Slice Epic Story Mapping

:::info Source Sourced from docs/roadmap/slice-epic-story-mapping.md in the documentation repo. :::

Execution-layer artifact. Companion to ROADMAP.md and 07 Epics & Stories.

This document maps every user story (US-1 … US-118) to the slice and milestone that delivers it. Each mapping includes cross-service dependencies, AI-first requirements, offline-first requirements, security + multi-tenant requirements, and testing + observability gates. Sprint-ready backlog tables follow each slice section.

0. Conventions

• Slice IDs map to milestones: S0→M0, S1→M1, S2+S3→M2, S4→M3, S5→M4, S6→M5.
• Story status per slice is one of: impl (full implementation), stub (minimal, will mature later), back-compat (maintain unchanged), deferred (not in this slice).
• Every story already has its full AC + DoD in doc 07; this doc adds slice-scoped scope and dependency detail.

1. Story-to-Slice Map (authoritative)

Story	Title	Slice	Status in slice	Later matures in
US-1	Provision org + invite users	S0	impl	—
US-2	Sign in via SAML/OIDC SSO	S4	impl	— (enhancements in S6)
US-3	Bind device for offline	S0	impl	—
US-4	Reset my password	S0	impl	—
US-5	Enroll WebAuthn	S4	impl	—
US-6	Define org units	S0	impl	—
US-7	Define custom role	S4	impl	—
US-8	Dynamic Group	S4	impl	—
US-9	Switch active tenant	S0	impl	—
US-10	SSO admin configuration	S4	impl	—
US-11	Bulk-import users via CSV	S4	impl	—
US-12	Create draft with modules/lessons	S2	impl	—
US-13	Add text/image/video/audio blocks	S2	impl	S5 (offline edits)
US-14	Add quiz + branching block	S2 (quiz) / S4 (branching)	impl/impl	—
US-15	Live multi-author collaboration	S5	impl	—
US-16	Submit for review + approve	S2	impl	—
US-17	Publish a course	S2	impl	—
US-18	Localize a course manually	S5	impl	—
US-19	Export to SCORM 1.2	S3	impl	—
US-20	Export to SCORM 2004 + xAPI	S4	impl	—
US-21	Export to HTML5 standalone	S2	impl	—
US-22	Build offline bundle + license envelope	S1	impl	—
US-22a	Tamper detection on bundle mount	S1	impl	—
US-22b	Import 3rd-party SCORM 1.2	S2 (v1) / S4 (full)	stub/impl	—
US-23	Browse + filter catalog	S2	impl	S6 (semantic re-rank)
US-24	Hide unpublished from non-authors	S0 (tests) / S2 (UI)	impl	—
US-25	Switch locale	S1	impl	—
US-26	Course detail page with instructor + reviews	S2	impl	—
US-27	Submit listing for approval	S3	impl	—
US-28	Define pricing plans	S3	impl	S6 (multi-currency depth)
US-29	Buy course as individual	S3	impl	—
US-30	Buy seat pack + assign	S3	impl	—
US-31	Refund within policy	S3	impl	—
US-32	Provider sees payouts	S3	impl	S6 (multi-currency)
US-33	Coupon redemption	S3	impl	—
US-34	Listing moderation AI + human	S4	impl	S6 (pattern v2)
US-35	One-shot assignment	S4	impl	—
US-36	Recurring assignment via RRULE	S4	impl	—
US-37	Team compliance dashboard	S4	impl	S6 (AI insight v2)
US-38	Overdue reminders + escalation	S4	impl	—
US-39	Reassign on failure	S4	impl	—
US-40	Resume from last position	S1	impl	—
US-41	Quiz with immediate feedback	S1	impl	—
US-42	Navigate branching scenario	S4	impl	—
US-43	Track progress in real time	S1	impl	—
US-44	View transcript	S1	impl	—
US-45	Earn certificate on completion	S1	stub	S3 templates
US-46	Verify certificate via public link	S1	impl	—
US-47	Define certificate template	S4	impl	—
US-48	Recertification reminders	S4	impl	—
US-49	Configure notification preferences	S1	impl	—
US-50	Multi-channel notifications	S1	impl	—
US-51	Manage suppression after bounce	S1	impl	—
US-52	Tenant-configurable webhooks	S3	impl	—
US-53	Lexical + semantic search	S5	impl	—
US-54	Personalized recommendations	S6	impl	—
US-55	Next-step recommendation	S6	impl	—
US-56	Org admin completion dashboard	S1 (mini) / S4 (full)	stub/impl	—
US-57	Provider sales dashboard	S3 (v1) / S6 (full)	stub/impl	—
US-58	Scheduled email report	S6	impl	—
US-59	Export raw events	S6	impl	—
US-60	AI generate lesson from prompt	S2	impl	—
US-61	Convert PDF/policy to draft course	S4	impl	—
US-62	AI quiz from lesson	S2	impl	—
US-63	AI branching scenario	S4	impl	—
US-64	Auto-translate course	S4 (pilot) / S5 (production)	stub/impl	—
US-65	AI TTS narration	S4	impl	—
US-66	AI image/diagram generation	S5	impl	—
US-67	AI rewrite/simplify	S2	impl	—
US-68	AI SCORM metadata + objectives	S4	impl	—
US-69	In-course AI tutor	S1	impl	—
US-70	AI lesson summary	S1	impl	—
US-71	AI worked example	S1	impl	—
US-72	AI adaptive practice	S4	impl	—
US-73	AI feedback on open-ended	S4	impl	—
US-74	AI recommends trainings for role	S4	impl	S6 depth
US-75	Predict at-risk non-completion	S6	impl	—
US-76	Auto-generate board compliance summary	S6	impl	—
US-77	Detect outdated content + suggest updates	S4	impl	—
US-78	Auto-tag + classify uploads	S4	impl	—
US-79	Block unsafe input/output	S0	impl	—
US-80	Redact PII before cloud calls	S0	impl	—
US-81	Per-tenant AI budget + degradation	S0	impl	—
US-82	AI provenance visible to author + admin	S0 (invariant) / S2 (UI)	impl	—
US-83	Export AI audit log	S4	impl	—
US-84	Browse downloadable courses	S1	impl	—
US-85	Download course bundle	S1	impl	—
US-86	Take course offline	S1	impl	—
US-87	AI tutor offline	S1	impl	—
US-88	Offline provisional certificate	S1	impl	—
US-89	Sync queued progress + certificate	S1	impl	—
US-90	Bundle revoked while offline	S1	impl	—
US-91	Manage offline storage + pinning	S1	impl	—
US-92	Edit draft offline	S5	impl	—
US-93	Resolve offline conflict (LWW + diff)	S5	impl	—
US-94	Offline AI in authoring	S5	impl	—
US-95	Local model for tutor	S1	impl	—
US-96	Local model for summary	S1	impl	—
US-97	Cloud refresh on demand	S1	impl	—
US-98	Pre-warm local cache per course	S4	impl	—
US-99	Local AI telemetry replay	S1	impl	—
US-100	Resume on another device	S1	impl	—
US-101	See conflicts across devices	S5	impl	—
US-102	Wipe device remotely	S0 (primitive) / S4 (full UI)	impl/impl	—
US-103	Bundle license revocation propagates	S1	impl	—
US-104	GDPR data export	S0 (primitive) / S4 (full)	stub/impl	—
US-105	GDPR erasure request	S4	impl	—
US-106	Daily Merkle anchoring	S0 (primitive) / S4 (verify UI)	impl/impl	—
US-107	Data residency change	S6	impl	—
US-108	Compliance officer reviews AI provenance	S4	impl	—
US-109	Keyboard navigation across player	S1	impl	—
US-110	Screen-reader compatibility	S1	impl	—
US-111	RTL parity	S1	impl	—
US-112	Captions + transcripts	S1	impl	—
US-113	Reduced-motion preference	S1	impl	—
US-114	High-contrast theme	S4	impl	—
US-115	In-house identity provider (extract native implementation)	S0	impl	—
US-116	Keycloak identity provider adapter	S4	impl	—
US-117	OIDC placeholders (Firebase, Okta, Cognito)	S4	stub	harden post-M3
US-118	Enterprise external IdP & directory federation	S4	impl	—

2. Slice 0 — Platform Foundation (M0)

2.1 Epics in scope

EP-1 (partial: US-1, US-3, US-4, US-115), EP-2 (partial: US-6, US-9), EP-16 (core), EP-20 (primitive: US-102), EP-21 (primitive: US-104, US-106), EP-22 (baseline across all UIs).

Repo implementation note (2026-04-18): Branch features/ep-2-org-management in Ghasi-EdTech implements EP-2 backend slices: US-6 org units, US-7 custom roles + DSL subset, US-8 dynamic groups + evaluate/snapshot, US-9 GET /me/tenants, POST /me/active-tenant + POST /auth/me/active-tenant, refresh rotation id + audit payload, US-10 SSO attribute maps validated against tenant roles where tenant.roles exists, US-11 CSV import jobs + result CSV. Client/UI (TanStack Query, IndexedDB, treegrid, axe, E2E) and full platform DoD (Pact broker, mutation gates, etc.) still belong to CI and frontend repos.

2.2 Cross-service dependencies

• Kong Gateway fronts all north-south HTTP to microservices (JWT/rate limits at edge); see ADR 0001.
• tenant-service depends on identity-service for user lookup on invite.
• ai-gateway-service depends on tenant-service for budget records.
• sync-service registers with every service that later adds replicable aggregates (registration is data-only at M0).
• notification-service used by invite emails and password reset.

2.3 AI-first requirements (slice-scoped)

• Gateway live with complete, embeddings, moderate, redact-pii.
• Prompt registry with 10 system prompts (bot-detection, slug suggestion, tenant description, invite copy, reset abuse check, safety classifier harness, PII redactor test harness).
• Provenance invariant enforceable by domain helper.
• Budget math per tenant functional.
• Local-inference SDK stub (returns "not-ready" to validate port).

2.4 Offline-first requirements (slice-scoped)

• Sync pull/push/resolve endpoints live.
• Client LocalStore port defined for web + mobile.
• Device-binding (identity-service) producing device cert.
• No bundles yet; sync only carries per-user preferences + tenant membership cache + conflict samples (for test).

2.5 Security + multi-tenant requirements

• RLS on every table of every schema.
• Two-tenant isolation test suite passing for every endpoint.
• KMS-per-tenant envelope encryption keys.
• JWT signing + JWKS.
• Audit log append-only + daily Merkle anchor.

2.6 Testing + observability

• Unit + integration + Pact + E2E harness per service.
• OTel end-to-end with traceparent.
• Default SLOs per service (availability, latency, error budget).
• Runbook skeleton per service.

2.7 Sprint-ready backlog for Slice 0 (priority order)

#	Story	Owner team	Dep	Est	Sprint
1	US-79 Block unsafe in/out	AI	—	3	1
2	US-80 Redact PII	AI	—	3	1
3	US-81 AI budget	AI	—	3	2
4	US-82 AI provenance invariant	AI + Platform	—	2	2
5	US-1 Provision org + invite	Platform	—	5	2
6	US-3 Device binding	Platform	US-1	3	3
7	US-4 Password reset	Platform	US-1	2	3
8	US-6 Org units	Platform	US-1	3	4
9	US-9 Tenant switcher	Platform	US-1	2	4
10	US-102 Wipe device (primitive)	Platform	US-3	3	4
11	US-104 GDPR export primitive	Platform	—	5	5
12	US-106 Merkle anchoring	Platform	Audit log	3	5
13	US-109–US-114 (a11y baseline across all M0 UIs)	Design + FE	—	rolling	1–6

3. Slice 1 — Minimal Learner (M1)

3.1 Epics in scope

EP-8 (US-40–44), EP-9 (US-45 stub, US-46), EP-10 (US-49–51), EP-12 (US-56 mini), EP-14 (US-69–71), EP-16 (full: US-79–82), EP-17 (US-84–91), EP-19 (US-95–97, US-99), EP-20 (US-100), EP-22 (US-109–113 on player).

3.2 Cross-service dependencies

delivery-service → content-service (PlayPackage manifest read)
delivery-service → progress-service (statements via events)
delivery-service → ai-gateway-service (tutor calls)
delivery-service → sync-service (offline mount + statement queue)
content-service → identity-service (device pubkey)
content-service → media-service (asset retrieval)
enrollment-service → catalog-service (validate version)
notification-service consumes events from all above

3.3 AI-first requirements (slice-scoped)

• All tutor flows through ai-gateway with pinned prompt delivery/tutor@1.0.0.
• Local-inference SDK produces real outputs (small model) for tutor + summary + worked example.
• SSE streaming from gateway.
• Local AI telemetry replay on sync.
• Provenance attached to every AssistantTurn.
• Budget visible in learner settings; soft-degrade functional.

3.4 Offline-first requirements (slice-scoped)

• PlayPackage Bundles signed + encrypted + device-bound.
• License envelope enforced (expiry, features).
• Player runs fully offline: navigation, quiz scoring (bundled rules), branching (deferred to S4), provisional certificate issuance.
• Statement outbox + AI telemetry outbox flushed on sync.
• Tamper detection on mount + report-tamper on reconnect.
• Revocation propagation on sync.
• Multi-device resume via max(cursor).

3.5 Security + multi-tenant

• Device cert renewal daily.
• Cross-tenant play-session resume blocked.
• Bundle encryption keys derived per (tenant, device, bundle).
• Tamper events audited.
• Chaos tests: corrupted bundle, expired license, revoked license, interrupted sync.

3.6 Testing + observability

• Airplane-mode E2E (J-12).
• Multi-device E2E (J-20).
• Player a11y axe-clean + NVDA + VoiceOver pass.
• Visual regression LTR + RTL + dark.
• Prompt regression on tutor/summary/worked-example.
• Safety eval + red-team corpus run weekly.
• Load: 10 k concurrent online sessions; 1 k concurrent offline-sync pushes.

3.7 Sprint-ready backlog for Slice 1 (selection)

#	Story	Owner	Dep	Est	Sprint
1	PlayPackage schema freeze	Content	M0	8	1
2	US-22 Bundle with license envelope	Content + Security	#1	8	1–2
3	US-22a Tamper detection	Content	#2	3	2
4	US-40 Resume from last position	Learner	—	3	2
5	US-41 Quiz immediate feedback	Learner + Assessment	—	5	2–3
6	US-43 Progress real-time	Learner	Events	3	3
7	US-44 Transcript	Learner	—	2	3
8	US-69 AI tutor in-course	Learner + AI	Gateway + prompts	5	3–4
9	US-70 AI summary	Learner + AI	US-69	3	4
10	US-71 AI worked example	Learner + AI	US-69	3	4
11	US-84–US-91 Offline journey	Learner + Content + Sync	Bundle	34	4–6
12	US-95–US-97, US-99 Local AI	AI + Mobile	Local SDK	16	4–6
13	US-100 Resume on another device	Learner + Sync	Sync	3	5
14	US-49–US-51 Notification prefs + channels + suppression	Platform + Comms	—	7	2–3
15	US-45 stub + US-46 Cert verify	Certification	—	5	5
16	US-56 mini dashboard	Data	—	3	5
17	US-109–US-113 A11y on player + RTL parity	Design + FE	—	13	1–6

4. Slice 2 — Authoring MVP + AI Co-Author MVP (M2, first half)

4.1 Epics in scope

EP-3 (US-12–14 quiz, US-16–17, US-21), EP-4 (US-21, US-22b stub), EP-5 (US-23–26, US-24), EP-13 (US-60, US-62, US-67, US-82 UI).

4.2 Cross-service dependencies

authoring-service → media-service (asset upload)
authoring-service → ai-gateway-service (AI Co-Author flows)
authoring-service → assessment-service (quiz bank ops)
Publish saga: authoring → content → catalog → search

4.3 AI-first requirements

• Prompts: authoring/block_from_intent@1.0.0, assessment/quiz_from_lesson@1.0.0, authoring/simplify@1.0.0.
• AI block generation SSE with accept/reject flow.
• Provenance badge visible in editor + learner view.

4.4 Offline-first requirements

• Authoring editor cached for read-only view offline (full offline edit comes in S5).
• Previously-published bundles continue to work offline.

4.5 Security + multi-tenant

• Block content encrypted in transit; answer keys encrypted at rest.
• Publish saga rolls back cleanly on partial failure.
• Tenant-scoped prompts.

4.6 Testing + observability

• Publish saga chaos tests (mid-saga failure at each step).
• Prompt regression gate on accept rate (≥ 50 %).
• End-to-end E2E: author creates course with AI → publishes → learner sees in catalog.
• Block registry Pact for editor + runtime.

4.7 Sprint-ready backlog for Slice 2

#	Story	Owner	Dep	Est	Sprint
1	Block schema freeze	Authoring	—	5	1
2	US-12 Draft w/ modules + lessons	Authoring	#1	3	1
3	US-13 Core blocks (text/img/video/audio)	Authoring + Media	#1	5	1–2
4	US-14 Quiz block (branching deferred)	Authoring + Assessment	#3	5	2
5	US-16 Submit for review	Authoring	#1	3	2
6	US-17 Publish saga	Authoring + Content + Catalog + Search	Saga infra	8	3
7	US-21 Export HTML5	Content	#6	3	3
8	US-22b SCORM import v1	Content	—	5	4
9	US-23 Browse + filter	Catalog + FE	#6	3	3
10	US-24 Hide unpublished	Catalog	#6	1	3
11	US-25 Locale switch	FE	—	2	2
12	US-26 Course detail	Catalog + FE	#6	3	4
13	US-60 AI generate block	Authoring + AI	Prompts	5	3
14	US-62 AI quiz from lesson	Authoring + AI + Assessment	Prompts	5	4
15	US-67 AI rewrite	Authoring + AI	Prompts	3	4
16	US-82 Provenance badges in editor	Authoring + FE	—	2	3

5. Slice 3 — Marketplace MVP (M2, second half)

5.1 Epics

EP-6 (US-27–33), EP-9 (US-45 issue + US-46 verify), EP-10 (US-52 webhooks), EP-4 (US-19 SCORM 1.2).

Engineering note (2026-04-23): EP-6 M2 slice is in place (listings approve/publish/suspend/appeal, coupons redeem, refund window, payout list + statement PDF placeholder, OpenAPI green). M3+ gaps (PSP/3DS, J-8 bulk, J-17 full, workers, partial refunds) are deferred — see 07 — EP-6 status and the Ghasi-EdTech monorepo ADR planning/adr/ADR-EP6-MKT-001-m3-deferrals.md.

5.2 Cross-service dependencies

Purchase saga: marketplace → billing → marketplace.license.granted → enrollment → notification
Refund flow: marketplace → billing → marketplace.license.revoked → enrollment.revoked → notification

5.3 AI-first

• AI moderation pipeline on listing marketing copy.
• AI fraud-signal classifier on payment intents.

5.4 Offline-first

• Licenses sync to client (read-only) so offline users see assigned courses.
• Refunds propagate to client via sync.
• Marketplace browse cache for offline tiles.

5.5 Security + multi-tenant

• PCI scope minimized via processor-tokenized cards.
• Webhook signatures HMAC-SHA256 + nonce + 5-min window.
• Refund policy per tenant; audit per transition.

5.6 Testing + observability

• SCORM Cloud conformance gate.
• Saga integration tests with mid-payment failure + compensations.
• Processor sandbox tests.
• Public verify page rate-limited; bot mitigation tested.

5.7 Sprint-ready backlog for Slice 3

#	Story	Owner	Dep	Est	Sprint
1	US-19 SCORM 1.2 export	Content	PlayPackage	5	4
2	US-27 Submit listing	Commerce	Catalog	3	5
3	US-28 Pricing plans	Commerce + Billing	—	5	5
4	US-29 Buy as individual (saga)	Commerce + Billing + Enrollment	—	5	5–6
5	US-30 Buy seat pack + assign	Commerce + Enrollment	#4	8	6
6	US-31 Refund within policy	Commerce + Billing	#4	5	6
7	US-32 Provider payouts	Commerce + Billing	#5	3	6
8	US-33 Coupon redemption	Commerce	—	3	6
9	US-45 Certificate issue	Certification	#4	3	6
10	US-46 Public verify link	Certification	#9	3	6
11	US-52 Webhooks (tenant-configurable)	Comms + Platform	—	5	5

6. Slice 4 — Compliance + Enterprise (M3)

6.1 Epics

EP-1 (US-2 SAML, US-5 WebAuthn, US-116 Keycloak, US-117 vendor OIDC stubs, US-118 enterprise IdP federation), EP-2 (US-7, US-8, US-10, US-11), EP-4 (US-20 SCORM 2004/xAPI, US-22b full), EP-7 (US-35–US-39), EP-9 (US-47, US-48), EP-13 (US-61, US-63–US-66 stub, US-68, US-77, US-78), EP-14 (US-72–US-73), EP-15 (US-74, US-76 stub), EP-16 (US-83), EP-20 (US-102 full), EP-21 (US-104 full, US-105, US-108, US-106 UI).

6.2 Cross-service dependencies

assignment-service → tenant-service (DynamicGroup eval) → enrollment-service → notification-service
identity-service SAML + SCIM → tenant-service (membership sync)
Branching block S2 → assessment-service scenario graph
Content 2004/xAPI export → progress-service LRS endpoints

6.3 AI-first

• Full AI Co-Author flows including PDF → course, branching gen, TTS (stub), images (stub).
• AI admin insight v1: assignment recommender.
• AI grading for short-answer with confidence threshold → human review.
• AI audit export from analytics-service.

6.4 Offline-first

• Branching scenario rules bundled for offline play.
• Compliance windows cached for current user offline.
• Pre-warm AI cache per bundle (US-98) enabled.

6.5 Security + multi-tenant

• SAML/OIDC configured per tenant; secrets rotation.
• RBAC + ABAC + DynamicGroup evaluated; two-tenant tests extended to ABAC predicates.
• GDPR erasure saga operational across all services built so far.
• Merkle verify UI operational for compliance officer.

6.6 Testing + observability

• ADL conformance on SCORM 2004 + xAPI + cmi5.
• RRULE fixture suite (≥ 1 000 cases incl. DST + leap).
• SAML tested against Okta, Azure AD, Google, custom ADFS.
• AI bias eval for recommender + short-answer grading.
• Erasure saga E2E covering all in-scope services.

6.7 Sprint-ready backlog for Slice 4 (selection)

#	Story	Owner	Dep	Est	Sprint
1	RRULE engine freeze	Enterprise	—	5	1
2	US-35 One-shot assignment	Enterprise + Enrollment	#1	3	1
3	US-36 Recurring assignment	Enterprise	#2	5	2
4	US-37 Team compliance dashboard	Enterprise + Data	#3	5	2–3
5	US-38 Reminders + escalation	Enterprise + Comms	#3	3	3
6	US-39 Reassign on failure	Enterprise	#3	3	3
7	US-7 Custom role	Platform	—	5	2
8	US-8 Dynamic Group	Platform	#7	5	3
9	US-10 SSO admin config	Platform	—	5	3
10	US-2 SAML SSO	Platform	#9	5	4
11	US-5 WebAuthn	Platform	—	5	4
12	US-11 Bulk CSV import	Platform	—	3	5
13	US-20 SCORM 2004 + xAPI	Content + Progress	—	5	4
14	US-22b SCORM import full	Content	—	5	5
15	US-14 branching (finish)	Authoring + Assessment	—	5	4
16	US-42 Play branching	Learner + Assessment	#15	5	5
17	US-61 PDF → course	Authoring + AI	Prompts	8	5–6
18	US-63 AI branching gen	Authoring + AI	Prompts	5	5
19	US-65 TTS narration	Media + AI	—	3	5
20	US-68 AI SCORM metadata	Content + AI	—	3	6
21	US-72 AI adaptive practice	Learner + AI	—	5	6
22	US-73 AI feedback on open-ended	Assessment + AI	—	5	6
23	US-74 AI assignment recommender	Enterprise + AI	—	5	6
24	US-77 Outdated content detect	Authoring + AI	—	5	6
25	US-78 Auto-tag classify	Authoring + AI	—	3	6
26	US-47 Cert templates	Certification	—	5	5
27	US-48 Recert reminders	Certification + Comms	—	3	6
28	US-83 AI audit export	Data + Security	—	3	6
29	US-102 Full wipe UX	Platform	—	3	4
30	US-104 GDPR export full	Platform + all services	Erasure saga	5	5
31	US-105 GDPR erasure	Platform + all services	—	8	5–6
32	US-106 Merkle verify UI	Security	—	3	6
33	US-108 Compliance AI provenance review	Data + Security	—	3	6
34	US-98 Pre-warm local cache	AI + Content	—	5	6
35	US-114 High-contrast theme	Design + FE	—	3	6

7. Slice 5 — Full Authoring + Offline Authoring (M4)

7.1 Epics

EP-3 full (US-13 refine, US-15, US-18), EP-13 (US-64 production, US-66), EP-18 (US-92–US-94), EP-4 (imports mature), EP-11 (US-53), EP-22 (final depth).

7.2 Cross-service dependencies

authoring-service + sync-service (Yjs + offline mutations)
media-service: AI image, TTS voices, video transcoding matrix
search-service: hybrid search with pgvector index

7.3 AI-first

• Image generation with style profiles per tenant.
• AI merge suggestions in conflict UI.
• AI translation productionized with per-tenant glossaries.

7.4 Offline-first

• Full offline authoring: block edit, reorder, AI assist (local), conflict resolve.
• Offline asset pre-uploads queued; metadata syncs after upload.
• Y.Doc offline merges; non-CRDT fields → LWW with diff UI.

7.5 Security + multi-tenant

• Collab session authenticated + per-message ACL.
• Offline conflict backups retained 30 days per audit policy.
• Tenant-glossary pinning for AI translation.

7.6 Testing + observability

• Yjs + offline merge chaos (two devices, partition, merge).
• Multi-author collab E2E.
• Offline authoring conflict E2E (J-25).
• Hybrid search ranker eval.

7.7 Sprint-ready backlog for Slice 5 (selection)

#	Story	Owner	Dep	Est	Sprint
1	US-15 Live collab (Yjs)	Authoring	—	8	1–2
2	US-92 Offline draft edits	Authoring + Sync	#1	5	2–3
3	US-93 Resolve offline conflict	Authoring + Sync	#2	5	3
4	US-94 Offline AI in authoring	Authoring + AI	#2	5	3
5	US-18 Manual localization	Authoring + FE	—	5	4
6	US-64 AI translate productionized	Authoring + AI	—	5	4
7	US-66 AI image	Media + AI	—	5	4
8	US-53 Hybrid search	Search + AI	—	5	4

8. Slice 6 — Scale + Advanced Insight + Mobile (M5)

8.1 Epics

EP-11 (US-54, US-55), EP-12 full (US-57, US-58, US-59), EP-15 depth (US-75, US-76), EP-20 (US-101, US-103), EP-21 (US-107), EP-6 expansion, mobile native, HIPAA.

8.2 Cross-service dependencies

analytics-service + ai-gateway → insights
search-service hybrid + ranker training
Residency saga: every service + KMS + storage + vector index
Mobile native features: biometrics + background sync + widgets

8.3 AI-first

• AI admin insight v2: board summaries, outdated detection v2, at-risk learner prediction with explainability.
• Personalized recommendations with fairness monitoring.
• Cloud refresh per feature available everywhere local AI runs.

8.4 Offline-first

• Background sync on mobile with OS schedulers.
• Biometric-gated bundle unlock.
• Widget shows next-due compliance.
• Cross-device conflict UX on mobile.

8.5 Security + multi-tenant

• Multi-region data residency saga covering every service.
• HIPAA tier restricts AI providers (explicit no-training configs) and adds BAA-only providers.
• White-label branding per provider with CSP scoping.

8.6 Testing + observability

• Residency saga rehearsal against production-size fixture.
• Mobile app store rating KPI ≥ 4.5.
• ISO 27001 attestation readiness.

8.7 Sprint-ready backlog for Slice 6 (selection)

#	Story	Owner	Dep	Est	Sprint
1	US-54 Personalized recs	Data + AI	—	5	1
2	US-55 Next-step recs	Data + AI	—	3	1
3	US-57 Provider sales dashboard full	Data + Commerce	—	5	2
4	US-58 Scheduled email report	Data + Comms	—	5	2
5	US-59 Raw event export	Data + Security	—	3	2
6	US-75 At-risk prediction	Data + AI	—	5	3
7	US-76 Board compliance narrative	Data + AI	—	5	3
8	US-101 Cross-device conflicts	Sync	—	3	3
9	US-103 Revocation propagation	Sync + Content	—	3	3
10	US-107 Residency change	Platform + SRE + all services	—	13	4–6
11	Mobile native biometric + widget + background sync	Mobile + Platform	—	8	3–5
12	HIPAA tier + BAA providers	Security + AI	—	8	3–5
13	White-label + custom domain	Platform + Design	—	5	4
14	Developer marketplace + SDK v1	Platform + DevEx	—	8	5–6

9. Inter-Slice Dependency Graph

S0 ──► S1 ──► S2 ─┬─► S3
                  │
                  ▼
                  S4 ──► S5 ──► S6

• S2 and S3 overlap within M2 (S2 delivers authoring + HTML + SCORM import stub; S3 delivers marketplace + SCORM 1.2 + billing + certs).
• S4 depends on S2 publish saga + S3 payment rails.
• S5 depends on S2 block registry + S1 sync protocol.
• S6 depends on everything before.

10. Slice Exit Criteria (concise)

Slice	Exit when…
S0	Two-tenant isolation, AI gateway, sync protocol, event envelope, schema registry, observability all green ≥ 2 weeks
S1	3+ design partners completing offline courses; AI tutor accept-rate > 50 %; airplane-mode E2E green
S2	First tenant publishes a course end-to-end via UI; HTML export verified; AI co-author accept-rate > 50 %
S3	First paid purchase → fulfilled enrollment → certificate; SCORM Cloud pass; payout to provider
S4	First enterprise customer using SAML + recurring assignments in production; SOC 2 Type I; ADL conformance
S5	Offline authoring conflict rate < 1 %; live collab stable; LTI 1.3 ready for embedding partners
S6	Multi-region residency migration on live tenant; HIPAA tier offered; mobile native rating ≥ 4.5