FE-574: Knowledge workspace review surface + lifecycle API

[lunelson]

Amp-Thread-ID: https://ampcode.com/threads/T-019d776b-9a7d-7521-b45c-ee24c4e45fca
Co-authored-by: Amp amp@ampcode.com

[linear]

FE-574 Knowledge workspace review surface + lifecycle API

CRUD/review endpoints for the broader knowledge layer plus a fit-for-purpose workspace for inspecting, editing, and reviewing graph-shaped knowledge. The sidebar may remain a summary/navigation view, but the primary interaction model should no longer assume a narrow tab strip. This slice assumes the redesigned knowledge ontology/graph from 7a + 7b.

[lunelson]

• FE-580: Collapse remaining shared transport and review-status type duplication #45
• FE-579: Greenfield/brownfield first-screen + exploration #44
• FE-545: Local-first storage; npx distribution #43
• FE-578: UI refinements; design adaptations & alignments #42
• FE-543: Expanded rich fixture scenarios + reviews + replannings #41
• FE-574: Knowledge workspace review surface + lifecycle API #39 👈 (View in Graphite)
• FE-573: Project dashboard workflow state #38
• FE-542: Criteria-review mode #37
• FE-568: Requirements-review mode #36
• FE-540: Design mode: commitment and exploration #35
• FE-572: Canonical knowledge model foundation + cutover seam #34
• FE-539: Explicit phase outcomes and scope closure #33
• FE-567: Phase-aware observer extraction #32
• FE-566: Generic knowledge layer schema and sidebar projection #28
• FE-564: Flexible turn-response model #27
• FE-559: Migrate: @anthropic-ai/claude-agent-sdk query() -> @anthropic-ai/sdk client.messages.stream() #25
• FE-538: Entity sidebar (read-only) #24
• FE-537: Observer agent + entity persistence #23
• FE-558: UI foundation: shadcn/ui + Tailwind 4 + AI Elements #22
• FE-557: Spike: observer extraction fidelity #21
• FE-556: Structured interview: client UI #20
• FE-555: Parts-based persistence + context builders #19
• FE-554: Structured interview: scope phase #17
• FE-553: Multi-project routing #16
• FE-541: Rich chat UI with tool calls and reasoning rendering #15
• FE-552: Drizzle ORM and core extraction #14
• FE-544: Turn tree schema and API #13
• FE-535: SQLite foundation and project persistence #11
• FE-534: Walking skeleton SDK to SSE to React #10
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[augmentcode]

🤖 Augment PR Summary

Summary: Adds a read-only knowledge review workspace and a markdown spec export flow, plus richer fixture seeding to support manual/outer-loop testing.

Changes:

• Introduced /project/:id/knowledge client route rendering kind-grouped knowledge items with requirement/criteria review badges and basic dependency context.
• Implemented GET /api/projects/:id/export gated by “all phases closed”, and a server-side markdown renderer (src/server/export.ts).
• Upgraded the export preview UI to fetch export state, show a not-ready message, render the markdown preview, and allow downloading spec.md.
• Added manifest-driven fixture seeding (issue-tracker domain) and merged these scenarios into the seed CLI’s scenario list.
• Updated README/PLAN/SPEC to document the new seeding workflow and reflect completed FE-574 slices.

Technical Notes: Export rendering uses md-pen and the shared knowledgeKindRegistry ordering; the knowledge workspace reuses the existing workspace loader and remains read-only.

_{🤖 Was this summary useful? React with 👍 or 👎}

[augmentcode]

Review completed. 4 suggestions posted.

Comment augment review to trigger a new review at any time.

[augmentcode]

selectedOptionIds is being populated from selectedOptionPositions (positions), but hydration/projection code treats selectedOptionIds as DB option IDs. This will cause seeded fixture turns to lose their persisted selections (e.g., getPersistedSelectedPositions won’t match any options).

Severity: high

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

[augmentcode]

data-confirmation expects proposalTurnId: number, but this code allows proposalTurnId to be null, which can make userPartsSchema parsing drop the persisted user_parts entirely for that turn. That can also mask workflow closure provenance because parsePhaseClosureCommand will fail for invalid payloads.

Severity: medium

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

[augmentcode]

This bare catch {} will silently ignore any failure while loading manifest scenarios (including JSON parse/shape errors), which can make issue-tracker-* scenarios disappear even though they’re documented in the README. Consider whether this should only swallow “file missing” cases or surface/log unexpected errors.

Severity: medium

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

[augmentcode]

If the export fetch fails (404/500/network), the query throws and data stays undefined, so the page renders no error state beyond the header/back link. This can make export failures hard to notice and diagnose.

Severity: medium

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

[lunelson]

Merge activity

• Apr 12, 12:28 PM UTC: A user started a stack merge that includes this pull request via Graphite.
• Apr 12, 12:31 PM UTC: Graphite rebased this pull request as part of a merge.
• Apr 12, 12:32 PM UTC: @lunelson merged this pull request with Graphite.