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	>-
author	tobiu
state	Merged
createdAt	Apr 13, 2026, 1:43 PM
updatedAt	Apr 13, 2026, 1:54 PM
closedAt	Apr 13, 2026, 1:54 PM
mergedAt	Apr 13, 2026, 1:54 PM
branches	`dev` ← `agent/9921-n_ctx-exhaustion`
url	https://github.com/neomjs/neo/pull/9964

Merged

tobiu commented on Apr 13, 2026, 1:43 PM

Resolves #9921. Implements strict text length truncation on Session Service session history aggregation to prevent raw context payload sizes from crashing local MLX OpenAiCompatible endpoints via n_ctx exhaustion limits. Includes Unit Testing coverage as specified by the testing skill guidelines.

tobiu commented on Apr 13, 2026, 1:45 PM

Input from gemini-exp-1206 (Antigravity):

✦ # PR Review Summary

Status: Approved

Great job knocking this out! Implementing a hard constraint on the prompt envelope is the most robust, defensive way to avoid hitting local model context exhaustion limits while staying lightweight. I've reviewed the diff and it aligns cleanly with the Neo Agent OS architecture.

📊 Evaluation Metrics

[ARCH_ALIGNMENT]: 100 - Perfect alignment with the Neo Memory Core architecture. It intercepts the prompt aggregation exactly where it needs to without mutating the underlying vector abstractions.

[CONTENT_COMPLETENESS]: 90 - The inline comments clearly explain the purpose of the negative length slice as a preventative measure for n_ctx.

[EXECUTION_QUALITY]: 100 - Clean execution. The unit test explicitly intercepts the mock's prompt generation to mathematically prove the length falls within safe boundaries, ensuring no slow live-model API calls block the test suite.

[PRODUCTIVITY]: 100 - Successfully prevents MLX context crashes, which directly unblocks the RLAIF backend.

[IMPACT]: 85 - High impact for system stability. Resolves an immediate infrastructure blocker.

[COMPLEXITY]: 20 - Very targeted, focused implementation.

[EFFORT_PROFILE]: Quick Win - High ROI/Low Complexity.

🕸️ Context & Graph Linking

Target Epic / Issue ID: Resolves #9921

Related Graph Nodes: SessionService

🧠 Graph Ingestion Notes

[KB_GAP]: N/A

[TOOLING_GAP]: N/A

[RETROSPECTIVE]: The decision to truncate strings by prioritizing the tail end slice(-MAX_CONTENT_LENGTH) is a significant strategic takeaway for conversational RLAIF extraction. Because agents resolve their tasks at the very end of a turn sequence, the chronological tail holds the highest summarization value. Always preserve the trailing edges when forced to truncate sequences.

📋 Required Actions

To proceed with merging, please address the following:

All PR components approved and ready for Squash Merge by the Commander.

Excellent work resolving the n_ctx constraints! Once this is merged, we can proceed to #9913.