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	13355
title	Render first-widget transcript and blueprint evidence pane
state	Closed
labels	enhancementaiarchitecture
assignees	neo-opus-ada
createdAt	Jun 15, 2026, 5:14 PM
updatedAt	Jun 17, 2026, 8:35 AM
githubUrl	https://github.com/neomjs/neo/issues/13355
author	neo-gpt
commentsCount	5
parentIssue	13349
subIssues	[]
subIssuesCompleted	0
subIssuesTotal	0
blockedBy	[x] 13439 Fix neuralLink connectToApp e2e fixture for childapp SharedWorker topology (target.toLowerCase TypeError), [x] 13362 Create first grid through Neural Link in a live app
blocking	[x] 13357 Capture first-widget requests in a safe chat surface
closedAt	Jun 16, 2026, 6:20 PM
milestone	v13.1

Render first-widget transcript and blueprint evidence pane

Closed v13.1.0/archive-v13-1-0-chunk-3 enhancementaiarchitecture

neo-gpt commented on Jun 15, 2026, 5:14 PM

Context

Epic #13349 decomposes the H2 Agent Harness path: chat/work surface -> constrained blueprint -> live first Neo widget. The first implementation leaf #13353 / PR #13354 proves the constrained blueprint can render as a live Neo grid pane. The next H2 proof needs the surrounding conversation/evidence surface so the user can see what request produced the widget and what bounded blueprint was accepted, without treating generated content as trusted HTML or arbitrary code.

Release classification: post-release Agent Harness product line; belongs on the Agent Harness project, not the v13 release board.

The Problem

A live grid pane alone can be mistaken for a hardcoded demo. H2 needs visible provenance: the request, the accepted blueprint summary, and the rendered result need to sit together so a reviewer can verify this is the start of a conversational creation loop rather than an isolated widget.

The surface must stay deliberately constrained. Transcript/evidence text may be agent-authored or user-authored, so it must render through Neo components / safe text surfaces, not innerHTML, and it must not introduce persistence or write-through before those boundaries are explicitly owned by later leaves.

The Architectural Reality

#13349 owns the H2 conversational app creation epic and names transcript/evidence presentation as a missing child capability.
ADR 0020 names the canonical H2 demo: chat pane -> "build me a neo grid" -> a separate live pane/window containing a mutable grid.
#13018 constrains transcript rendering toward safe VDOM / no unsafe HTML rendering.
#13022 seeded AgentOS visual tokens for chat, transcript, state, and grid-pane surfaces.
#13353 / PR #13354 is the prerequisite first live-widget pane; this ticket should build on that shape after it lands, not race the same files while the PR is under review.

The Fix

Add a one-PR H2 child slice that renders a transcript / blueprint evidence pane adjacent to or above the first-widget result inside the AgentOS widget child app.

Expected shape:

Show a deterministic user request such as "build me a neo grid" and a deterministic agent response summary.
Show the accepted blueprint schema / title / row-count / column-count as inspectable evidence, not raw executable code.
Render all transcript/evidence content through bounded Neo component text surfaces or safe VDOM-equivalent rendering; no innerHTML or arbitrary code execution.
Keep persistence/write-through out of scope. This is visible provenance for the first widget, not durable chat history.
Consume the first-widget live pane from #13353 rather than replacing it.

Contract Ledger Matrix

Target Surface	Source of Authority	Proposed Behavior	Fallback	Docs	Evidence
AgentOS widget child app transcript/evidence pane	#13349, ADR 0020, #13018	Displays deterministic request/response provenance and accepted blueprint metadata around the first live widget	If blueprint evidence is invalid, show bounded rejected-state text	JSDoc / @summary on new helpers or component methods	Unit test plus local render smoke
Blueprint evidence projection	#13353 blueprint contract	Projects schema/title/row-count/column-count without exposing executable strings	Reject unknown evidence object keys	JSDoc / @summary	Unit test for accepted and rejected evidence
Rendering safety	#13018 safe transcript constraint	No unsafe HTML path for transcript/evidence text	Fail closed to safe text	Test source assertion	Unit test / source assertion

Decision Record impact

Aligned with ADR 0020. No ADR amendment expected; this ticket implements H2 evidence/provenance presentation without changing the product ordering, source placement, or co-habitation model.

Acceptance Criteria

The AgentOS widget child app renders a transcript/evidence area that shows a deterministic user request and agent response summary for the first widget path.
The surface shows accepted blueprint metadata (at minimum schema, title, column count, and row count) without exposing executable code or using unsafe HTML rendering.
Invalid evidence data fails closed to a bounded visible state.
The existing first-widget live grid pane remains present and uses the #13353 blueprint output rather than being forked into a separate demo path.
Focused unit coverage verifies valid evidence, invalid evidence, and safe-rendering/source constraints.
Local render smoke verifies the transcript/evidence pane and grid render together.

Out of Scope

Durable transcript persistence / Memory Core write-through.
Natural-language model invocation or streaming chat orchestration.
User-editable mutation path for the live widget.
Electron shell/build-root work; that remains #13033.
Any partner-private workflow or client-specific UI.

Avoided Traps / Gold Standards Rejected

Static demo dead end: rejected. This leaf explicitly ties the widget to visible request/evidence context.
Unsafe transcript rendering: rejected. Agent/user text must not become trusted HTML.
Persistence bundled too early: rejected. This leaf proves visible provenance only; persistence/write-through gets a later leaf with its own boundary.
Duplicate Electron support: rejected. #13033 remains the canonical Electron shell/build-root lane and is not touched here.

Parent: #13349. Grandparent: #13012. Prerequisite leaf: #13353 / PR #13354. Sibling lane: #13033 remains the Electron shell/build-root lane. ADR: learn/agentos/decisions/0020-agent-harness-concept.md.

Duplicate / Claim Sweep

Live latest-open sweep: checked the latest 20 open issues immediately before creation on 2026-06-15T15:15Z. Relevant results were #13349 (parent H2 epic), #13353 (first live-widget prerequisite), #13012 (grandparent harness epic), #13033 (Electron lane, not duplicate), and unrelated assigned backlog items; no equivalent transcript/evidence child leaf was present.

Targeted live GitHub searches: "first widget transcript evidence pane" and "H2 transcript blueprint evidence Agent Harness" found #13349 and #13353 but no duplicate child leaf.

A2A in-flight sweep: native list_messages / add_message tools are not exposed in this Codex Desktop surface. I could not inspect private in-flight A2A claims from this harness. Public GitHub issue/PR state and repo content were checked immediately before filing, and no public duplicate or collision was visible.

Semantic KB sweep: ask_knowledge_base(query='Agent Harness H2 first widget blueprint chat transcript persistence live edit pane', type='all') surfaced the H2 blueprint/transcript/persistence concepts and confirmed this ticket is the transcript/evidence slice, distinct from persistence/write-through and Neural Link live-edit leaves.

Exact repo/content sweep: rg "first-widget|first widget|transcript evidence|blueprint evidence|H2 transcript" resources/content/issues resources/content/discussions .agents/workflows learn -n returned no existing transcript/evidence child leaf.

Origin Session ID: 019ec8a7-1f8e-75a3-b223-fe59cc444776

Retrieval Hint: "Agent Harness H2 first-widget transcript evidence blueprint metadata safe rendering no innerHTML"

tobiu removed the block by #13353 on Jun 15, 2026, 6:00 PM

tobiu referenced in commit c8255b1 - "feat(agentos): first-widget evidence pane + live grid render-together (#13355) (#13409) on Jun 16, 2026, 10:32 AM