2026-05-18 — UI/UX Phase C (160 round-two derivations)¶

Pure text deliverable: no code change. Round-one covered URL sync, indicators, banners, gates, deep links, sidebars and other "basics" (see 2026-05-18-ui-ux-phase-a.md). Round two deliberately moves into the second-order surfaces — performance, motion, error recovery, observability, mobile, theming, print/export, keyboard, undo/redo, batch, a11y depth, security UX, privacy/telemetry, onboarding, contextual help, advanced search, personalization, collaboration.

Each item is sized so a single change note can take it from idea to implementation. Numbering restarts per menu and uses the menu prefix used in phase A (D=Dashboard, N=New Search, J=Jobs, R=Results, C=Custom DB, L=Lab Tools, T=Terminal, A=API).

1. Dashboard (D21–D40)¶

D21 — prefers-reduced-motion audit. Disable the radial pulse on RefreshRing and the skeleton shimmer when the OS opts out; keep colour cues so meaning is preserved.
D22 — Card pin & reorder. Let the user pin frequently-used cards to the top and drag-reorder the remaining grid; persist to localStorage under dashboard.layout.v2.
D23 — Per-card refresh interval override. Each MonitorCard already has a refresh ring; allow long-press / right-click to pick 15s/30s/1m/5m for that card only, persisted per user.
D24 — "Why is this card empty?" contextual help. Hover on an empty card surfaces a popover with the exact RBAC role / configuration needed, cross-linked to docs/auth.md and the SetupWizard step.
D25 — Pause polling when tab hidden. Listen to document.visibilitychange and suspend useQuery refetch loops; resume with an immediate fetch on return.
D26 — Compact / expanded density toggle. Switch between three densities (compact / regular / roomy) for users on small laptops vs. 4K monitors.
D27 — Card-level error budget. When a single card fails 3 times in 10 minutes, automatically demote its polling to 5-minute interval and show a recovery affordance instead of hammering ARM.
D28 — Last-error timeline. A small clock-icon menu per card listing the last 5 errors with timestamps, so an intermittent 502 isn't lost.
D29 — Cross-card correlated alerts. If AKS is "stopped" and Jobs card is empty, render one combined banner ("Cluster off — jobs will resume after start") instead of two independent degraded states.
D30 — Export card payload. Per-card "Copy as JSON" / "Download .json" button for sharing screenshots in incident threads.
D31 — Print-friendly stylesheet. @media print styles that strip glass blur, expand all collapsibles, and add a generated cover page.
D32 — High-contrast theme. Provide an alternate token set with WCAG AAA contrast for users with low vision; toggle from settings.
D33 — Light-mode polish. Audit glass tokens for a sun-mode variant (off by default) for shared meeting-room displays.
D34 — Mobile / tablet breakpoint. Stack monitor cards single-column below 720 px; convert the action chips to a bottom sheet.
D35 — Keyboard shortcut palette. ? opens a command palette listing all keyboard shortcuts (G then J = Jobs, G then T = Terminal, R = refresh).
D36 — Onboarding tour. First-visit guided overlay (1-shot) that walks through cards 1→6 and ends on the Setup wizard.
D37 — Setup status checklist. Sticky right-rail panel that lists unfinished setup steps ("RBAC missing on ACR", "OpenAPI image not built") with one-click jumps.
D38 — Telemetry consent banner. Explicit opt-in for client-side performance telemetry (Vitals) sent to App Insights, with revoke flow and a "view what we collect" link.
D39 — Per-card observability metrics. Tooltip on RefreshRing shows median / p95 latency over the last hour computed from useQuery.dataUpdatedAt deltas.
D40 — Collaboration share link. Generate a deep link ?sub=…&rg=… (subscription scope only, no secrets) so a teammate opening it lands on the same configuration.

2. New Search (N21–N40)¶

N21 — Resumable upload for query FASTA. Block uploads of large query files into chunks; resume on socket drop without losing the staged blob.
N22 — Local cost estimator. Compute estimated compute hours + storage IO + ACR pull traffic before submit, broken down per node SKU.
N23 — Side-by-side draft comparison. When a draft exists, show a diff with the last successful submission's parameters so the user sees what's different.
N24 — Pre-flight cache. Cache pre-flight results for 60 s so flipping back and forth between summary and editing doesn't re-run the expensive AKS/storage probes.
N25 — Saved presets. Named parameter presets (DB, e-value, max targets, output format) that can be applied with one click.
N26 — Template gallery. Curated query templates per use case (16S rRNA, COVID variant, metagenomics).
N27 — Validation linter for FASTA. Inline warnings for ambiguous bases, too-short sequences, or duplicate IDs before submit.
N28 — Cost / time guardrail dialog. If estimated cost exceeds a user-configurable threshold (e.g. $5), require an explicit "I'm sure" confirmation.
N29 — Run-as scheduled. Pick "run now" vs. "run at " vs. "run when cluster is idle"; queue via Celery beat.
N30 — Resource lock indicator. If another submission is running and the AKS node pool is at capacity, show queued position + ETA.
N31 — Sandbox dry-run. "Dry run" mode that splits + plans but doesn't submit any K8s jobs; reports the planned shard count and the estimated cost only.
N32 — Sticky parameter explanations. A right-side help drawer that always shows the docs for the field currently focused.
N33 — Recent values dropdown. For e-value / max-targets fields, show the last 5 distinct values the user actually used.
N34 — Search-time DB freshness check. Warn if the selected BLAST DB has not been refreshed in N days, with a one-click "rebuild DB" shortcut to the Custom DB page.
N35 — Submit history button. Inline "view my last 5 submits" link in the footer, opens a popover not a new page.
N36 — Keyboard submit. ⌘/Ctrl-Enter to submit when canSubmit is true; the existing tooltip already names the requirement.
N37 — Mobile-friendly textarea. Auto-grow + safe-area padding for iPad/Android tablet use in the lab.
N38 — Draft scope picker. Save the draft to per-tab session storage (current) vs. localStorage (cross-tab) vs. server-side per-user draft (cross-device). Pick at footer.
N39 — File picker improvements. Drag-drop overlay highlights the entire form, not just the file input; reject non-FASTA MIME types early.
N40 — Onboarding sample submission. First-time users see a "Try with example query" CTA that pre-fills a tiny FASTA against a sample DB so they can validate the pipeline without leaving the page.

3. BLAST Jobs (J21–J40)¶

J21 — Bulk delete. Select multiple jobs, confirm once, queue a Celery task per row; respect concurrency limits.
J22 — Bulk export. Selected jobs → ZIP of results manifests + parameter sheets.
J23 — Saved views. Pick filter + search + grouping + columns; save as named "view" (e.g. "Failed yesterday", "My nightly runs").
J24 — Column chooser. Toggle elapsed-time / cost / shard count / DB version columns on/off; persisted per user.
J25 — Status-change toast log. Optional toast when a job a user owns transitions to Completed or Failed, even when the user is on a different page.
J26 — Status-change browser notification. Same as above using the Web Notifications API, opt-in.
J27 — Quick-filter from row. Click on a row's DB pill / region / cluster chip to filter to "same as this row".
J28 — Inline retry / re-submit. "Re-run this job" copies the parameters into a new draft in New Search.
J29 — Group-by toggle. Switch between "by date" (current), "by status", "by DB", "by submitter".
J30 — Timeline scrubber. Scroll horizontally through last 30 days, click a day to filter; visualises burst patterns.
J31 — Cost roll-up bar. Show summed estimated cost for the currently-filtered rows at the top of the table.
J32 — Pagination + virtualization. For users with > 500 jobs, switch to virtualized rendering instead of paging the API.
J33 — Server-side filter & search. Push filter/search down to the backend so the SPA doesn't pull the full job list.
J34 — Optimistic delete with undo. Local row strikethrough → delete after 5-second undo window; cancels the queued Celery task.
J35 — Bulk re-tag / annotate. Add a free-text note column persisted in Table Storage for end-of-experiment write-ups.
J36 — Export to spreadsheet. "Download CSV" with the visible columns and current filter; nothing the user can't already screenshot.
J37 — Owner-only filter. "Show my jobs only" persistent toggle for shared-tenant deployments; respects caller.object_id.
J38 — Job age warning. Row chip when a job has been "Running" for

2× median; clickable to open the AKS pod logs deep-link.
J39 — Mobile row layout. Below 720 px collapse non-essential columns into a "Details" expander.
J40 — Audit drawer. Right-side drawer that shows the row's full Table Storage history events without leaving the page.

4. Job Results (R21–R40)¶

R21 — Stream large hit tables. Render hits incrementally so a 100-row result doesn't block the main thread.
R22 — Virtualised hit table. Re-introduce react-virtual for

1 000 hits so scrolling stays smooth.
R23 — Filter hits by e-value / identity. Two-thumb sliders that filter without re-fetch.
R24 — Per-hit "view alignment". Inline expandable alignment block using the existing pairwise renderer.
R25 — Highlight matching DB version. Visual chip when the result used a DB version that is no longer the latest, with a re-run shortcut.
R26 — Copy-as-table. Copy the visible hit table as TSV or Markdown for pasting into a lab notebook.
R27 — Export ASN.1 / pairwise / XML. Existing outfmt selector + download path; new buttons in the toolbar instead of the menu.
R28 — Print one-pager. @media print layout that drops chrome and prints just the parameter sheet + hit table + alignments.
R29 — Permalink to a hit. #hit-${accession} deep-link expands the row + scrolls into view (mirrors API A2).
R30 — Annotation overlay. Toggle GeneBank annotations on top of alignments when available (via the OpenAPI proxy).
R31 — Save annotated session. Persist user annotations / highlights back to results blob (append-only) for resume.
R32 — "Open in IGV" launcher. Generate an IGV session URL for hits whose accession resolves to a UCSC track.
R33 — Compare two jobs. Side-by-side view of two submissions' hit tables, with a Venn-style overlap header.
R34 — Pipeline timing breakdown. Stacked bar of provisioning / DB download / split / search / merge so the user sees where time went.
R35 — Cost actuals. Replace the estimate with the actuals once the job is Completed; expose the delta.
R36 — Reproducibility manifest. "Download manifest" that bundles parameters + DB version + image digests + caller identity for archival.
R37 — Share-this-result link. Read-only deep link a teammate can open (no PII; respects RBAC server-side).
R38 — Storage-locked recovery polish. Existing StorageLockedPanel should additionally surface the IP-allowlist helper script command line and a one-click "Copy command".
R39 — Streaming stderr tail. Live tail of any merge errors via the terminal_exec server while results are being assembled.
R40 — Empty-result hint. When 0 hits, distinguish "actually 0 hits" from "filter excluded all hits" — show both states clearly.

5. Custom DB Builder (C21–C40)¶

C21 — Resumable FASTA upload. Same chunking story as N21; large custom DBs are the biggest upload pain point.
C22 — Background build with notify. Detach the build step from the page so the user can navigate away and get a notification on completion.
C23 — Validation report. After upload, show count of sequences, total length, ambiguous-base ratio, duplicate IDs — before the user pays the makeblastdb cost.
C24 — Cost estimate before build. Bytes → blob storage cost + estimated build CPU minutes.
C25 — Versioned DB ID. Auto-append vN suffix when re-publishing the same name so previous submissions can still reference the old version.
C26 — Side-by-side DB diff. When re-building, diff the new sequence set against the previously published version.
C27 — Build logs viewer. Stream makeblastdb stderr/stdout via the terminal_exec server with collapsible sections per phase.
C28 — Retry from any step. If "Build & publish" fails, allow retry from that step alone without re-uploading the FASTA.
C29 — Saved configurations. Persist (db type, masking options, description) per user so a recurring weekly build is one click.
C30 — Database catalogue page. Link out to a Lab Tools subpage that lists all custom DBs with metadata, build date, size, and an "use in new search" CTA.
C31 — Soft-delete with restore window. 7-day restore window for deleted DBs; permanent only after explicit purge.
C32 — Multi-FASTA merge wizard. Upload N files, dedupe IDs, build as one DB.
C33 — Schedule weekly rebuild. Celery beat schedule for the saved config; toggle on/off in the page.
C34 — Quota awareness. Warn when the Storage account is approaching its quota or maxIngressMbps before allowing the upload.
C35 — Mobile read-only view. Show recent builds and statuses on mobile (no upload), so a lab manager can check without a laptop.
C36 — Keyboard "Next" navigation. ⌘/Ctrl-→ moves to the next wizard step when its done predicate is true.
C37 — Onboarding tour. First-visit tour that explains the 3 steps and links to the elastic-blast docs for FASTA format.
C38 — In-page docs sidebar. Sticky right-rail that pulls the matching azure-prereq.md section based on the current step.
C39 — Privacy classification. Pick "internal / collaborator / public" tag; warn if user picks "public" with a non-empty IP allowlist.
C40 — Audit log link. Surface the DB's append-blob audit history with caller identity hashes for compliance evidence.

6. Lab Tools (L21–L40)¶

L21 — Tool catalogue search. Top-of-page input that filters the Cost / Preprocess / Primer / Taxonomy / Schedules / DbVersions / Audit tabs by tool name.
L22 — Keyboard tab navigation. Left/Right arrow keys move between tabs when focus is in the tab strip.
L23 — Recently used tools. A small "Recent" tab that lists the last 5 tabs the user actually used.
L24 — Per-tool last-error indicator. Red dot on a tab when its underlying endpoint last returned a degraded payload.
L25 — Tools status board. A "tool health" header strip listing which tools are backed by real Celery tasks vs. still 503-stubs.
L26 — Pin a tool as default. Set any tab as the landing tab, overriding the global DEFAULT_TAB.
L27 — Tool-level dark/light variants. Some tools render graphs; pick a matching theme per tool.
L28 — Cost tool — what-if simulator. Slider for node count + SKU
hours, real-time recompute.
L29 — Cost tool — month-to-date roll-up. Sum of estimated and actual costs across all completed jobs this month.
L30 — Preprocess tool — sequence stats panel. Length distribution histogram + GC content for staged queries.
L31 — Primer tool — Tm calculator. Quick melting-temperature calculator with salt + dNTP inputs.
L32 — Taxonomy tool — TaxID lookup. Input a TaxID, surface lineage + child counts.
L33 — Schedules tool — calendar view. Replace the table with a week/month calendar of beat schedules.
L34 — Schedules tool — disable all toggle. "Pause all schedules" master switch (with confirmation) for incident response.
L35 — DbVersions tool — diff. Compare two DB versions' size, sequence count, build date.
L36 — Audit tool — full-text search. FTS-style search over the append-blob audit stream with date / actor filters.
L37 — Audit tool — export. Download the filtered audit as CSV.
L38 — Audit tool — anomaly highlights. Visually flag rows where the caller's IP / role differs from their usual pattern.
L39 — Onboarding tour per tool. First-visit tooltip per tab when the user opens it for the first time.
L40 — Cross-tool deep links. Each tool's row in another tool can link back (e.g. Cost row → Schedules' matching beat entry).

7. Browser Terminal (T21–T40)¶

T21 — Session list & multiplex. Multiple labelled tmux windows side by side with a tab strip.
T22 — Reconnect-with-state. On disconnect, replay the last N lines of scrollback before resuming live (uses tmux's capture-pane).
T23 — Search in scrollback. ⌘/Ctrl-F search across the scrollback buffer.
T24 — Copy / paste polish. "Copy on select" toggle and an explicit paste-from-clipboard menu (some lab browsers block automatic paste).
T25 — Slash-command suggestions. Type "/" to get a popover of common elastic-blast and kubectl recipes.
T26 — Pre-baked recipe palette. "Run elastic-blast status every 10 s" as a one-click action.
T27 — Output capture as artefact. "Save last command output to Files share" button so terminal logs don't have to live in scrollback.
T28 — Font / contrast settings. xterm.js theme + font-size picker with live preview.
T29 — Reduced-motion cursor. Disable cursor blink when prefers-reduced-motion.
T30 — Mobile-friendly soft keyboard row. Esc / Tab / arrows / pipe / Ctrl row above the keyboard for iPad use.
T31 — Session timeout policy. Surface the remaining session TTL
warn 60 s before disconnect.
T32 — Per-keystroke audit toggle. Off by default; when enabled every keystroke is timestamped to the audit blob (incident-response use only).
T33 — Onboarding banner. First-time hint to run az login --use-device-code (complements T2 indicator).
T34 — In-terminal help. ? keystroke opens a side drawer listing all elastic-blast / elb commands.
T35 — Active-job ribbon. Sticky strip showing the currently active elastic-blast submission with progress.
T36 — Restart-terminal confirmation. Confirm dialog with explicit "This drops your tmux session" warning before sidecar restart.
T37 — Reset-home guard. Multi-step confirmation + typed-name guard for resetting /home/azureuser (drops cached az login).
T38 — Shareable session id. "Share session id" copy button so collaborators can join with tmux attach -t <id> (RBAC enforced).
T39 — Telemetry: command exit codes. Aggregate exit-code chart over last 24h for the cockpit.
T40 — Privacy redaction. Auto-redact tokens / SAS strings rendered in the terminal output (mirrors the api-side sanitiser).

8. API Reference (A21–A40)¶

A21 — Operation history. Per-endpoint "last 5 calls I made" log with status + duration; replays one-click.
A22 — Save request as preset. Name + persist a fully-populated body so the user can re-run with one click.
A23 — Compare two responses. Side-by-side JSON diff between two recent executions of the same endpoint.
A24 — Query playground panel. A /api/blast/search style playground (akin to GraphiQL) for ad-hoc exploration.
A25 — Per-endpoint cURL generator. Auto-generate a copy-pasteable cURL with the live token and params filled in.
A26 — Per-endpoint code samples. Toggle: cURL / TypeScript fetch / Python httpx.
A27 — OpenAPI version banner. Show spec version + last refresh time at the top of the sidebar, with a manual refresh button.
A28 — Mark deprecated endpoints. Cross-reference deprecated: true from the spec and render a strikethrough in the sidebar list.
A29 — Tag descriptions inline. Surface the OpenAPI tag description in the sidebar as a hover popover.
A30 — Try-it sandbox / production toggle. Run against the live cluster or against a recorded fixture (no side effects).
A31 — Saved environments. Cluster A / Cluster B / local — switchable from the sidebar.
A32 — Request schema linter. Inline warnings when the user's JSON body doesn't match the OpenAPI schema before they hit Try-It.
A33 — Response schema validator. Highlight fields in the response that don't match the spec — surfaces drift between spec and runtime.
A34 — Error catalogue. A dedicated tab listing all error responses' detail.code values + remediation steps.
A35 — RBAC matrix overlay. Tag each endpoint with the minimum RBAC role; filter "show me endpoints I can call".
A36 — Rate-limit indicator. Show the per-endpoint quota (when the backend publishes one) with a current-usage bar.
A37 — Print-friendly spec dump. Single-page printable spec view with @media print polish.
A38 — Onboarding for new endpoints. "What's new since you last visited" diff: endpoints added / changed / removed.
A39 — Mobile read-only view. Below 720 px hide the Try-It UI and render endpoints as a long-scroll spec viewer.
A40 — Collaboration / link-to-step. Deep link a particular request (method + path + body preset id) so a teammate opening it lands in the playground pre-filled.

For the next phase to schedule, the following are worth promoting from the per-menu lists into shared infrastructure:

Reduced-motion + high-contrast theme (D21, D32, T29). One token set serves everyone.
Mobile / tablet breakpoint pass (D34, J39, T30, A39, R/N for read-only). Same media-query rules.
Print stylesheet (D31, R28, A37). Single @media print file in web/src/theme/.
Resumable upload (N21, C21). Shared chunked-upload hook.
Onboarding tour engine (D36, N40, C37, T33, L39, A38). Single stepper component driven by per-page configuration.
Saved presets / personalisation (J23, L26, N25, R31, A22, A31). Backed by user-scoped settings in Table Storage.
Streaming / virtualisation (J32, R21, R22). TanStack Virtual in one place.

Phase B (when scheduled) should pick the top 5 of these to land the infrastructure once, then iterate the per-menu hookups.