All phases
19
Datasets & Training Orchestration
The 'Intelligence Foundry' gaps. Four sub-phases: 19A Garden UX (creator's home view), 19B Dataset registry (first-class datasets, public catalog, API/IoT connectors), 19C Training orchestration (real model invocation + training queue, replaces the trinary hash stub), 19D Feedback loops (outcome tracking, retraining triggers, A/B). Composes with the existing tissue primitive. Total ~25 SP across the four sub-phases.
Progress
100%
Tasks done
22 / 22
0 remaining
Owner
Eng lead + AI agents
Status: done
Tasks
Kanban for this phase. Move tasks between columns to update status.
Backlog
0
To do
0
In progress
0
In review
0
Done
22
19A.1 — 'Intelligence Garden' home view (apps/web/dashboard/garden)
p0
Eng lead + AI agents1.5 SP
19A.2 — Garden widgets (ANN card, tissue card, training-job card, revenue card)
p0
Eng lead + AI agents1 SP
19A.3 — Add /garden to dashboard nav, redirect /dashboard → /dashboard/garden
p1
Eng lead + AI agents0.5 SP
19A.4 — Garden empty state (first-run experience)
p1
Eng lead + AI agents0.5 SP
19B.1 — services/dataset service skeleton (Fastify + Drizzle + JWT)
p1
Eng lead + AI agents1 SP
19B.2 — Dataset schema (datasets + dataset_versions + dataset_access)
p1
Eng lead + AI agents1.5 SP
19B.3 — Dataset upload pipeline (S3-compatible, content-hash, schema sniff)
p1
Eng lead + AI agents2 SP
19B.4 — Public dataset catalog + browse UI
p2
Eng lead + AI agents1.5 SP
19B.5 — Connector registry (API + IoT + public sources)
p2
Eng lead + AI agents3 SP
19B.6 — SDK: Datasets resource (list, retrieve, create, uploadVersion, listVersions)
p1
Eng lead + AI agents0.5 SP
19C.1 — services/training service skeleton (job queue, worker pool)
p0
Eng lead + AI agents2 SP
19C.2 — Training recipe schema + catalog (architecture, hyperparameters, optimizer)
p0
Eng lead + AI agents1 SP
19C.3 — Real LLM invocation in ANN service (close the trinary stub gap)
p0
Eng lead + AI agents3 SP
19C.4 — Bridge services/training ↔ services/compute (allocate, monitor, release)
p0
Eng lead + AI agents2 SP
19C.5 — Training progress stream (WebSocket / SSE) for the 'Growing Intelligence' UX
p1
Eng lead + AI agents1.5 SP
19C.6 — On training success: auto-publish new ANN version + run validation
p0
Eng lead + AI agents1.5 SP
19D.1 — Decision outcome tracking (real-world feedback)
p1
Eng lead + AI agents1.5 SP
19D.2 — Auto-retrain trigger (drift + outcome rate)
p1
Eng lead + AI agents2 SP
19D.3 — A/B / shadow deployment for ANN versions
p2
Eng lead + AI agents2 SP
19D.4 — Garden UX: show 'actual accuracy' alongside predicted accuracy
p2
Eng lead + AI agents0.5 SP
ADR 003 — Dataset ownership + licensing model
p0
Eng lead + AI agents0.5 SP
Phase 3 Vision copy rewrite (brand-voice pass)
p1
Eng lead + AI agents0.5 SP