vectors → returns the producer's samples element shape, 1:1, no reshape.

A missing/absent embedding is NOT a dimension mismatch (that's coverage's domain, #14075) — only a present vector whose .length !== expectedDimension counts.

Contract Ledger

Target Surface	Source of Authority	Proposed Behavior	Fallback	Docs	Evidence
`auditCollectionVectorDimensions({collection, collectionName, expectedDimension, sampleSize})`	#14109 gather-step contract (Vega-confirmed 1:1) + the `buildDimensionConsistencyDiagnosis` input	Sample embeddings → `{collection: collectionName, expectedDimension, mismatchedVectorCount, sampledCount}`; counts present vectors whose `.length !== expectedDimension`	empty/unsampled collection → `mismatchedVectorCount: 0`; a `.get` failure surfaces as a probe error (never throws into the runner)	fn JSDoc + ticket	unit spec (mock collection)

Acceptance Criteria

Samples per-collection embeddings + returns {collection, expectedDimension, mismatchedVectorCount} (the producer's samples element, 1:1).
Counts only present-but-wrong-dimension vectors; missing embeddings are not counted (coverage's domain).
Empty/unsampled collection → mismatchedVectorCount: 0 (no false signal).
Exported for the #14109 runner to import. Unit-covered with a mock collection (mismatch, all-match, empty).

Out of Scope

The #14109 runner wiring (schedule/history/route — consumes this).
The pure buildDimensionConsistencyDiagnosis producer (#14102/#14104).
Other gatherers (coverage exists; monotonicity is runner-history-sourced, not a gatherer — per the #14109 de-dup).

#14109 (the runner that imports this), #14102 / #14104 (the producer this feeds), #14026 (the detect-signal class), #14039 (v13.1 epic).

Authored by Ada (Claude Opus 4.8, Claude Code).