Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS
JavaScript ai-agentsdashboarddeveloper-toolsdigital-twinquantified-selfsecond-brainself-hostedtailscale
import { useState, useEffect, useCallback, useMemo, useId, useRef } from 'react';
import { Link } from 'react-router';
import {
AlertTriangle, Bot, ChevronDown, ChevronRight, CircleDot, ClipboardCheck,
ExternalLink, MessageSquare, RefreshCw, Rocket, Search, Tag, User
} from 'lucide-react';
import BrailleSpinner from '../../BrailleSpinner';
import Banner from '../../ui/Banner';
import Pill from '../../ui/Pill';
import toast from '../../ui/Toast';
import { DEFAULT_PR_COMPLETION, PR_COMPLETION_OPTIONS, prCompletionOption } from '../../cos/constants';
import ProviderModelSelector from '../../ProviderModelSelector';
import { useThemeContext } from '../../ThemeContext';
import { useCosTaskUpdates } from '../../../hooks/useCosTaskUpdates';
import useRunWithPicker from '../../../hooks/useRunWithPicker';
import useClaimReviewers from '../../../hooks/useClaimReviewers';
import ClaimReviewerSource from '../ClaimReviewerSource';
import ScopeAdherenceCheck from '../ScopeAdherenceCheck';
import ReviewerPicker from '../../cos/ReviewerPicker';
import useReviewerModelOptions from '../../../hooks/useReviewerModelOptions';
import { chipColors } from '../../../lib/chipContrast';
import * as api from '../../../services/api';
import { timeAgo, formatCount } from '../../../utils/formatters';
import { useAsyncAction } from '../../../hooks/useAsyncAction';
import AddToThreadButton from '../../threads/AddToThreadButton';
import RunActionButton from './RunActionButton';
function ClaimReviewOverride({ defaults, overrides, onChange, modelOptions }) {
return (
<ReviewerPicker
{...defaults}
{...overrides}
defaults={defaults}
modelOptions={modelOptions}
showRunFlags={false}
onChange={({ reviewers, usernames, optionalReviewers, reviewerMaxRounds, reviewerModels, reviewerEfforts }) =>
onChange(Object.fromEntries(Object.entries({ reviewers, usernames, optionalReviewers, reviewerMaxRounds, reviewerModels, reviewerEfforts })
.filter(([, value]) => value !== undefined)))}
/>
);
}
const FORGE_LABEL = { github: 'GitHub', gitlab: 'GitLab' };
const RUN_FIELD_LABEL = 'block text-[10px] uppercase tracking-wide text-gray-500';
const RUN_STATUS_RANK = { queuing: 0, queued: 1, active: 2, completed: 3, blocked: 3 };
// Compact, content-width row buttons (right-aligned, wrap on narrow rows); touch
// devices get a taller 36px target instead of full-width 44px slabs.
const ISSUE_ACTION_BUTTON = 'px-2.5 py-1 min-h-7 pointer-coarse:min-h-9 rounded-md text-xs';
const RUN_STATUS_LABEL = {
queued: 'Queued — view',
active: 'Active — view',
completed: 'Completed — view',
blocked: 'Blocked — view'
};
const runStatusForTask = (status) => ({
pending: 'queued',
in_progress: 'active',
completed: 'completed',
blocked: 'blocked'
}[status] || null);
// The two per-issue agent runs this tab can launch. Both POST the same
// `/tasks/slashdo` shape — a slashdo command pinned to one issue via `target` —
// and differ only in which command, which task-metadata key carries the pinned
// issue back over the socket, and how the row reads.
//
// They are tracked as SEPARATE runs (the state map is keyed `<action>:<number>`,
// and the server persists `claimTarget` / `replanTarget` under distinct keys) so
// replanning an issue never disables its Claim button, and neither run's
// lifecycle events can light up the other's.
// Every message this tab prints about a run is "<verb> #42 <happened>", so the
// per-action table carries only the two words that vary — the imperative the
// button promises and the noun the notice names it by — and one formatter
// renders all four sentences from them. A third action supplies two strings.
const runMessages = ({ verb, noun }) => ({
queued: (number) => `Queued a CoS agent to ${verb} #${number}`,
failed: (number) => `Failed to queue a ${verb} for #${number}`,
transition: (number, to) => ({
active: `${noun} #${number} is now active`,
completed: `${noun} #${number} completed`,
blocked: `${noun} #${number} was blocked`
}[to])
});
const ISSUE_ACTIONS = {
replan: {
command: 'replan',
targetKey: 'replanTarget',
label: 'Replan',
icon: ClipboardCheck,
tone: 'bg-port-border/60 text-gray-300 enabled:hover:bg-port-border',
title: (number) => `Queue a CoS agent to review the plan on issue #${number} and comment its refinements — it writes no code`,
...runMessages({ verb: 'replan', noun: 'Replan of' })
},
claim: {
command: 'next',
targetKey: 'claimTarget',
label: 'Claim',
icon: Rocket,
tone: 'bg-port-accent/20 text-port-accent enabled:hover:bg-port-accent/30',
title: (number, appName) => `Queue a CoS agent to claim issue #${number} for ${appName}`,
...runMessages({ verb: 'claim', noun: 'Claim' })
}
};
// Buttons render in table order — no second list to keep in sync.
const ACTION_ORDER = Object.keys(ISSUE_ACTIONS);
/** Run-map key. One issue can carry an independent run per action. */
const runKey = (action, issueNumber) => `${action}:${issueNumber}`;
/** Split a run key back into its action + issue number (as a string). */
function parseRunKey(key) {
const at = key.indexOf(':');
return { action: key.slice(0, at), issueNumber: key.slice(at + 1) };
}
// `in-progress` is the forge label a `/do:next` claim stamps on an issue it is
// actively working (server/services/issueReconcile.js#IN_PROGRESS_LABEL), and
// `blocked` marks work that should not be picked up by default. `needs-input`
// and `needs-decision` park issues awaiting human choices. Those rows aren't
// useful claim candidates, so the tab hides them until the user toggles the
// chip back on.
const DEFAULT_HIDDEN_LABELS = ['blocked', 'in-progress', 'needs-decision', 'needs-input'];
// Row background tint for the two claim-state labels. Uses the theme's
// warning/error tokens (alpha-tinted so every theme keeps its own palette
// rather than a hardcoded color), blocked taking precedence over in-progress.
export function issueRowTint(issue) {
const names = new Set(issue.labels.map(l => l.name));
if (names.has('blocked')) return 'bg-port-error/10';
if (names.has('in-progress')) return 'bg-port-warning/10';
return 'bg-port-card';
}
const defaultLabelFilter = () => ({
mode: 'exclude',
names: new Set(DEFAULT_HIDDEN_LABELS),
});
/** Exclude: drop any issue that carries a named label. Include: keep only issues that carry at least one named label. */
function issuePassesLabelFilter(issue, { mode, names }) {
const labels = issue.labels || [];
if (mode === 'include') return labels.some(l => names.has(l.name));
return names.size === 0 || !labels.some(l => names.has(l.name));
}
// Why the forge returned nothing, in the user's terms. Sentinel-aware: a
// definitive "no open issues" and a failed probe are different sentences, so an
// unreachable CLI can never read as an empty tracker.
const EMPTY_REASONS = {
'no-repo-path': 'This app has no repo path configured.',
'unsupported-forge': 'This app\'s git origin isn\'t GitHub or GitLab, so there are no forge issues to list.',
'tracker-not-a-forge': 'This app\'s Work Tracker isn\'t a forge issue tracker, so a claim here wouldn\'t touch forge issues. Change it under Edit App → Workflow.',
'tracker-forge-mismatch': 'This app\'s Work Tracker is pinned to a different forge than its git origin, so neither list would match what a claim runs against. Reconcile them under Edit App → Workflow.',
'no-open-issues': 'No open issues on this tracker.'
};
/**
* Label chips carry the forge's own color, which is arbitrary per-repo data —
* so it can't be a Tailwind class (those must be literal strings in source).
* Render it as a tinted chip via inline style: a low-alpha wash of the label's
* own color for the background, and `chipContrast` for the text/border, which
* keeps the hue but moves its lightness until it clears WCAG AA on the ACTIVE
* theme mode. Verbatim label color only ever worked on night themes — GitHub's
* pale defaults (`plan` #fef2c0, `effort:*` #c5def5) rendered as ~1.1:1 on the
* day themes, i.e. invisible.
*
* A label with no color — or a color we can't parse — gets no inline style, so
* the chip keeps its own neutral look (the `muted` Pill tone on a row chip).
*/
function LabelChip({ label }) {
const { theme } = useThemeContext();
const style = chipColors(label.color, theme?.mode);
return (
<Pill
size="xs"
tone={style ? 'bare' : 'muted'}
title={label.description || undefined}
style={style}
>
{label.name}
</Pill>
);
}
/**
* One label's show/hide toggle. Pressed (default) = issues carrying the label
* are listed; un-pressed strikes the chip through and drops every issue that
* carries it — so an issue tagged `bug` + `in-progress` disappears while
* `in-progress` is off, which is what "hide in-progress work" has to mean.
*
* The count renders as a node adjacent to the name, so the computed accessible
* name would be "bug1" — `aria-label` spells it out instead.
*
* NOT the shared `ui/ToggleChip`: that one is a fixed-accent pill wrapping a
* real checkbox, which can't carry the forge's per-label color (the thing the
* user recognizes a label by) and stacks a checkbox per chip into a row that
* routinely runs 20+ labels wide.
*/
function LabelFilterChip({ facet, hidden, onToggle }) {
const { theme } = useThemeContext();
const style = hidden ? null : chipColors(facet.color, theme?.mode);
// The theme utilities carry `!important` (index.css remaps `.bg-port-bg`,
// `.border-port-border`, and day-mode `.text-gray-300`), and author
// `!important` beats an inline declaration — so a chip that ships BOTH renders
// in theme neutrals with its graded color silently dead. Emit the neutrals
// only when there's no graded style to paint, the way `Pill tone="bare"` does.
const neutral = hidden
? 'border-port-border bg-port-bg text-gray-500 line-through opacity-60 hover:opacity-100'
: 'border-port-border bg-port-bg text-gray-300 hover:opacity-80';
return (
<button
type="button"
onClick={() => onToggle(facet.name)}
aria-pressed={!hidden}
aria-label={`${facet.name} (${facet.count})`}
title={facet.description
? `${facet.description} — click to ${hidden ? 'show' : 'hide'} these issues`
: `Click to ${hidden ? 'show' : 'hide'} issues labeled ${facet.name}`}
className={`inline-flex items-center gap-1 rounded border px-1.5 py-0.5 text-[10px] transition-opacity ${
style ? 'hover:opacity-80' : neutral
}`}
style={style}
>
{facet.name}
<span className="font-mono opacity-70">{facet.count}</span>
</button>
);
}
/**
* Open issues from the app's forge (GitHub or GitLab, resolved from the git
* `origin` remote), each with its labels, assignees, expandable description, and
* a one-click "Claim with CoS agent" button.
*
* Claiming queues the SAME `/do:next` task the Agent Operations panel does,
* pinned to this issue via `target` — so the run honors the app's configured
* Work Tracker, worktree, and PR settings instead of a parallel code path.
*
* Replan is the same mechanism pointed at `replan`: a SECOND model re-derives
* the plan on one already-planned issue (and, for an epic, its children) and
* comments its refinements. It writes no code and claims nothing, so it is
* offered alongside Claim rather than instead of it.
*/
export default function IssuesTab({ appId, appName }) {
const [trackAssigned, trackingAssigned] = useAsyncAction(async () => {
const result = await api.syncGithubThreads({ appId }, { silent: true });
const summary = `${appName}: ${formatCount(result.created)} threads added, ${formatCount(result.updated)} updated`;
if (result.possiblyTruncated) toast(`${summary}. More assigned issues may exist beyond this batch.`, { icon: '⚠️' });
else toast.success(summary);
});
const searchId = useId();
const filedById = useId();
const overrideContextId = useId();
const prCompletionId = useId();
const [data, setData] = useState(null);
const [loading, setLoading] = useState(true);
const [error, setError] = useState('');
const [query, setQuery] = useState('');
const [unassignedOnly, setUnassignedOnly] = useState(false);
const [filedBy, setFiledBy] = useState('');
// Exclude mode tracks labels to hide (a later refresh's new label stays
// visible). Include mode — entered by Hide all labels — tracks labels to
// keep, so turning `critical` back on lists every issue that carries it,
// even when the issue also has labels that are still off.
const [labelFilter, setLabelFilter] = useState(defaultLabelFilter);
const [expanded, setExpanded] = useState(() => new Set());
// Per-issue run lifecycle: 'queuing' while the POST is in flight, then the
// task's live CoS state. Keyed `<action>:<issue number>` so neither one row's
// run nor its sibling action can disable the others' buttons.
const [runs, setRuns] = useState({});
const runsRef = useRef({});
// Generation guard: a forge list can take seconds, so a Refresh (or a switch to
// a different app, which updates this component in place rather than
// remounting it) can leave an older request in flight. Without this, that older
// response lands last and shows one app's issues under another's Claim buttons.
const requestRef = useRef(0);
// Page-level provider/model/effort pin for every Claim AND Replan button on this tab —
// left untouched (blank), a claim resolves the install's active provider,
// same as the bare button always did (POST /tasks/slashdo -> resolveAgentProviderAndModel;
// this manual path does NOT consult the app's scheduled claim-work override for
// the PROVIDER — that pin is read only by the automated claim-work task).
// Scoped to the provider deliberately: the REVIEWERS below do come from that
// override, which is precisely the mismatch `claimReviewers` exists to surface.
// This picker never persists across a reload; it's a session convenience for
// "claim the next several issues with model X" without reopening the Agent
// Operations drawer each time.
const picker = useRunWithPicker();
const [overrideContext, setOverrideContext] = useState('');
// The reviewers a Claim launched from this tab will actually run — NOT the
// Models → Code Reviewers list, whenever a claim-work override is in play (see
// `GET /apps/:id/claim-reviewers`). Untouched fields remain inherited.
const claimReviewers = useClaimReviewers(appId);
const [prCompletion, setPrCompletion] = useState(DEFAULT_PR_COMPLETION);
const [prCompletionTouched, setPrCompletionTouched] = useState(false);
useEffect(() => {
if (!prCompletionTouched && claimReviewers?.prCompletion) {
setPrCompletion(claimReviewers.prCompletion);
}
}, [claimReviewers?.prCompletion, prCompletionTouched]);
const [reviewOverrides, setReviewOverrides] = useState({});
const reviewerModelOptions = useReviewerModelOptions();
// A claim needs someone it can wait on: a non-Copilot AI reviewer or a
// GitHub/GitLab user. Copilot alone has no CLI for the claim agent to run.
const invalidReviewOverride = prCompletion === 'review-then-merge'
&& Array.isArray(reviewOverrides.reviewers)
&& !reviewOverrides.reviewers.some(reviewer => reviewer !== 'copilot')
&& !(reviewOverrides.usernames ?? claimReviewers?.usernames ?? []).length;
const [showReviewOverride, setShowReviewOverride] = useState(false);
const [showSettings, setShowSettings] = useState(false);
const runSettingsId = useId();
// An invalid or active review override must never hide behind the collapsed summary.
// Overrides only render (and apply) under review-then-merge, so only they can lock the panel open.
const settingsForcedOpen = prCompletion === 'review-then-merge'
&& (invalidReviewOverride || Object.keys(reviewOverrides).length > 0);
const settingsOpen = showSettings || settingsForcedOpen;
useEffect(() => {
setReviewOverrides({});
setShowReviewOverride(false);
setPrCompletion(DEFAULT_PR_COMPLETION);
setPrCompletionTouched(false);
}, [appId]);
// Keep the event-driven path based on the latest runs without putting a
// mutable state snapshot in its effect dependencies. Socket callbacks can
// arrive between the POST response and the next React render.
const replaceRuns = useCallback((updater) => {
const next = updater(runsRef.current);
runsRef.current = next;
setRuns(next);
}, []);
useEffect(() => {
runsRef.current = {};
setRuns({});
setOverrideContext('');
setLabelFilter(defaultLabelFilter());
setUnassignedOnly(false);
setFiledBy('');
}, [appId]);
const applyTaskUpdate = useCallback((task) => {
if (!task?.id) return;
const nextStatus = runStatusForTask(task.status);
if (!nextStatus) return;
const currentRuns = runsRef.current;
const nextRuns = { ...currentRuns };
const transitions = [];
for (const [key, rawRun] of Object.entries(currentRuns)) {
const { action, issueNumber } = parseRunKey(key);
const spec = ISSUE_ACTIONS[action];
if (!spec) continue;
const run = typeof rawRun === 'string' ? { status: rawRun } : rawRun;
// Before the POST resolves there is no task id to match on, so fall back
// to the durable per-action target the server stamped. Reading THIS
// action's key is what keeps a claim's events off a replan's row.
const matchesTask = run.taskId === task.id || (
!run.taskId && task.metadata?.app === appId &&
String(task.metadata?.[spec.targetKey]) === issueNumber
);
if (!matchesTask) continue;
const currentStatus = run.status || 'queuing';
if ((RUN_STATUS_RANK[nextStatus] ?? 0) < (RUN_STATUS_RANK[currentStatus] ?? 0)) continue;
if (run.taskId === task.id && currentStatus === nextStatus) continue;
nextRuns[key] = { ...run, taskId: run.taskId || task.id, status: nextStatus };
transitions.push({ action, issueNumber, from: currentStatus, to: nextStatus });
}
if (transitions.length === 0) return;
replaceRuns(() => nextRuns);
for (const { action, issueNumber, from, to } of transitions) {
if (to === from) continue;
const message = ISSUE_ACTIONS[action].transition(issueNumber, to);
if (!message) continue;
if (to === 'active') toast(message, { icon: '▶️' });
else if (to === 'completed') toast.success(message);
else if (to === 'blocked') toast.error(message);
}
}, [appId, replaceRuns]);
useCosTaskUpdates(applyTaskUpdate);
const load = useCallback(async () => {
const generation = requestRef.current + 1;
requestRef.current = generation;
const isCurrent = () => requestRef.current === generation;
setLoading(true);
setError('');
const result = await api.getAppIssues(appId).catch(err => {
if (isCurrent()) setError(err?.message || 'Failed to load issues');
return null;
});
if (!isCurrent()) return;
setLoading(false);
// A failed refresh keeps the last good list rather than blanking it.
if (result) setData(result);
}, [appId]);
useEffect(() => { load(); }, [load]);
// Lowercase each issue's searchable text ONCE per fetch, not once per
// keystroke: bodies are capped at 8000 chars and the list at 200 issues, so
// re-deriving inside the filter churned up to 1.6 MB of throwaway strings on
// every character typed. Digits are case-invariant, so the issue number folds
// into the same haystack.
const haystacks = useMemo(() => (data?.issues || []).map(issue => ({
issue,
hay: [
String(issue.number),
issue.title,
issue.body || '',
issue.labels.map(l => l.name).join(' '),
issue.assignees.join(' ')
].join('\n').toLowerCase()
})), [data]);
// Every label present on the fetched issues, with how many issues carry it,
// alphabetical so the chip row doesn't reshuffle between refreshes.
const labelFacets = useMemo(() => {
const byName = new Map();
for (const issue of data?.issues || []) {
for (const label of issue.labels || []) {
const seen = byName.get(label.name);
if (seen) seen.count += 1;
else byName.set(label.name, { ...label, count: 1 });
}
}
return [...byName.values()].sort((a, b) => a.name.localeCompare(b.name));
}, [data]);
const authorOptions = useMemo(() => {
const authors = new Set();
for (const issue of data?.issues || []) {
if (issue.author) authors.add(issue.author);
}
return [...authors].sort((a, b) => a.localeCompare(b));
}, [data]);
const issues = useMemo(() => {
const q = query.trim().toLowerCase();
const rows = haystacks.filter(h =>
issuePassesLabelFilter(h.issue, labelFilter)
&& (!unassignedOnly || h.issue.assignees.length === 0)
&& (!filedBy || h.issue.author === filedBy)
&& (!q || h.hay.includes(q))
);
return rows.map(h => h.issue);
}, [haystacks, query, labelFilter, unassignedOnly, filedBy]);
const total = data?.issues?.length ?? 0;
// Only labels actually on this tracker count as "filtering something" — the
// seeded defaults must not claim to hide anything in a repo that never uses
// them. Include mode is always a filter, even with an empty keep-set.
const hidingByLabel = labelFilter.mode === 'include'
|| labelFacets.some(f => labelFilter.names.has(f.name));
const canHideAll = labelFilter.mode !== 'include' || labelFilter.names.size > 0;
const toggleLabel = (name) => setLabelFilter(prev => {
const names = new Set(prev.names);
if (names.has(name)) names.delete(name); else names.add(name);
return { mode: prev.mode, names };
});
const hideAllLabels = () => setLabelFilter({ mode: 'include', names: new Set() });
const showAllLabels = () => setLabelFilter({ mode: 'exclude', names: new Set() });
const labelChipHidden = (name) => (labelFilter.mode === 'include'
? !labelFilter.names.has(name)
: labelFilter.names.has(name));
const toggleExpanded = (number) => setExpanded(prev => {
const next = new Set(prev);
if (next.has(number)) next.delete(number); else next.add(number);
return next;
});
// One launcher for both per-issue actions: they POST the same body to the same
// endpoint and differ only in the slashdo command. Keeping them on one path is
// what guarantees the provider pin, override context, and prefetched issue
// content reach a replan exactly as they reach a claim.
const handleRun = async (issue, action) => {
const spec = ISSUE_ACTIONS[action];
if (action === 'claim' && invalidReviewOverride) return;
const key = runKey(action, issue.number);
replaceRuns(prev => ({ ...prev, [key]: { status: 'queuing', taskId: null } }));
const trimmedOverrideContext = overrideContext.trim();
const result = await api.createSlashdoTask(spec.command, appId, {
target: String(issue.number),
issueContext: {
number: issue.number,
title: issue.title || '',
body: issue.body || '',
...(issue.url ? { url: issue.url } : {})
},
...picker.pin,
...(action === 'claim' ? {
...(prCompletionTouched ? { prCompletion } : {}),
...(prCompletion === 'review-then-merge' ? reviewOverrides : {}),
} : {}),
...(trimmedOverrideContext ? { overrideContext: trimmedOverrideContext } : {}),
}, { silent: true })
.catch(err => {
toast.error(err?.message || spec.failed(issue.number));
return null;
});
if (!result) {
replaceRuns(prev => {
const next = { ...prev };
delete next[key];
return next;
});
return;
}
replaceRuns(prev => {
const current = prev[key];
const currentStatus = typeof current === 'string' ? current : current?.status;
const resultStatus = runStatusForTask(result.status) || 'queued';
const status = (RUN_STATUS_RANK[resultStatus] ?? 0) >= (RUN_STATUS_RANK[currentStatus] ?? 0)
? resultStatus
: currentStatus;
return {
...prev,
[key]: {
...(typeof current === 'object' ? current : {}),
status,
taskId: current?.taskId || result.id || null
}
};
});
toast.success(spec.queued(issue.number));
};
if (loading && !data) return <BrailleSpinner text="Loading issues" />;
const forgeLabel = FORGE_LABEL[data?.forge] || 'Issues';
// A transient failure keeps the "couldn't ask" framing — never the lie that
// the tracker is empty. The SENTENCE comes from the server, beside the
// classifier that produced the reason: the reason vocabulary is open-ended
// (`gh-${status}`), and only the classifier knows whether the forge was
// unreachable or merely answered in a dialect we couldn't parse. A client-side
// reason→copy table can only shadow one entry of that, which is how an
// authenticated user got told to authenticate.
const unavailable = data?.transient === true;
const reviewerCsv = !claimReviewers || prCompletion !== 'review-then-merge' || invalidReviewOverride ? null
: Object.keys(reviewOverrides).length
? [
...(reviewOverrides.reviewers ?? claimReviewers.reviewers),
...(reviewOverrides.usernames ?? claimReviewers.usernames ?? []).map(username => `@${username}`),
].join(', ')
: claimReviewers.csv;
const pickedProvider = picker.providers?.find(p => p.id === picker.selectedProviderId);
const runSettingsSummary = [
pickedProvider?.name || picker.selectedProviderId || 'Auto',
picker.selectedModel || pickedProvider?.defaultModel || null,
picker.effort || pickedProvider?.effort || null,
prCompletionOption(prCompletion)?.label ?? prCompletion,
reviewerCsv,
overrideContext.trim() ? '+ context' : null,
].filter(Boolean).join(' · ');
return (
<div className="space-y-4">
<div className="flex flex-wrap items-center gap-3">
<h3 className="text-lg font-semibold text-white">
{forgeLabel} Issues
{data?.fullName && <span className="ml-2 text-xs font-mono text-gray-500">{data.fullName}</span>}
</h3>
{total > 0 && (
<span className="text-xs text-gray-500">
{issues.length === total ? `${total} open` : `${issues.length} of ${total} open`}
</span>
)}
<button
type="button"
onClick={() => setUnassignedOnly(prev => !prev)}
aria-pressed={unassignedOnly}
title={unassignedOnly ? 'Show assigned issues too' : 'Show only issues without assignees'}
className={`px-3 py-1.5 rounded-lg text-xs border flex items-center gap-1.5 transition-colors ${unassignedOnly
? 'bg-port-accent/20 text-port-accent border-port-accent/40'
: 'bg-port-bg text-gray-400 border-port-border hover:text-white'
}`}
>
<User size={13} /> Unassigned only
</button>
<div className="flex items-center gap-1.5">
<label htmlFor={filedById} className="text-xs text-gray-400 shrink-0">
Filed by
</label>
<select
id={filedById}
aria-label="Filed by"
value={filedBy}
onChange={e => setFiledBy(e.target.value)}
className="px-2.5 py-1.5 bg-port-bg border border-port-border rounded-lg text-xs text-white focus:border-port-accent focus:outline-hidden"
>
<option value="">All</option>
{authorOptions.map(author => (
<option key={author} value={author}>{author}</option>
))}
</select>
</div>
<div className="relative ml-auto">
<label htmlFor={searchId} className="sr-only">Filter issues</label>
<Search size={14} className="absolute left-2.5 top-1/2 -translate-y-1/2 text-gray-500" />
<input
id={searchId}
type="search"
value={query}
onChange={e => setQuery(e.target.value)}
placeholder="Filter by title, label, assignee…"
className="w-full sm:w-72 pl-8 pr-3 py-1.5 bg-port-bg border border-port-border rounded-lg text-white text-sm focus:border-port-accent focus:outline-hidden"
/>
</div>
<button
onClick={load}
disabled={loading}
className="px-3 py-1.5 bg-port-border hover:bg-port-border/80 text-white rounded-lg text-xs flex items-center gap-1 disabled:opacity-50"
>
<RefreshCw size={14} className={loading ? 'animate-spin' : ''} /> Refresh
</button>
</div>
{error && (
<Banner tone="error" size="md" icon={AlertTriangle}>
Couldn't load issues — {error}
</Banner>
)}
{!error && unavailable && (
<Banner tone="warning" size="md" icon={AlertTriangle}>
{data.headline || `Couldn't reach ${forgeLabel}`} ({data.reason})
{data.remedy ? ` — ${data.remedy}.` : ''}
</Banner>
)}
{data?.forge === 'github' && (
<div className="flex flex-wrap items-center gap-3 text-xs">
<button
type="button"
onClick={trackAssigned}
disabled={trackingAssigned || loading}
className="px-3 py-1.5 rounded-lg border border-port-border text-port-accent hover:bg-port-accent/10 disabled:opacity-50"
>
{trackingAssigned ? 'Tracking assigned issues…' : 'Track my assigned issues in Brain'}
</button>
<Link to="/brain/threads" className="text-port-accent hover:underline">Open Threads</Link>
</div>
)}
{labelFacets.length > 0 && (
<div
role="group"
aria-label="Label filters"
className="flex flex-wrap items-center gap-1.5 px-3 py-2 bg-port-card border border-port-border rounded-lg"
>
<span className="flex items-center gap-1.5 text-xs text-gray-500 uppercase tracking-wide shrink-0">
<Tag size={14} /> Labels
</span>
{labelFacets.map(facet => (
<LabelFilterChip
key={facet.name}
facet={facet}
hidden={labelChipHidden(facet.name)}
onToggle={toggleLabel}
/>
))}
{(canHideAll || hidingByLabel) && (
<div className="ml-auto flex items-center gap-3">
{canHideAll && (
<button
type="button"
onClick={hideAllLabels}
className="text-xs text-gray-500 hover:text-port-accent transition-colors"
>
Hide all labels
</button>
)}
{hidingByLabel && (
<button
type="button"
onClick={showAllLabels}
className="text-xs text-gray-500 hover:text-port-accent transition-colors"
>
Show all labels
</button>
)}
</div>
)}
</div>
)}
<div className="px-3 py-2 bg-port-card border border-port-border rounded-lg space-y-3">
<div className="flex flex-wrap items-center gap-x-3 gap-y-1">
<Bot size={14} className="text-gray-500 shrink-0" />
<span data-testid="run-settings-summary" className="flex-1 min-w-0 text-xs text-gray-300 break-words">{runSettingsSummary}</span>
{!settingsForcedOpen && (
<button
type="button"
aria-expanded={settingsOpen}
aria-controls={runSettingsId}
onClick={() => setShowSettings(value => !value)}
className="text-xs text-port-accent hover:underline shrink-0"
>
{settingsOpen ? 'Hide' : 'Change'}
</button>
)}
</div>
{settingsOpen && (
<div id={runSettingsId} className="space-y-3">
<div className="space-y-1">
<span className={RUN_FIELD_LABEL}>Agent</span>
<ProviderModelSelector
{...picker.selectorProps}
compact
/>
</div>
<div className="space-y-1">
<label htmlFor={prCompletionId} className={RUN_FIELD_LABEL}>After PR</label>
<select
id={prCompletionId}
value={prCompletion}
title={prCompletionOption(prCompletion)?.description}
onChange={e => { setPrCompletion(e.target.value); setPrCompletionTouched(true); }}
className="w-full sm:w-auto min-h-9 px-2.5 bg-port-bg border border-port-border rounded-lg text-xs text-white focus:border-port-accent focus:outline-hidden"
>
{PR_COMPLETION_OPTIONS.map(opt => (
<option key={opt.value} value={opt.value}>{opt.label}</option>
))}
</select>
</div>
{claimReviewers && prCompletion === 'review-then-merge' && (
<div className="space-y-1">
<div className="flex items-center gap-2">
<span className={RUN_FIELD_LABEL}>Reviewers</span>
<span className="flex-1" />
{Object.keys(reviewOverrides).length > 0 && (
<button type="button" aria-label="Use configured reviewers" onClick={() => setReviewOverrides({})} className="text-xs text-port-accent hover:underline min-h-7">
Reset
</button>
)}
<button
type="button"
aria-expanded={showReviewOverride}
aria-label={showReviewOverride ? 'Done editing reviewers' : 'Edit reviewers'}
onClick={() => setShowReviewOverride(value => !value)}
className="text-xs text-port-accent hover:underline min-h-7"
>
{showReviewOverride ? 'Done' : 'Edit'}
</button>
</div>
{!showReviewOverride && (
<p className="text-xs text-gray-300 break-words">
{reviewerCsv ? <span>{reviewerCsv}</span> : <span className="text-gray-500">None</span>}
{!Object.keys(reviewOverrides).length && <ClaimReviewerSource source={claimReviewers.source} />}
</p>
)}
{invalidReviewOverride && (
<p className="text-xs text-port-warning" role="alert">Add an AI reviewer or a GitHub/GitLab user.</p>
)}
{showReviewOverride && (
<ClaimReviewOverride defaults={claimReviewers} overrides={reviewOverrides} onChange={setReviewOverrides} modelOptions={reviewerModelOptions} />
)}
</div>
)}
<div className="space-y-1">
<label htmlFor={overrideContextId} className={RUN_FIELD_LABEL}>Extra context</label>
<textarea
id={overrideContextId}
value={overrideContext}
onChange={e => setOverrideContext(e.target.value)}
maxLength={4000}
rows={2}
placeholder="Optional guidance for Claim and Replan"
className="w-full px-3 py-2 bg-port-bg border border-port-border rounded-lg text-white text-sm placeholder:text-gray-600 focus:border-port-accent focus:outline-hidden resize-y"
/>
</div>
</div>
)}
</div>
{!error && !unavailable && total === 0 && (
<div className="px-3 py-2 text-sm text-gray-500 bg-port-card border border-port-border rounded-lg">
{EMPTY_REASONS[data?.reason] ?? EMPTY_REASONS['no-open-issues']}
</div>
)}
{/* A failed REFRESH deliberately shows the error banner above AND the last
good list below — blanking issues the user was reading is worse than
showing them alongside a "couldn't refresh" notice. */}
{issues.length === 0 && total > 0 && (
<div className="px-3 py-2 text-sm text-gray-500 bg-port-card border border-port-border rounded-lg">
{query.trim()
? <>No open issues match “{query}”{hidingByLabel ? ' with the current label filters' : ''}.</>
: unassignedOnly && filedBy
? <>No unassigned open issues filed by {filedBy}{hidingByLabel ? ' match the current label filters' : ''}.</>
: unassignedOnly
? <>No unassigned open issues{hidingByLabel ? ' match the current label filters' : ''}.</>
: filedBy
? <>No open issues filed by {filedBy}{hidingByLabel ? ' match the current label filters' : ''}.</>
: 'No open issues match the current label filters.'}
</div>
)}
{issues.length > 0 && (
<div className="border border-port-border rounded-lg divide-y divide-port-border overflow-hidden">
{issues.map(issue => {
const isOpen = expanded.has(issue.number);
return (
<div key={issue.number} className={`${issueRowTint(issue)} @container`}>
<div className="flex flex-col @min-[40rem]:flex-row @min-[40rem]:items-start gap-2 @min-[40rem]:gap-3 p-3">
<button
onClick={() => toggleExpanded(issue.number)}
aria-expanded={isOpen}
aria-label={`${isOpen ? 'Collapse' : 'Expand'} description for issue ${issue.number}`}
className="flex text-gray-500 hover:text-white transition-colors mt-0.5 shrink-0"
>
{isOpen ? <ChevronDown size={16} /> : <ChevronRight size={16} />}
</button>
<div className="flex-1 min-w-0">
<div className="flex items-start gap-2">
<CircleDot size={14} className="text-port-success shrink-0 mt-0.5" />
<span className="text-xs font-mono text-gray-500 shrink-0 mt-0.5">#{issue.number}</span>
<button
onClick={() => toggleExpanded(issue.number)}
aria-expanded={isOpen}
className="text-sm font-medium text-white text-left hover:text-port-accent transition-colors break-words flex-1 min-w-0"
>
{issue.title || '(no title)'}
</button>
</div>
{issue.labels.length > 0 && (
<div className="flex flex-wrap gap-1 mt-1.5">
{issue.labels.map(l => <LabelChip key={l.name} label={l} />)}
</div>
)}
<div className="flex flex-wrap items-center gap-x-3 gap-y-1 mt-1.5 text-xs text-gray-500">
{issue.author && <span>opened by {issue.author}</span>}
{issue.updatedAt && <span>updated {timeAgo(issue.updatedAt)}</span>}
{issue.milestone && <span className="text-cyan-400">{issue.milestone}</span>}
{issue.commentCount > 0 && (
<span className="flex items-center gap-1">
<MessageSquare size={12} /> {issue.commentCount} comment{issue.commentCount === 1 ? '' : 's'}
</span>
)}
{issue.assignees.length > 0 ? (
<span className="flex items-center gap-1 text-gray-300">
<User size={12} /> {issue.assignees.join(', ')}
</span>
) : (
<span className="italic">unassigned</span>
)}
</div>
</div>
<div className="flex flex-wrap items-center justify-end gap-1.5 shrink-0 @min-[40rem]:max-w-[45%]">
{ACTION_ORDER.map(action => {
const spec = ISSUE_ACTIONS[action];
const run = runs[runKey(action, issue.number)];
const state = typeof run === 'string' ? run : run?.status;
const Icon = spec.icon;
// A launched run swaps its button for a link to the agent
// queue: the run is no longer this row's to start, and the
// one thing the user wants next is to watch it.
if (state && state !== 'queuing') {
return (
<Link
key={action}
to="/cos/agents"
aria-label={`${spec.label} #${issue.number}: ${RUN_STATUS_LABEL[state] || 'Queued — view'}`}
className={`${ISSUE_ACTION_BUTTON} bg-port-success/20 text-port-success hover:bg-port-success/30 border border-port-border flex items-center gap-1.5 whitespace-nowrap transition-colors`}
>
<Icon size={14} /> {spec.label} · {RUN_STATUS_LABEL[state] || 'Queued — view'}
</Link>
);
}
return (
<RunActionButton
key={action}
onClick={() => handleRun(issue, action)}
busy={state === 'queuing'}
disabled={action === 'claim' && invalidReviewOverride}
title={spec.title(issue.number, appName)}
icon={Icon}
className={`${ISSUE_ACTION_BUTTON} ${spec.tone} border border-port-border`}
>
{spec.label}
</RunActionButton>
);
})}
{(issue.url || issue.html_url) && (
<AddToThreadButton
refItem={{
kind: data?.forge === 'gitlab' ? 'gitlab.issue' : 'github.issue',
id: issue.url || issue.html_url,
label: `#${issue.number} ${issue.title || ''}`.trim()
}}
buttonText="Thread"
className={`${ISSUE_ACTION_BUTTON} bg-port-bg text-gray-300 hover:text-white border border-port-border flex items-center gap-1.5 transition-colors`}
/>
)}
{issue.url && (
<a
href={issue.url}
target="_blank"
rel="noreferrer"
role="button"
aria-label={`Open issue ${issue.number} on ${forgeLabel}`}
title={`Open issue ${issue.number} on ${forgeLabel}`}
onKeyDown={(e) => {
if (e.key === ' ' || e.key === 'Spacebar') {
e.preventDefault();
window.open(issue.url, '_blank', 'noreferrer');
}
}}
className={`${ISSUE_ACTION_BUTTON} inline-flex items-center gap-1.5 border border-port-border bg-port-bg text-gray-300 hover:text-white hover:border-port-accent/40 hover:bg-port-border/40 transition-colors`}
>
<ExternalLink size={14} />
<span>{forgeLabel}</span>
</a>
)}
</div>
</div>
{isOpen && (
<div className="px-3 pb-3 sm:pl-10">
{issue.body?.trim() ? (
<pre className="text-xs text-gray-300 whitespace-pre-wrap break-words font-sans bg-port-bg border border-port-border rounded-lg p-3 max-h-96 overflow-auto">
{issue.body}
</pre>
) : (
<p className="text-xs text-gray-500 italic">No description.</p>
)}
{/* Advisory only, and only under an expanded row: the
reader has the description in front of them, so the
clause the scorer cites is something they can check. */}
<ScopeAdherenceCheck
appId={appId}
kind="issue"
title={issue.title || ''}
body={issue.body || ''}
/>
</div>
)}
</div>
);
})}
</div>
)}
</div>
);
}