COSMOS-CQAPublic Research Portal

Hosted demo manual

Demo workbook for the public Core Pack workflow.

Use this workbook to review the sample workbench as research infrastructure: inspect the Core Pack, exercise tile labeling, export evidence, and stay inside the public claim boundary.

Open Demo Workbench Repository Workbook

Selective-access application planned. A separate COSMOS-CQA application is planned for verified researchers and institutions. This public workbook covers only the static local-first demo and does not describe active onboarding, authenticated review, or server submission.

Cite this demo release: 10.5281/zenodo.21315776 Project DOI 10.5281/zenodo.21112698

Before you begin

Research-only boundary

The hosted demo is not a production decision system and does not publish validated cosmology diagnostics.

Public label policy

Public truth labels remain hidden in the visible workflow and public DOM text. Dev truth review is only for explicit local development mode.

Local-first data

The static portal has no account system. Downloads, bookmarks, and imports stay in the browser session unless you share them.

Reviewer boundary

Reviewer handoff packets are local JSON artifacts. The public demo does not authenticate reviewers, submit observations, or operate an expert queue service.

Evidence focus

Use exports to inspect contracts, provenance hashes, SBOM references, known limitations, and deterministic replay surfaces.

Audio safety

The optional sonic loop is off until Play is selected, looping is off by default, and sound must be treated as a review aid rather than a diagnostic or therapeutic signal.

Twenty-minute walkthrough

  1. Open the hosted demo. Confirm the banner reports that the synthetic Core Pack fixture is ready.
  2. Inspect Core Pack intake. Review manifest summary, tile passports, evidence references, SBOM references, and diagnostic references.
  3. Review the first tile. Change overlay and palette once, then confirm the caption and tile passport remain readable.
  4. Use viewer tools. Try Zoom, Pan, Rotate, and Reset; transformed clicks should still export normalized source-tile coordinates.
  5. Check audio safety before playback. Read the Audio and Captions notice, keep device volume low, and leave Loop off unless you intentionally need repeat playback.
  6. Pin an observation. Click a tile sector, confirm the note field adds normalized source-tile coordinates, add observation text after the cue, submit the label, and confirm Evidence Workspace shows the observation map.
  7. Review the observation. Select the row in Observation Review, edit note/class/severity, set review status and confidence, and confirm the ledger updates.
  8. Exercise adjudication carefully. Mark an observation as needing adjudication only as a workflow placeholder, enter a queue note, and record one queue decision without treating it as consensus.
  9. Try the calibration wizard. In Evidence Drawer, start Calibration in Wizard mode, complete the three-step gold-tile mini-review, and confirm score/reliability feedback updates the caption and metrics.
  10. Create a bookmark. Use the bookmark control, reload the copied state URL, and confirm tile, overlay, palette, rate, loop, and caption settings return.
  11. Check diagnostic caveats. Read placeholder status and limitations before treating any diagnostic surface as evidence.
  12. Prepare local review packets. Export Reviewer Handoff JSON or Review Return JSON only as local artifacts; no authenticated reviewer access or network submission occurs in the public demo.
  13. Refresh report preview. Use the Reports panel to regenerate the validation preview before export.
  14. Export evidence. Download SBOM JSON, validation report JSON, session JSON, and evidence bundle JSON for local review.

What to record during review

Workflow state Demo URL, browser, viewport size, timestamp, and whether public mode or dev mode was active.
Core Pack evidence Manifest id, tile passport ids, evidence reference paths, SBOM reference ids, and provenance hash subjects.
Interaction evidence Tile navigation, viewer transforms, overlay/palette changes, pin/submit/review behavior, calibration wizard result, bookmark reload result, and export filenames.
Boundary notes Known limitations, diagnostic caveats, public truth-label hiding, and any accessibility or visual issues observed.

Local verification commands

npm --prefix apps/web run test:browser -- test/browser/hosted-demo.spec.mjs
npm --prefix apps/web run test:browser -- test/browser/portal.spec.mjs
npm --prefix apps/web run check:portal-deploy
npm --prefix apps/web run check

The focused hosted-demo test verifies public boundaries and evidence export behavior. The portal test verifies public pages, metadata, handoff, and responsive usability.

Review outcome

A completed workbook pass should leave you with a validation report JSON, optional SBOM JSON, optional session/evidence bundle JSON, and a short note describing any visual, accessibility, or research-boundary concerns. File implementation bugs as GitHub issues and route public research questions through the contact page.

For public notes or reproducibility review, cite the exact demo release DOI 10.5281/zenodo.21315776 alongside exported validation, SBOM, session, or evidence bundle artifacts.