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	10153
title	Lazy back-fill: Memory + Session nodes for pre-migration Chroma rows
state	Closed
labels	enhancementaiarchitecturecore
assignees	tobiu
createdAt	Apr 21, 2026, 11:31 AM
updatedAt	Apr 22, 2026, 2:00 PM
githubUrl	https://github.com/neomjs/neo/issues/10153
author	tobiu
commentsCount	1
parentIssue	10143
subIssues	[]
subIssuesCompleted	0
subIssuesTotal	0
blockedBy	[x] 10151 DreamService: deterministic ingestion phase for Memory + Session graph nodes
blocking	[]
closedAt	Apr 22, 2026, 1:21 AM

Lazy back-fill: Memory + Session nodes for pre-migration Chroma rows

Closed v13.0.0/archive-v13-0-0-chunk-5 enhancementaiarchitecturecore

tobiu commented on Apr 21, 2026, 11:31 AM

Lazy back-fill: Memory + Session nodes for pre-migration Chroma rows

Context

Parent sub-epic (#10143) phase 3. Phase 1 adds forward ingestion; phase 3 handles historical Chroma rows (8246 memories / 794 summaries at filing).

The Problem

Forward-only ingestion leaves permanent asymmetry: mailbox IN_REPLY_TO or identity AUTHORED_BY references to pre-migration artifacts have dangling targets. Eager back-fill at deploy time is bounded but expensive; lazy amortizes cost.

The Architectural Reality

Historical memories in neo-agent-memory; summaries in neo-agent-sessions. Total ~9040 rows.
Graph schema (post-phase-1) has Memory/Session node types; Chroma-ID → node mapping is deterministic.
graphDigested: true on summaries is the existing per-artifact processing flag — pattern to match.

The Fix

New marker property on Memory/Session nodes: backfilled: true (vs. liveIngested: true)
Lazy trigger: graph addEdge with missing target runs single-row ingestion synchronously, then completes the edge
Optional batch-priming CLI (ai/scripts/priorityBackfill.mjs) — eager back-fill for high-traffic sessions (e.g. last 30 days). Opt-in.

Acceptance Criteria

addEdge with missing target triggers lazy ingestion; edge completes atomically
backfilled: true marker distinguishable from live-ingested nodes in queries
Graph integrity check: no dangling edges post-shipping, even across historical boundary
CLI priority-backfill script documented
Performance observability: lazy ingestion latency instrumented; SLO defined post-phase-1 baseline, not pre-hardcoded

Out of Scope

Eager full-corpus back-fill at deploy (cost-prohibitive)
Re-running extractor on historical sessions for provenance edges (opportunistic; not required)
Deletion/archival policy for old nodes (separate concern)

Avoided Traps

Eager back-fill. Rejected. 9040 rows × ingestion + potential re-extraction = substantial deployment stall.
Forward-only (no back-fill). Rejected. Permanent dangling edges break invariants; downstream consumers each have to special-case null targets.

Parent: #10143
Blocked by: phase 1 (need mapping logic)
Files: graph edge-creation service, ai/scripts/priorityBackfill.mjs (new)

Origin Session ID: 71dc3cd8-d39d-48e1-ac62-e240ca67d1a5

tobiu added the enhancement label on Apr 21, 2026, 11:31 AM

tobiu added the ai label on Apr 21, 2026, 11:31 AM

tobiu added the architecture label on Apr 21, 2026, 11:31 AM

tobiu added the core label on Apr 21, 2026, 11:31 AM

tobiu added parent issue #10143 on Apr 21, 2026, 11:33 AM

tobiu marked this issue as being blocked by #10151 on Apr 21, 2026, 11:33 AM

tobiu cross-referenced by #10143 on Apr 21, 2026, 3:06 PM

tobiu cross-referenced by #10158 on Apr 21, 2026, 3:23 PM

tobiu cross-referenced by PR #10161 on Apr 21, 2026, 4:36 PM

tobiu cross-referenced by #9999 on Apr 21, 2026, 7:03 PM

tobiu cross-referenced by #10145 on Apr 21, 2026, 8:43 PM

tobiu cross-referenced by PR #10165 on Apr 21, 2026, 8:59 PM

tobiu assigned to @tobiu on Apr 21, 2026, 10:28 PM

tobiu Apr 21, 2026, 11:02 PM

Input from Antigravity (Gemini 3.1 Pro):

✦ # Epic Review by Antigravity (Gemini 3.1 Pro)

Stage 1 — Roadmap Fit: ✅ Vital for the transition to Graph-first Memory artifacts without breaking historical provenance or causing a deploy-time stall.

Stage 2 — Approach Elegance: ✅ The lazy-evaluation model for backfilling on addEdge is highly elegant. It amortizes the cost of 9040 rows beautifully, while priorityBackfill.mjs allows optional batch-warming.

Stage 3 — Sub-Structure Coherence: ✅ This ticket is highly focused; its ACs define a clear, self-contained atomic delivery.

Stage 4 — Prescription Layer: ✅ Fixing the dangling edge problem exactly at the addEdge boundary is the right substrate-level interception point.

Stage 5 — Avoided Traps Completeness: ✅ Properly identifies the extremes of Eager full-corpus and Forward-only as catastrophic/expensive, hitting the lazy sweet spot.

Verdict: Greenlight

Origin Session ID: ada4aecb-60fc-4b25-a76b-cb92600835a2

tobiu cross-referenced by PR #10171 on Apr 22, 2026, 12:47 AM

tobiu cross-referenced by #10172 on Apr 22, 2026, 1:19 AM

tobiu referenced in commit 8f52831 - "fix(ai): recover orphaned .draining file from SIGKILL edge case (#10153) on Apr 22, 2026, 1:19 AM

tobiu referenced in commit 3dbea00 - "feat(ai): lazy back-fill for Memory + Session graph nodes (#10153) (#10171) on Apr 22, 2026, 1:21 AM

tobiu closed this issue on Apr 22, 2026, 1:21 AM

Lazy back-fill: Memory + Session nodes for pre-migration Chroma rows

Lazy back-fill: Memory + Session nodes for pre-migration Chroma rows

Context

The Problem

The Architectural Reality

The Fix

Acceptance Criteria

Out of Scope

Avoided Traps

Related