Skip to content

DONE-5c8f2a57.md

Latest commit

 

History

History
138 lines (120 loc) · 7.42 KB

DONE-5c8f2a57.md

File metadata and controls

138 lines (120 loc) · 7.42 KB
Raw

DONE evidence — task 5c8f2a57 (doc workspace for transcripts, timecodes, summaries: txt/docx export, guest review)

Verified against origin/main fdfc6929 (2026-11-05) in a dedicated worktree, plus read-only prod checks through the dataops CLI. All three asks are already implemented; the only real gaps found were two export Content-Type misalignments, fixed in this PR (two one-line route changes + assertions in the existing tests).

(a) Transcript/timecodes/summary doc workspace — DONE

The same four kinds exist in both doc families, and both are reachable per card and per task:

  • Kinds are transcript, timecodes, summary, note in both families: DOCUMENT_KINDS (backend/src/db/documents.ts:8) and CARD_DOCUMENT_KINDS (backend/src/db/cardDocuments.ts:8); enforced by the create routes (backend/src/routes/documents.ts:59, backend/src/routes/cardDocuments.ts:69).
  • Task documents (guest-reviewable): dataops doc list|show|create|update|delete|export (cli/src/commands/documents.mjs:7-17) over /api/documents (backend/src/routes/documents.ts); portal UI creates them from a kind picker and edits/exports them (frontend/src/surfaces/work-detail/task-documents.js:25-28, 33-46).
  • Card documents (staff-only, the per-card workspace): dataops card-doc list --card ID / create --card ID --kind transcript … (cli/src/commands/cardDocuments.mjs:6-15); the API lists by cardId and creates attached to the card (backend/src/routes/cardDocuments.ts:39-42, 67-72).
  • Podcast episode link: dataops podcast create-card <episode-id> creates the episode Card and stores its id on the episode (cli/src/commands/podcast.mjs:54-56; PodcastEpisode.cardId, backend/src/db/podcastEpisodes.ts:77). Prod has 10 episodes with linked Cards; card-doc list on those cards works (empty today — the feature exists, nobody has attached transcripts yet). The episode record itself also carries the website page transcript data (EpisodeTranscriptEntry[] with sec/time/who, backend/src/db/podcastEpisodes.ts:31-39, 90) and the readiness transcript observation (backend/src/podcast/publication.ts).
  • Guest-facing materials for an episode live in the third family: dataops podcast doc create <card-id> --template podcast|podcast-guest-research|podcast-run-sheet|podcast-episode-brief (cli/src/commands/sponsorDocs.mjs:47-68), linked on the Card and served at /podcast-docs/:id.
  • Copy within a family: --from ID on both doc create and card-doc create (backend/src/routes/documents.ts:49-53, backend/src/routes/cardDocuments.ts:59-63; verified by 0101cb2b).

The smallest slice the task asked for ("create doc of kind transcript linked to an episode card, list docs per card") is exactly card-doc create --card <episode-card> --kind transcript … + card-doc list --card <episode-card>, both present. No new flow was needed.

(b) txt export — DONE (Content-Types aligned in this PR)

Every doc-family export endpoint accepts txt:

  • Task documents: GET /api/documents/:id/export?format=txt → text/plain; charset=utf-8 (backend/src/routes/documentExports.ts:14, 21).
  • Card documents: GET /api/card-documents/:id/export?format=txt → text/plain; charset=utf-8 (backend/src/routes/cardDocuments.ts:25, 29).
  • Sponsor/podcast docs: GET /api/sponsor-docs/:id/export?format=txt (backend/src/routes/sponsorDocs.ts:827-840).
  • Knowledge/registry docs: GET /docs/export?format=txt → text/plain (backend/src/docs/contentApi.ts:561, 571).
  • CLI: dataops doc export <id> --format txt, dataops card-doc export <id> --format txt, dataops podcast doc export <id> --format txt (cli/src/commands/documents-transfer.mjs:9-10); DOCX verified by 0101cb2b.

Fixed here: sponsor-docs served txt with Content-Type: text/markdown (the format was appended to the markdown branch in f2780c8e without adjusting the header) — now txt/text export as text/plain; charset=utf-8 (backend/src/routes/sponsorDocs.ts:834). Card-documents served md as text/plain — now text/markdown; charset=utf-8, matching the task-document and docs endpoints (backend/src/routes/cardDocuments.ts:29).

(c) Guest review — DONE (view + comment without an account)

Two existing mechanisms cover it; no new auth was added:

  • Document review links (task documents of every kind, so transcripts and summaries included): staff mints a reviewer token via POST /api/documents/:id/reviewers and receives /document-review/:docId?review=<token> (backend/src/documents/reviews.ts:57-59); the guest page sends the token from the URL as x-document-review-token (frontend/src/document-review-page.js:7-20). Anonymous guest requests are allowed exactly for GET /api/documents/:id/review (document read) and GET|POST /api/documents/:id/comments(/…/reply|resolve) (threaded comments) — isAnonymousDocumentReviewRequest (backend/src/documents/reviews.ts:13-15). Guests cannot manage reviewers (reviews.ts:43) or delete threads (reviews.ts:68-69); the review read strips taskId so the link leaks nothing else (asserted in backend/tests/document-reviews.test.ts:40-66). Tokens are scoped, hashed, expiring and revocable (backend/tests/document-reviews.test.ts:13-23). Staff UI: "Guest review and comments" panel on every task document (frontend/src/surfaces/work-detail/task-documents.js:77-86). Agents can mint links without the portal via the generic API passthrough (dataops api POST /api/documents/<id>/reviewers; MCP dataops_write reaches the same routes, mcp/src/tools.mjs:12-39).
  • Sponsor-docs guest-code flow (webinar/workshop/podcast guest materials): anonymous GET|PUT /api/sponsor-docs/:id and the export are gated by the per-doc access code and rate limiter (backend/src/routes/sponsorDocs.ts:175-190), and guests see the speaker view of the content (speakerContentView, backend/src/routes/sponsorDocs.ts:352, 800-806). Card documents are deliberately outside both: card docs are staff-only and the review loader cannot see them (getDocument returns null for card docs; asserted in backend/tests/card-documents.test.ts:36).

Changes in this PR

  • backend/src/routes/sponsorDocs.ts — txt/text export Content-Type is text/plain; charset=utf-8 (was text/markdown).
  • backend/src/routes/cardDocuments.ts — md export Content-Type is text/markdown; charset=utf-8 (was text/plain); txt unchanged.
  • backend/tests/sponsor-docs.test.ts — the all-template export test now asserts md → text/markdown, txt → text/plain, and that both bodies are identical.
  • backend/tests/card-documents.test.ts — the export loop asserts the per-format Content-Type.

Tests

  • backend/tests/card-documents.test.ts — 6 pass, 0 fail.
  • backend/tests/sponsor-docs.test.ts — 21 pass, 0 fail.
  • backend/tests/documents.test.ts + backend/tests/document-reviews.test.ts — 15 pass, 0 fail.
  • npx tsc --noEmit (backend) — clean.

Remaining gaps

  • No bridge converts an episode's website transcript data (episode.transcript) into a card transcript document; staff paste or upload it (card-doc create --file). The task scoped (a) to create-plus-list per episode card, which exists, so this is a candidate follow-up, not a gap against the ask.
  • There is no dedicated CLI command for reviewer management (review links are minted via the portal panel or the generic dataops api passthrough).