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	feat(ai): replay message WAL into mailbox graph (#13892)
author	neo-gpt
state	Merged
createdAt	Jun 23, 2026, 1:08 PM
updatedAt	Jun 23, 2026, 5:33 PM
closedAt	Jun 23, 2026, 5:33 PM
mergedAt	Jun 23, 2026, 5:33 PM
branches	`dev` ← `codex/13892-message-wal-projection`
url	https://github.com/neomjs/neo/pull/13921
contentTrust
projected
quarantined	0
signals	[]

Merged

neo-gpt commented on Jun 23, 2026, 1:08 PM

Resolves #13892 Related: #13889

Projects accepted A2A message WAL records into the mailbox graph idempotently. The message WAL now has graph-projection marker files, pending-record reads skip already-projected rows, MailboxService.addMessage() shares the replay projection path, and both local daemon and in-process Memory Core drain hosts now wire the mailbox projection processor.

Evidence: L2 (focused unit coverage over JSONL WAL markers, SQLite-backed mailbox graph projection, daemon processor wiring, and config/server import coverage) -> L2 required (idempotent replay, pending retention on required-edge failure, send-time broadcast fan-out, and host-loop wiring). Residual: A2A message vector/search indexing remains outside #13892 and stays deferred to the parent epic.

Deltas from ticket

Reused the memory-WAL marker pattern for message graph projection instead of inventing a second reconciliation model.
Made optional semantic edge failures non-blocking during projection; delivery-critical SENT_BY, SENT_TO, and broadcast DELIVERED_TO still fail loudly and leave the WAL row pending.
Replay uses the WAL's send-time broadcast recipient snapshot, not the current AgentIdentity graph audience.

Test Evidence

node --check ai/services/memory-core/helpers/messageWalStore.mjs
node --check ai/services/memory-core/MailboxService.mjs
node --check ai/daemons/message/drainCycle.mjs
node --check ai/daemons/message/daemon.mjs
node --check ai/mcp/server/memory-core/Server.mjs
node --check test/playwright/unit/ai/services/memory-core/MailboxService.spec.mjs
node --check test/playwright/unit/ai/daemons/message/drainCycle.spec.mjs
git diff --check
npm run agent-preflight -- ai/services/memory-core/helpers/messageWalStore.mjs ai/services/memory-core/MailboxService.mjs ai/daemons/message/drainCycle.mjs ai/daemons/message/daemon.mjs ai/mcp/server/memory-core/Server.mjs test/playwright/unit/ai/services/memory-core/MailboxService.spec.mjs test/playwright/unit/ai/daemons/message/drainCycle.spec.mjs
npm run test-unit -- test/playwright/unit/ai/daemons/message/drainCycle.spec.mjs test/playwright/unit/ai/services/memory-core/MailboxService.spec.mjs test/playwright/unit/ai/mcp/server/memory-core/Server.spec.mjs test/playwright/unit/ai/mcp/server/memory-core/config.template.spec.mjs test/playwright/unit/ai/daemons/message/drainLock.spec.mjs (107 passed)

Post-Merge Validation

Confirm a deployed message drain host projects a seeded pending message WAL row, writes the graph marker, and leaves no pending row for that message.

Authored by Euclid (GPT-5, Codex Desktop). Session 019ef378-527d-7393-bc74-ec3a1d3f2ddf.

neo-gpt commented on Jun 23, 2026, 1:17 PM

[review-request] #13921 is ready for primary review.

Review role: primary-reviewer Requested action: use /pr-review on PR #13921 Reviewer: @neo-opus-grace Current head: 05a4027d99297c34167f5e844779f1c7640f565f CI: all checks green as of 2026-06-23T11:16:40Z

Scope: #13892 message WAL graph projection replay/idempotency under #13889. The PR wires marker-based pending reads, mailbox projection replay, local/in-process daemon processors, and focused replay tests.

A2A note: the current Codex harness exposes no add_message tool; GitHub review request + this public handoff comment are the lifecycle route.

tobiu APPROVED reviewed on Jun 23, 2026, 5:33 PM

No review body provided.