# Web client — parity checklist The browser client's acceptance gate. **The target is the Android app's feature set** (`android/`), rendered for the web. A row is only `done` when the feature works in the browser against a live server **and** a review stage captured the rendered result through `test/ui-smoke.mjs` (or an extension of it). A green unit test, or a code change without a capture, is `partial` at best. Status legend: `todo` · `partial` (some of it works) · `done` (working + captured). ### Share artifact files, not links (2026-10-21) Gallery/list/visit and viewer Share actions now fetch the artifact and pass a named, MIME-typed `File` to `navigator.share({files})`. No URL is substituted. Copy link remains separate. Browsers without file-sharing APIs disable Share; unsupported types, failed fetches, cancellation and blocked sharing have truthful feedback. A slow download that outlasts user activation asks for a fresh Share tap using the prepared file, rather than silently failing or sharing a URL. Preparation is deduplicated per button and canceled if its view goes away. `test/artifactSharing.test.mjs` checks exact binary bytes, filename/MIME, file-only payloads, HTTP/network failures, unsupported types/APIs, cancellation/denial, expired activation, navigation away and concurrent clicks. The real Chrome `test/artifact-browser-probe.mjs` drives gallery/list/viewer Share and feedback states with a stubbed native Share API; captures at 390/420/1100px are in approved OpenCode temporary `artifact-file-share-captures/`. Native OS chooser/recipient delivery is **not tested**. Shell cache is v16. This supersedes earlier URL-share behavior without changing the artifact's external-open action or authentication. ### Image artifact primary preview (2026-10-21) Image thumbnails and image list rows now open the existing in-app viewer. The former eye action is replaced by an external-arrow **Open in browser** action, which does not also activate the viewer. This applies to both the run gallery and individual visit artifacts; other file types retain their current actions. The viewer's own Open in browser, Copy and navigation actions remain unchanged. `test/artifact-browser-probe.mjs` verifies gallery/list/visit clicks, keyboard activation and return focus, preview without a new tab, actual browser-tab opening from the external image action, and installed-Android intent routing without double activation. Captures at 390/420/1100px are under the approved OpenCode temporary `image-primary-preview-captures/`. Android OS hand-off is not tested. This supersedes prior external-primary image behavior; shell cache is v15. ### Chat metrics behind Info (2026-10-21) The four usage/performance values now live in an explicit **Chat information** panel, opened by the always-visible Info button at the far right of the chat's top bar. The permanent metric rail is removed on mobile and desktop. The panel overlays rather than shrinking the conversation, still receives live numeric updates, and keeps unknowns distinct from reported zeros. `test/chat-usage-browser.mjs` verifies collapsed/open/unknown/partial/zero states at 390/420/1100px and a short viewport; metrics arriving while closed remain available on open. It checks unchanged transcript height/DOM, preserved draft and selection, trapped panel focus, return focus, Escape, browser Back, Close and backdrop dismissal, accurate expanded state, and the crowded mobile toolbar with both history and approval controls. Captures: approved OpenCode temporary `chat-info-captures/`. A modal lifecycle cleanup hook also handles programmatic closure without leaving stale Info state. Shell cache is v14. This supersedes the permanent-header-rail placement described in the earlier usage entry below; the native metrics protocol/accounting is unchanged. An authenticated live 390px browser check confirmed the collapsed default, four-metric panel, and browser Back closing it without leaving the conversation. This frontend-only update required no backend/service restart. ### Run graph opens recorded visits (2026-10-21) Run-graph stage clicks now call the same visit-pane action as Timeline, selecting the stage's latest recorded visit. Earlier visits remain selectable from Timeline; messages/tool turns within one visit remain together in its Session audit. A graph click no longer fetches or defaults to the authored definition. Definition remains an explicit Audit & debug tab, pinned to the run's pipeline version. Unvisited nodes do not fabricate visits or sessions. Workflow-definition graph clicks still open the definition viewer. `test/run-graph-visit-browser.mjs` compares the graph and latest Timeline pane contents/routes at 390/1100px, verifies a loop's latest session identity, no definition request on default open, unvisited-node safety, and explicit pinned definition access. Captures: approved OpenCode temporary `run-graph-visit-captures/`. `test/ui-stage-definition.mjs` retains its source, missing/loading/error and keyboard checks through the now-explicit run audit. This supersedes the older run-graph definition-popup default described below. The deployed authenticated browser also compared run #19's actual graph and Timeline pane contents successfully, without altering that run or its sessions. ### Independent schedules and provenance filters (2026-10-21) Pipeline Summary/Schedules manages multiple independent run configurations for the same pipeline/version: name, prompt, server workspace, timer and optional run options. Schedule this run seeds original input/title/options and preserves run-level write restrictions without changing workflow tool permissions. Creates reuse one request key across response-loss retries; updates/deletes use revision CAS and preserve conflicting drafts. No workflow publication occurs. Pipelines filters distinguish configured schedules (including paused) from none; Runs filters distinguish actual timer origins from nonscheduled runs. A manual run of a configured pipeline is not tagged scheduled, and deletion retains old run provenance. Older peers without fields remain unclassified. Sparse live frames cannot clear/retarget known provenance. The header provides Refresh; no per-step scheduling poll or new workflow is used to recover skipped dispatches. `test/schedules-browser.mjs` covers edit/create/pause/delete, conflict handling, unknown-server responses, keyboard/focus and 390/1100px. The real scratch-server `test/schedule-integration-browser.mjs` checks two independent topics, both filters, retained CAS drafts, Schedule this run preserving explicit title/input/ write scope, an actual claimed timer's first-frame provenance versus manual execution, unchanged workflow source, and history after deletion. It uses only human stages, explicit titles, notify=off and a fixture-only NextDue advancement; it initiates no provider turns or live-user runs. Captures: `schedule-editor-captures/` and `schedule-integration-captures/` under the approved OpenCode temporary directory. Shell cache v13 refreshes the new modules/styles. After deployment, authenticated read-only checks verified the schedule API, typed pipeline/run scheduling metadata and the actual editor. No production schedule, workflow version or run was created for verification. ### Always-visible chat row actions (2026-10-21) Star is always visible at the top right and Chat actions at the bottom right of a reserved trailing column that stretches with the row. The non-shrinking harness mark precedes the title inside the native session button. Native sibling action buttons have 44px targets (48px for coarse pointers), common ghost-button background/border/hover/active/focus tokens, and no hover reveal or opacity/layout transition. Only starred chats use an accented star. Session interaction feedback belongs to the outer row, matching an ordinary row without an inner rectangle. Only explicit activation opens the action menu; its existing Rename/Close/Reopen/Hide items and shared star API are unchanged. Menus preserve list scroll and anchor inside the viewport outside scroller clipping. `test/chat-row-actions-browser.mjs` uses actual `createChatView` with mock repo/rt, not native chats. It captures long-title/path, selected/starred, streaming, new, closed, hover, keyboard, explicit-menu and scroller-bottom states at 390/420/1100px, plus coarse-pointer 390px. Geometry and hit-tests check top/right and bottom/right anchoring (including taller content), left-before-title placement, non-overlap, target sizes, no hover-open/reveal, keyboard order, shared-star calls, retained selected chat, Rename activation and exactly one route dispatch per thread/logo click. CDP-forced normal/hover/active/hover+active/focus-visible states compare computed styles with ordinary ghost and row buttons (selected and unselected), including coarse input; real pointer movement, clicks and keyboard Tab/Enter separately verify behavior. Capture manifest: approved OpenCode temporary `chat-row-layout-followup-captures/manifest.json` (22 `topright-leftlogo-bottomright-*.png` captures). Usage-rail, composer-fit and miniatures browser regressions pass unchanged; 723 unit tests pass. Live authenticated/native-chat verification and deployment are not performed by this isolated fixture change; no parity status is promoted. ### Chat performance and usage (2026-10-21) The chat header now shows reported output tokens/sec, total processed tokens, last prompt/context usage and context-window limit, and reported USD cost. Missing values are explicitly **Unavailable**; a genuine reported zero stays zero. Native snapshots replace previous values, including duplicate/reconnect updates and context reductions after compaction. Cumulative processed tokens are never used as context size, output text is never used to estimate tokens, and unreported subscription costs are not guessed from a pricing table. Metrics-only updates patch the header rail without repainting the transcript or composer, changing focus/selection, creating messages, or advancing a queued prompt. Last reported values remain available when the turn settles. `test/chat-usage-browser.mjs` captures known, missing, partial and zero-valued snapshots at 390/420/1100px plus a 390px short viewport. It checks formatting, no horizontal overflow, unchanged message count, no spurious Stop control, duplicate-snapshot replacement and preserved composer/transcript DOM and cursor. Captures: approved OpenCode temporary `chat-usage-captures/` directory. These fixtures initiate no provider turns. Native adapters have separate regression fixtures and retain/replay numeric snapshots without synthetic transcript items. OpenCode V2 reports session counters/cost and model metadata; measured rates use non-reasoning output and native request-to-stream timing (including request latency, excluding tool execution). Codex reports token/context snapshots but provides no compatible generation-rate or USD billing data, so those values remain unavailable. Shell cache v12 refreshes the changed modules on reload. After deployment, an authenticated live browser received all four metrics for the existing owned OpenCode chat and formatted them through the shipped module. No provider turn/session was created for that check; private usage figures are not included in the committed evidence. Native requests and credentialed access remain server-side. ### Context compaction activity (2026-10-20) **Partial against the live-provider gate.** A native compaction frame shows **Compacting context…** with the active turn's spinner and Stop control, even before the first answer or alongside existing reasoning/tools. Previously stored reasoning is labeled Reasoning, not actively Thinking, while context compacts. The label clears on compaction completion/failure; that does not finish the surrounding model turn or count as a tool/message. Disconnect/turn settlement clear stale activity. Reconnect clears stale state before subscribing, then keeps fresh replayed activity while authoritative history loads. OpenCode V2's native compaction lifecycle and Codex's `contextCompaction` items are normalized by the server, with retained active replay and native turn/item identity checks where available. No state is inferred from slow responses or model-authored text. Backend regression fixtures exercise the native schemas; no live provider compaction was triggered for verification. `test/chat-compaction-browser.mjs` captures before-answer and with-history states at 390/1100px plus cleared-on-disconnect. It checks polite status announcements, no horizontal overflow, preserved message counts, Stop lifecycle, and actual reconnect-handler replay surviving the history read. Captures: approved OpenCode temporary `chat-compaction-captures/` directory. Fixtures do not trigger compaction, inference or changes to real chats. A server update is needed for the new native activity frames; shell cache v11 refreshes the browser modules. ### Tool summary error scope (2026-10-20) The aggregate **N tools** label stays neutral and borderless even when a call fails. Only the separate **· N errors** count is red; individual failed calls remain marked inside the tool popup. Keyboard focus keeps its accessible focus ring. This is an intentional browser presentation correction, not a protocol or Android change. `test/tool-summary-browser.mjs` verifies 82 calls with zero, one and two errors at 390/1100px, neutral aggregate color, transparent border, no resting outline, correct accessible error counts, and exactly one failed row among 82 popup rows. Captures are in the approved OpenCode temporary `tool-summary-captures/` directory. Fixture-only browser evidence; no live chat/session was changed. Shell cache v10 refreshes the changed modules on reload. ### Generic alpha starters (2026-10-20) The embedded catalogue now presents small code change/review, research/fact-check, document summary, option comparison, and plan/checklist starters, alongside the retained legacy keys. No project path, stack, provider account or model is baked into these starter definitions. No existing published definitions are changed. `test/ui-smoke.mjs --scratch` verified all eight entries and the real validate/create/list flow against a current-source Go server on a free loopback port with a scratch database and explicitly enabled authoring. Desktop and 390px gallery captures check overflow and card metadata containment; grid rows now keep their content height rather than clipping the stage counts. The browser driver scrolls instantly before measuring click targets, avoiding stale positions under CSS smooth scrolling. Captures: approved OpenCode temporary `alpha-template-browser-captures/51-template-create-dialog.png`, `52-template-created.png`, `53-pipeline-created-list.png`, `54-alpha-starter-gallery-desktop.png`, `55-alpha-starter-gallery-mobile.png`. These checks create no live definitions and perform no inference. Real research tools, provider outcomes and deployment are not established by this check. ### Session capture and export recovery (2026-10-20) **Partial against the live-server gate.** The Session tab now distinguishes a recorded session ID from its missing/failed export, selects a verified complete repair for that same ID, and refreshes an already-open audit when evidence changes. A failed or partial later repair does not hide a complete earlier one. **Retry export only** reads the exact recorded visit; it never reruns its model or tools. Versioned repairs stay in Audit, not the stage deliverables gallery. Export-gated waits are nonterminal and show no human decision controls. `test/session-audit-browser.mjs` captures the recorded-but-failed state and the successful export-only retry at 390px, with sandboxed iframe and unchanged-frame checks. Captures: approved OpenCode temporary `session-audit-captures/`, `01-recorded-identity-export-failed.png` and `02-complete-repair-refreshes-open-audit.png`, plus `03-export-wait-is-not-a-human-decision.png`. These use fixture session data, not a deployed-service claim. Separately, read-only API verification exported all five original run #19 sessions completely (14/11/10/16/11 messages), without rerunning stages. No deployment or live evidence mutation was performed. ### Artifact browser hand-off (2026-10-20) Artifact opens now use one validated navigation helper across gallery/list, verdict, audit/session/repair, search, viewer and chat links. Android standalone apps request an explicit browser-family VIEW intent: an implicit HTTPS intent or `_blank` alone can be captured by the current WebAPK. Browser packages are fixed by the known browser family, never supplied by artifact/model text; other environments retain native `_blank` links with external/noopener/noreferrer. Copy/Share keep ordinary HTTP(S) URLs. No code replaces the app's own location. Markdown root-relative artifact links are clickable; executable/local URLs remain unlinked. The viewer has an explicit **Open in browser** link and wrapping mobile actions. All file types open in the browser by default; **Preview in app** is an explicit secondary gallery/list action for media and Markdown/text. On touch screens these actions are visible, separate from Open, and at least 44px. `test/artifact-browser-probe.mjs` verifies an actual new browser target while the app URL/history remain unchanged, then inspects Android standalone hand-offs for tiles (HTML/Markdown/image), list, search, chat, session and viewer controls. Copy retains HTTP, unsafe links stay inert, and actions fit at 390/420px. Captures are in the approved OpenCode temporary `artifact-browser-captures/` directory. **Android's actual OS/browser launch and Back behavior require the user's phone check**; the desktop test does not claim to reproduce an installed Android WebAPK. Shell cache version is bumped so the new navigation modules refresh on the next app reload. ### List miniature follow-up (2026-10-20) Workflow and run miniatures now share the existing graph layout, right-aligned inside their SVG as well as their row, with a 6px extra trailing inset. Run LIST nodes now use each stage's last visit: success green, failure red, running accent, waiting amber and queued a distinct muted tint. Unvisited/unknown nodes are neutral hollow/dashed; **no aggregate run-status tint is inherited**. Taken edges use recorded outcome/target routes, including loops and branches. The additive `graph_evidence` summary projection reads only five routing/status fields from `run_steps`, compacts repeated loops, and batches 100 run IDs per query (not N detail fetches). Subscribe snapshots seed the same evidence; coalesced live frames preserve every changed visit in `graph_visits` and the client maintains last-stage/distinct-route evidence without a detail fetch. Older summaries remain honestly unknown. Topology comes from the run's own graph, an exact-version pipeline summary, or one cached `get_pipeline(id, version)` read per endpoint/version (never `get_run` per row, never a fallback to latest). Unavailable graphs stay absent. No schema migration or new private input/output/transcript payload is needed. Scratch browser extension `test/miniatures-browser.mjs` verifies 390/420/1100px alignment, two successful stages before failure plus never-visited branches, waiting/running/queued states, legacy unknown evidence, wrapped titles, separate applied filters and no document overflow; captures and `geometry.json` are in `/private/var/folders/jp/q0qvjyg54fl_c5vhhk1ylz2m0000gn/T/opencode/per-stage-run-miniatures-captures/`. These are deterministic fixture captures, not a new live-server parity claim. Collapsed desktop navigation additionally gives Expand its own 40px flow slot above the Pipelines brand instead of overlaying it. The same scratch browser extension asserts non-overlap with brand and first destination, expands again through the real button at 1100px, and checks the rail remains hidden at 390/420px. Captures: `rail-collapsed-1100.png`, `rail-expanded-1100.png` in the directory above. Mobile and the expanded rail are unchanged. ### Stage definition popup follow-up (2026-10-20) P6/R5 follow-up: **partial** against the live-server parity gate; verified with the scratch-only `node test/ui-stage-definition.mjs ../.ui-stage-definition` browser extension. Clicking a stage in either graph now opens the managed audit popup, without replacing the current pane or changing its route. Timeline visit panes remain available, and their Audit & debug Definition tab shares the viewer. Rendered uses the existing Markdown/code renderer; Raw is the exact authored stage source slice (including whitespace, custom blocks and shorthand), not a serialization of the normalized graph. Run definitions request `get_pipeline` with the recorded pipeline id/version, never latest. Missing source and load errors are explicit; terminal nodes have no authored stage definition. Scratch captures: `.ui-stage-definition/01-run-definition-rendered.png`, `02-run-definition-raw.png`, `03-pipeline-definition-rendered.png`, `04-visit-definition-nested.png`, `05-definition-missing-source.png`, `06-definition-load-failed.png`, `07-definition-terminal.png`, and `08-definition-narrow.png`, `09-definition-loading.png`, and `10-definition-missing-version.png`. Real browser assertions cover exact source, pinned version, unchanged route, nested-pane preservation, Tab trapping and Escape/focus restoration. No deployed service, live workflow or device was used. ### Chat audit follow-up (2026-10-20, partial — fresh UI captures pending) Composer fit and vertical alignment follow-up (2026-10-20): attachment, Send and Stop controls (44px square; 48px with a coarse pointer) are vertically centered on the textarea, including multiline drafts. The single-line input matches their height, and six complete auto-grown text lines fit before scrolling. Send remains available while running; its adjacent Stop slot is reserved even while idle so state changes do not rewrap a draft. Monochrome CSS SVG masks replace the platform-dependent button glyph rendering, preserving existing accessible names. Back/title, protocol and keyboard geometry are untouched. The visible placeholder is shortened to `Message the agent…` so an empty field stays one aligned line; the full slash-command hint remains its accessible name. `node test/composer-fit-browser.mjs ../.ui-composer-fit-after` verifies the real chat view and input handlers using deterministic fixtures at 390/420/1100px: empty, single/multiline, running, six/ten-line drafts, attachments and simulated keyboard geometry. `--before` serves HEAD's CSS for comparison (run before committing, or on the parent revision). Captures and `geometry.json` are in `.ui-composer-fit-before/` and `.ui-composer-fit-after/` in the isolated composer worktree. This is scratch-browser evidence, not a live-server parity upgrade; real iOS/Android keyboards were not tested. Integration additionally captured 28 states under the approved OpenCode temporary `final-composer-fit/` directory, including browser-emulated coarse pointers at 390/420px: empty input and buttons are exactly 48px, and adding Stop does not shift the input. Normal-pointer empty fields and controls are exactly 44px. These are not real-phone keyboard claims. Codex model follow-up (partial, fixture-tested; fresh UI captures pending): model controls now include Codex, using the app-server `model/list` catalog and current `thread/read`/`thread/list` model metadata. Selecting a model saves a per-chat Pipelines registry choice, explicitly labelled **next turn**, and adds the verified `model` override to the next user-initiated `turn/start`, including after a Pipelines restart. This does not call `thread/resume`, start a goal, edit Codex configuration, or claim the remote thread changed before a turn is sent. Versioned protocol evidence and synthetic fixtures: `internal/codex/testdata/`. The following supersedes older C8/B18 implementation details below: - Web attachments are limited to **1 MiB total and eight files**, matching the server. Unsupported harnesses reject attachments before upload storage/queueing. - Uploads live in a private durable `~/.pipelines-chat-attachments` directory (override: `PIPELINES_CHAT_ATTACHMENT_DIR`, must already have a parent and be private). Storage is capped at 64 MiB/512 files; the next upload collects unleased files older than seven days. Queued/in-flight uploads are leased, never collected. Restart retains files, but not in-memory leases or the existing prompt queue. - OpenCode provider-step ends are intermediate progress; foreground completion is read from V2 `/api/session/active`. Failed readback does not advance a queue. - Create replies report `instruction_status` as `applied`, `unsupported`, `failed` or `not_requested`; the UI shows applied/unsupported/failed as chat notices. - The model picker uses the shared modal manager for focus trapping and Escape. Unit/HTTP-fixture regressions cover these changes. No live workflow or deployed service was used; new rendered-state captures are **not yet available**. Audit additions (2026-10-20): **partial** — explicit run Report format selector (`output_format`, HTML/Markdown/no requirement), and session export empty/failed/partial explanations. Unit-tested; browser captures missing under the audit's no-live-runs constraint. Existing captured states below are unchanged. Automatic session export recovery additionally shows pending/running/complete/ exhausted job status and its bounded attempt count in the visit pane. Backend recovery is fixture-tested; this new status line remains **partial**, without a browser capture (no live runs were authorized). ## Transport and connection | # | Feature (Android counterpart) | Protocol | Status | | --- | --- | --- | --- | | T1 | Server-URL entry + validation rules (`data/transport/ServerUrl.kt`) | — | done | | T2 | One reused socket, id matching, request timeout, not-connected guard (`PipelinesClient.kt`) | — | done | | T3 | Connect / disconnect / manual reconnect | `ping` | done — connection now lives only in the Settings Connection section (L2–L4 moved it out of the chrome). `Connect` is connect-only, `Disconnect` closes the shared socket, and `Reconnect` re-opens the validated endpoint immediately, resetting the client's attempt/backoff; all three drive the page's one client. Captured 2026-10-19: `captures/84-shell-settings-connection-1440x900.png` (the connected status card and the Disconnect/Connect pair) and `captures/L1-settings-idle.png` (idle: Connect + Reconnect). The shell's rail/dock dot+word control mirrors the state and is a Settings shortcut (`captures/85-shell-status-indicator-1440x900.png`); the old header URL field, Connect/Disconnect button and state badge are gone | | T4 | Backoff recovery + foreground reconnect after sleep/blur (`ConnectionRecovery.kt`, `ForegroundReconnector.kt`) | — | done — the client's capped backoff is surfaced honestly rather than rewritten: `captures/29-disconnected-banner.png` shows the Settings status card `Disconnected` / `Connection closed (1006)` / `Retrying with backoff…`, while the shell's compact indicator reads `Disconnected`. The attempt number is named only while Connecting (`Connecting (attempt N)…`), matching `ConnectionBanner.kt`; the disconnected card deliberately omits it rather than diverging from Android's wording. The foreground-reconnect **decision** (`foregroundAction`) is unit-tested in `test/connection.test.mjs` (Idle→LeaveAlone; Connecting/Disconnected→Reconnect; Connected→`confirmedAlive ? LeaveAlone : Reconnect`); the **rendered recovery** is captured in `captures/30-reconnect-recovered.png`, where the page (asserted still `Disconnected`) is pointed back at the live server and the browser's own `online` event drives it to a green Connected. The browser uses its `online`/`visibilitychange` event as the liveness signal, so a still-Connected socket is left alone (no in-page positive ping probe is needed; a documented deviation from `ForegroundReconnector`'s positive ping) | | T5 | Connection status + live badge; build/protocol mismatch warning | `pong` build/protocol | done — the full status card lives in Settings (never overlaying content): green `Connected` + `Server build …` + `Server time …` (`captures/84-shell-settings-connection-1440x900.png`), `Disconnected` + the server's message + `Retrying with backoff…` (`captures/L1-settings-disconnected.png`), `Connecting (attempt N)…` (`captures/L1-settings-connecting.png`) and idle `Not connected` (`captures/L1-settings-idle.png`). The shell keeps only a compact dot+word indicator (connected / connecting / disconnected / not connected / not configured) in the rail footer and the narrow bar; it is never colour-only. The protocol-mismatch warning is rendered in the same Settings Connection section (and remains unit-tested via `protocolMismatch`); no live mismatch is expected because the server reports protocol 1. A minimal dot+word live badge still marks the run stream in the runs list/detail (`live` / `reconnecting` / `not live`, per `LiveBadge.kt`/`RunStream.kt`). The build line comes from the `pong` in `SettingsScreen.kt`'s own shape, `Server build {version} ({commit})`; a dirty tree is already marked in `version` by `internal/buildinfo` (`-dirty`), so no duplicate `+dirty` marker is added | | T6 | Artifact URL resolution against the configured server (`resolveArtifactUrl`) | `GET /artifacts/…` | done (URL only) | ## Pipelines | # | Feature | Protocol | Status | | --- | --- | --- | --- | | P1 | Pipeline list | `list_pipelines` | done — row: name; `id {id} · v{version} · {N stages}`; `updated d MMM HH:mm`; last-run line joined from `list_runs` (status chip + `{ago} · {duration} · run {id}`, or `no runs recorded`) | | P2 | Filter + sort (status/name/updated) | client-side | done — Filter menu (name contains, Runs Any/Has runs/No runs, Last run status over the 7 server states, Clear filters) + Sort menu (Server order default, Name A–Z/Z–A, Most recent run, Fewest/Most stages, Reset); server order preserved | | P3 | Pipeline detail: stages, graph, versions, last run, content hash | `get_pipeline` | done — Summary/Graph/Source tabs; Summary shows content hash (monospace), latest version (`v{n}` or blank), created/updated, and a collapsed-by-default Versions section. Graph tab shows the browser guidance note plus the rendered workflow graph (P6) | | P4 | Templates gallery + template detail | `list_templates`, `validate` | done | | P5 | Create pipeline from a template (validate then submit) | `validate`, `submit` | done — a template detail offers the action `Create pipeline from this template`; its dialog's title is `Create from {name}` (a blank name falling back to the template key), it has one required field labelled `New pipeline name (required)`, the note `Creates a new draft pipeline on the server from this template's source; it is not run.`, and Create/Cancel buttons (Cancel reads `Close` after a success). Create is disabled while the name is blank (whitespace counts) or while sending, and the field locks while sending and after a success. Submitting validates the template's source first (`validate`); a validation refusal shows the server's own error in the still-open dialog and sends no `submit`. A passing validate submits exactly `{type:'submit',name:,source: