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.

Neo.mjs - Self-Evolving Software Organism

Frontmatter

title	feat(ai): add Chroma test-collection catalog diagnostics (#14010)
author	neo-gpt
state	Closed
createdAt	Jun 25, 2026, 9:07 PM
updatedAt	Jul 27, 2026, 12:07 AM
closedAt	Jun 25, 2026, 10:24 PM
mergedAt
branches	`dev` ← `codex/14010-chroma-test-collection-diagnostics`
url	https://github.com/neomjs/neo/pull/14019
contentTrust
projected
quarantined	0
signals	[]

Closed

neo-gpt commented on Jun 25, 2026, 9:07 PM

Resolves #14010

Adds read-only Chroma SQLite catalog diagnostics to the existing purgeTestCollections maintenance tool. The tool now distinguishes production default_database test-row bleed from isolated neo-unit-test residue, reports each test-* collection with tenant/database, collection id, metadata/vector/queue counts, vector segment path presence, and latest relevant timestamp, and exposes a --fail-on-production-test-collections guard for post-run checks.

Evidence: L3 (focused unit coverage plus live read-only unified-store catalog probe) -> L3 required (the incident is a live Chroma catalog recurrence where tenant/database placement determines whether this is production bleed or isolated test residue). Residual: post-merge operator validation can run the full CLI report during a quiet window.

Deltas from ticket

Live catalog evidence corrected the initial premise: the current 448 test-* collection rows are not in default_database. They are all in the isolated neo-unit-test database; only one is non-empty, and the latest relevant timestamp is 2026-06-06 01:14:46. The implementation therefore keeps the production guard and diagnostic, but does not treat the rows as active production namespace bleeding.

Test Evidence

npm run test-unit -- test/playwright/unit/ai/scripts/maintenance/purgeTestCollections.spec.mjs -> 14 passed.
node --input-type=module -e '...' live read-only diagnostic probe -> totalCollections=448, productionDatabaseCollections=0, isolatedTestDatabaseRows=448, nonEmptyCollections=1, latestRelevantTimestamp=2026-06-06 01:14:46.
git diff --check passed.
npm run agent-preflight -- ai/scripts/maintenance/purgeTestCollections.mjs test/playwright/unit/ai/scripts/maintenance/purgeTestCollections.spec.mjs passed.

Post-Merge Validation

Run npm run ai:purge-test-collections -- --fail-on-production-test-collections against the live store and confirm productionDatabase=0.
If cleanup is desired, run npm run ai:purge-test-collections -- --drop-test-db --apply in a separate operator window after the current defrag/backup work is quiescent.

Authored by Euclid (GPT-5, Codex Desktop). Session 9280140f-8b54-4462-9342-49cca7e226f4.

neo-gpt commented on Jun 25, 2026, 10:24 PM

Withdrawn. Operator correction accepted: this PR judged success by tenant/database separation and mostly added diagnostics. That is the wrong success criterion for #14010. Unit tests leaving neo-unit-test Chroma collections inside the shared live persist store are still live-store test bleed/bloat, even when they are not in default_database. This PR should not be reviewed or merged; the replacement needs to remove/prevent that residue instead of adding another smoke detector.