Sprint 8: Industry detection, AI skill suggestions & onboarding blueprint graph (2026-08-26 to 2026-09-09)

[adoLime]

Sprint Goal

Sprint 8 lays the personalization foundations: the system detects each project's industry and weaves it into skill suggestions, onboarding paths, and blueprints; AI skill suggestions give PMs a head start on role skill sets; and the system-wide onboarding blueprint becomes a curated, admin-editable graph from which project-specific blueprints are derived. The Knowledge Base gains an org-metadata filter. We also finish four carry-over items and fix two production bugs from the dev cluster (stuck GitHub sync runs, Chroma fingerprint regression). Scope is priority-ordered — industry foundation and skill suggestions come first; the blueprint-graph backend and UI are the most likely to slip to Sprint 9.

In scope

A. Industry detection — foundation (highest priority)

• Industry detection: data model & AI evaluation (Backend/AI/Frontend) [Story]: Industry detection — data model & AI evaluation #294
• Industry detection: triggers — auto after ingest + lazy evaluation (Backend) [Story]: Industry detection — triggers (auto after ingest + lazy evaluation) #295
• Industry detection: consumer chains — onboarding path & blueprints (Backend/AI) [Story]: Industry detection — consumer chain (onboarding path, re-anchored) #296

B. AI skill suggestions

• Default skill pool & skill schema extension (Backend) [Story]: Default skill pool & skill schema extension #297
• AI skill suggestion endpoint (AI) [Story]: AI skill suggestion endpoint #298
• Skill suggestion integration — backend + frontend review panel (Backend/Frontend) [Story]: Skill suggestion integration (backend + frontend review panel) #299

C. Onboarding blueprint as a graph

• System-wide blueprint graph — data model & admin CRUD (Backend) [Story]: System-wide onboarding blueprint graph — data model & admin CRUD #300
• System blueprint — AI node derivation engine (AI) [Story]: System blueprint — AI node derivation engine #301
• Project blueprint derivation pipeline & summary corpus feedback (Backend/AI) [Story]: Project blueprint derivation pipeline & summary corpus feedback #302
• Onboarding blueprint graph UI (Frontend) [Story]: Onboarding blueprint graph UI #303
  Note: largest block (4 stories, "High" effort per concept); if it doesn't finish, [Story]: Project blueprint derivation pipeline & summary corpus feedback #302/[Story]: Onboarding blueprint graph UI #303 slip to Sprint 9.

D. Knowledge Base polish

• Filter Knowledge Base artifacts by GitHub org-level metadata (Frontend) [Story]: Filter Knowledge Base artifacts by GitHub org-level metadata #293

E. Carry-overs — finish

• Evaluate multi-agent frameworks to support the team (AI/PO) [Story]: Evaluate multi-agent frameworks to support the team developing SprintStart #241
• Link existing sources to additional projects without re-ingesting (Backend/Frontend/AI) [Story]: Link existing sources to additional projects without re-ingesting #257
• Sidebar — clearer hover-state differentiation (Frontend/UX) [Story]: Sidebar — clearer hover-state differentiation #278
• Delete connected repos/sources from a project (Backend/Frontend) [Story]: Delete connected repos/sources from a project #280

F. Bugs & platform health

• Scheduled GitHub sync leaves ingestion runs stuck in RUNNING forever (Backend) [Bug]: Scheduled GitHub sync leaves ingestion runs stuck in RUNNING forever — sources show endless "syncing" sprintstart-backend#195
• Fingerprint check loads document text on every retrieval — regression from the Chroma paging fix (AI) Fingerprint check loads document text on every retrieval (regression from #172) sprintstart-ai#174

Demo outcomes (Sprint Review 2026-09-09)

1. Create/patch a project → the AI evaluates and stores its industry from ingested artifacts; manual re-evaluation works; the wizard offers an industry fallback field. [[Story]: Industry detection — data model & AI evaluation #294]
2. Industry flows into onboarding: path generation and blueprints are industry-aware. [[Story]: Industry detection — triggers (auto after ingest + lazy evaluation) #295 + [Story]: Industry detection — consumer chain (onboarding path, re-anchored) #296]
3. As a PM creating a role: the AI suggests fitting skills (universal + project-specific, with citations); a review panel accepts/rejects before committing. [[Story]: Default skill pool & skill schema extension #297 + [Story]: AI skill suggestion endpoint #298 + [Story]: Skill suggestion integration (backend + frontend review panel) #299]
4. As an Admin: the system-wide blueprint graph (19 nodes, DAG-validated) is seeded and editable. [[Story]: System-wide onboarding blueprint graph — data model & admin CRUD #300]
5. Derive a project blueprint from the graph: grounded per-node summaries with artifact links become part of the project corpus; first-role-assignment triggers derivation. [[Story]: System blueprint — AI node derivation engine #301 + [Story]: Project blueprint derivation pipeline & summary corpus feedback #302]
6. The onboarding view renders the graph (or its linearized form) with dependency gating and the "My Next Task" end node. [[Story]: Onboarding blueprint graph UI #303]
7. The Knowledge Base filters org-level metadata via the new "Organization" tab. [[Story]: Filter Knowledge Base artifacts by GitHub org-level metadata #293]
8. Carry-overs done: sources linkable to additional projects and deletable from a project; sidebar states distinct; multi-agent framework evaluation documented. [[Story]: Evaluate multi-agent frameworks to support the team developing SprintStart #241 + [Story]: Link existing sources to additional projects without re-ingesting #257 + [Story]: Sidebar — clearer hover-state differentiation #278 + [Story]: Delete connected repos/sources from a project #280]
9. Nightly GitHub syncs complete (no more stuck RUNNING runs); the Chroma fingerprint check no longer loads document texts on every retrieval. [sprintstart-backend#195 + sprintstart-ai#174]