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	12204
title	Epic: Portal news — PR + Discussion views & chunked lazy-load tree
state	Closed
labels	epicaiarchitectureperformancebuild
assignees	[]
createdAt	May 30, 2026, 7:24 PM
updatedAt	Jun 7, 2026, 7:17 PM
githubUrl	https://github.com/neomjs/neo/issues/12204
author	neo-opus-ada
commentsCount	2
parentIssue	null
subIssues	12205 Neutralize the portal offscreen renderer into a shared TimelineCanvas 12206 Promote portal canvas view-glue (bridge + CanvasWrapper) to shared 12207 Rename timeline-node model/store TicketTimelineSection(s) → TimelineSection(s) 12208 Extract the shared timeline contract into Neo.app.content.Component 12209 Make portal content cross-link URLs per-content-type config 12210 Portal index generators for Pull Requests + Discussions (chunked output) 12211 Portal Discussion view — Portal.view.news.discussions (thin subclass) 12213 Portal Pull Request view — Portal.view.news.pulls (two-section parser) 12214 Document the canonical PR + Discussion markdown grammar by the syncers 12215 DiscussionSyncer: accepted-answer support (GraphQL query + emit marker) 12216 DiscussionSyncer: structured threaded replies (depth-aware emit) 12217 Refactor tickets index emitter → chunked-by-folder + path de-duplication 12218 Opt-in load-on-folder-expand for the shared portal TreeList 12219 Make tickets controller deep-link + default-route lazy-aware 12220 Correct tickets.mjs module JSDoc to the chunked-index reality 12265 Fix portal timeline import paths after shared move 12304 Switch portal tickets store to chunked lazy index 12309 Portal chunked trees: chunk folders display out of sequence (sortDate sort vs positional labels) 12317 Use bot avatar glyphs in Portal news summary views
subIssuesCompleted	19
subIssuesTotal	19
blockedBy	[]
blocking	[]
closedAt	Jun 2, 2026, 12:17 AM

Epic: Portal news — PR + Discussion views & chunked lazy-load tree

Closed v13.0.0/archive-v13-0-0-chunk-14 epicaiarchitectureperformancebuild

neo-opus-ada commented on May 30, 2026, 7:24 PM

Context

The portal "news" section surfaces four content types (Medium, Blog, Tickets, Release Notes) but not Pull Requests or Discussions — both are synced to resources/content/{pulls,discussions}/ (+ archives) yet invisible to human visitors and SEO crawlers (originating ticket #10056). Separately, the tickets tree-index JSON (apps/portal/resources/data/tickets.json) has grown to ~1.5 MB / ~8,772 leaf rows, eager-loaded up front (autoLoad:true) — a portal-news performance problem.

A 6-agent analysis workflow produced a unified design tying both tracks together; this epic is its execution home. The two tracks share one emitted JSON shape, so they reinforce each other.

The Problem

Track A (views): add PR + Discussion views, reusing the existing ticket timeline + offscreen "Neural Timeline" canvas. The view stack is ~80% already generalized; the only irreducible specialization is the per-type markdown→timeline parser.
Track B (perf): the per-issue markdown is already lazy (fetched on leaf-select) and already split into chunk-N/ folders (ADR 0004). What is not chunked is the tree-index JSON — one monolith loaded whole. Map it to per-chunk-folder index files, lazy-loaded on tree-node expand.

The Architectural Reality

View shell is framework-level + content-agnostic: src/canvas/Base.mjs, src/app/SharedCanvas.mjs (already multi-renderer via rendererClassName/rendererImportPath), src/app/content/{Container,PageContainer,Component}, and the sections store bridge (a load-bearing producer/consumer contract — must stay stable). The offscreen renderer apps/portal/canvas/TicketCanvas.mjs is purely geometric (only its name says "Ticket"). The Releases tab already proves the shell generalizes to a non-issue type.
Per-type parser differences are real (verified against the synced corpus): Issue `