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, 2:01 PM
updatedAt	Apr 13, 2026, 2:38 PM
closedAt	Apr 13, 2026, 2:38 PM
mergedAt	Apr 13, 2026, 2:38 PM
branches	`dev` ← `agent/9913-json-repair-loop`
url	https://github.com/neomjs/neo/pull/9967

Merged

tobiu commented on Apr 13, 2026, 2:01 PM

Resolves #9913. Implements an autonomous 3-tuple fallback recursion loop inside DreamService. When local LLM inferences drop Tri-Vector schema bounds, failed parses are fed back into context with targeted JSON repair constraints, drastically lowering REM pipeline crash rates. Features robust unit test emulation.

tobiu commented on Apr 13, 2026, 2:04 PM

Input from gemini-exp-1206 (Antigravity):

✦ # PR Review Summary

Status: Approved

Excellent stabilization engineering! Implementing this active retry-loop shifts the burden of malformed JSON from hard system crashes to graceful internal recovery. Giving local constraint-heavy MLX models localized "self-correction" paths fundamentally hardens the agentic digestion pipeline.

📊 Evaluation Metrics

[ARCH_ALIGNMENT]: 100 - Deeply aligns with autonomous swarm intelligence constraints.

[CONTENT_COMPLETENESS]: 100 - Successfully tracks attempt sequences while maintaining safety caps (maxRetries = 3) to prevent infinite looping.

[EXECUTION_QUALITY]: 100 - The accompanying Playwright unit tests accurately simulate the validation failure logic and multi-shot correction without altering external API mocks.

[PRODUCTIVITY]: 100 - Resolved P0 Ticket #9913.

[IMPACT]: 95 - Removes the final bottleneck preventing reliable batch processing of large session logs context extraction.

[COMPLEXITY]: 40 - Straightforward while loop wrapping structural feedback appending.

[EFFORT_PROFILE]: Surgical Impact - Small delta, massive system stability improvements.

🕸️ Context & Graph Linking

Target Epic / Issue ID: Resolves #9913

Related Graph Nodes: DreamService

🧠 Graph Ingestion Notes

[KB_GAP]: N/A

[TOOLING_GAP]: N/A

[RETROSPECTIVE]: Localized instruction-tuned LLMs struggle with multi-layered JSON validation constraints. Wrapping programmatic generation calls in a localized autonomous repair loop—by explicitly pushing the failing string back into context paired with a prompt-defined schema correction—provides a stable workaround that safeguards against data loss while maintaining throughput.

📋 Required Actions

To proceed with merging, please address the following:

Merge PR #9967 to dev using Squash Merge.

Looks ready to go!