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	9375
title	Improve ScrollSync Lock Release Mechanism (rAF vs setTimeout)
state	Closed
labels	enhancementaiarchitecture
assignees	tobiu
createdAt	Mar 7, 2026, 5:54 PM
updatedAt	Mar 7, 2026, 5:57 PM
githubUrl	https://github.com/neomjs/neo/issues/9375
author	tobiu
commentsCount	1
parentIssue	9369
subIssues	[]
subIssuesCompleted	0
subIssuesTotal	0
contentTrust
projected
quarantined	0
signals	[]
blockedBy	[]
blocking	[]
closedAt	Mar 7, 2026, 5:57 PM

Improve ScrollSync Lock Release Mechanism (rAF vs setTimeout)

Closed v12.1.0/archive-v12-1-0-chunk-1 enhancementaiarchitecture

tobiu commented on Mar 7, 2026, 5:54 PM

Part of Epic #9369

Currently, the granular lock (coordinatingNodes Set) in ScrollSync.mjs is released on the next requestAnimationFrame. However, GridDragScroll utilizes a continuous requestAnimationFrame loop for momentum scrolling. Because native DOM scroll events are asynchronous to the render pipeline, they might dispatch to the event queue after the lock is released on the next frame, defeating the gatekeeper and causing a ping-pong loop.

Investigate and implement a safer lock release mechanism (e.g., using setTimeout to push the release to the end of the macrotask queue) to ensure the lock remains held across continuous 16ms programmatic mutations until all resulting native scroll events have been flushed by the browser.

tobiu Mar 7, 2026, 5:56 PM

Input from Gemini 3.1 Pro:

✦ I have updated the lock release mechanism in ScrollSync.mjs to use a debounced setTimeout (default 50ms) instead of requestAnimationFrame.

Why: requestAnimationFrame releases the lock on the very next render tick. However, native DOM scroll events are dispatched asynchronously. If GridDragScroll's momentum loop fires an update every 16ms, a native scroll event might arrive after the rAF tick has already released the lock, causing a false positive in the interceptor and triggering a feedback loop.

The Fix: By using a Map of node IDs to timeout IDs, the lockNode method now clears any existing timeout and sets a new 50ms timer. This guarantees that during a continuous momentum throw (which updates the DOM every 16ms), the lock remains firmly held the entire time. It only releases 50ms after the final momentum update has settled, providing plenty of time for all trailing native asynchronous scroll events to be safely ignored by the gatekeeper.