Merge branch 'main' into feat/streaming-chat

Reviewed-on: #48

apps/desktop/e2e/live-backlog.spec.ts

@@ -41,6 +41,18 @@ test.describe('live backlog', () => {
     await expect(
       win.getByText(/cold-start priors · \d+\/20 closed issues estimated|calibrated on \d+ closed/),
     ).toBeVisible()
+    // Real per-milestone forecasts — these milestone names come from gitea, not the
+    // fixture (which lists Beta / Pilot-ready / v1.0).
+    await expect(win.getByText(/P2 — Scheduler/)).toBeVisible()
+    await win.screenshot({ path: join(here, '.artifacts', 'screens', 'live-runway.png'), fullPage: true, animations: 'disabled' })
+
+    // Milestone drill-in — clicking a real milestone opens its real detail
+    await win.getByText(/P2 — Scheduler/).click()
+    await expect(win.getByRole('heading', { name: 'P2 — Scheduler + Monte Carlo' })).toBeVisible()
+    await expect(win.getByText(/\d+ issues · est \d+d/)).toBeVisible()
+    await win.screenshot({ path: join(here, '.artifacts', 'screens', 'live-milestone.png'), fullPage: true, animations: 'disabled' })
+    await rail.getByRole('button', { name: 'Runway' }).click()
+
     await win.getByRole('button', { name: 'Full report' }).click()
     await expect(win.getByText(/cold-start · \d+\/20|curve active · n ≥ 20/)).toBeVisible()
     await win.screenshot({ path: join(here, '.artifacts', 'screens', 'live-calibration.png'), fullPage: true, animations: 'disabled' })

apps/desktop/src/main/model.ts

@@ -86,11 +86,15 @@ export function registerModelIpc(): void {
     return { configured: true, model }
   })
 
-  ipcMain.handle('model:chat', async (_event, messages: ChatMessage[]) => {
+  ipcMain.handle('model:chat', async (event, messages: ChatMessage[]) => {
     if (!router) return { ok: false as const, reason: 'unconfigured' as const }
     const client = getGiteaClient()
     const model = await resolveLoadedModel(router.small.baseUrl, router.small.model)
     const chat = createChatClient({ ...router.small, model }, fetch)
+    // stream the model's prose to the renderer token-by-token
+    const onToken = (delta: string) => {
+      if (!event.sender.isDestroyed()) event.sender.send('model:chat:token', delta)
+    }
 
     // Proposals the model formulates this turn; the renderer approves them (the
     // write happens through gitea:applyChange, never inside the loop).
@@ -129,10 +133,11 @@ export function registerModelIpc(): void {
 
     try {
       const turn = await runAgentTurn({
-        complete: (m, t) => chat.complete(m, t),
+        complete: (m, t, ot) => chat.complete(m, t, ot),
         messages: [{ role: 'system', content: REGINALD_SYSTEM }, ...messages],
         tools: REGINALD_TOOLS,
         execute,
+        onToken,
       })
       return { ok: true as const, content: turn.content, steps: turn.steps, proposals }
     } catch (e) {

apps/desktop/src/preload/index.ts

@@ -25,6 +25,12 @@ const api = {
     status: () => ipcRenderer.invoke('model:status'),
     /** One agent turn: messages in, Reginald's prose + the tools it consulted out. */
     chat: (messages: unknown) => ipcRenderer.invoke('model:chat', messages),
+    /** Subscribe to streamed prose tokens for the in-flight turn; returns an unsubscribe. */
+    onToken: (cb: (delta: string) => void) => {
+      const listener = (_e: unknown, delta: string) => cb(delta)
+      ipcRenderer.on('model:chat:token', listener)
+      return () => ipcRenderer.removeListener('model:chat:token', listener)
+    },
     /** Decompose a braindump into a proposed issue set (capture_work). */
     capture: (braindump: string) => ipcRenderer.invoke('model:capture', braindump),
   },

apps/desktop/src/renderer/src/components/screens/milestone-screen.tsx

@@ -1,22 +1,44 @@
 import React from 'react'
 
 import { COLUMNS, type IssueRef } from '../../data/fixtures.js'
+import { type MilestoneView } from '../../lib/backlog.js'
 import { BurnUpCone } from '../charts/chart.js'
 import { Badge, Button, Card, Icon, Tag } from '../ui/index.js'
 
-// Milestone detail — scope, cone, issues; forecasts stay ranges
-export function MilestoneScreen({ onBack, onOpenIssue }: { onBack: () => void; onOpenIssue: (issue: IssueRef) => void }) {
+// Milestone detail — scope, cone, issues; forecasts stay ranges.
+// `data` (real milestone forecast) overrides the demo fixture when present.
+export function MilestoneScreen({
+  onBack,
+  onOpenIssue,
+  data,
+}: {
+  onBack: () => void
+  onOpenIssue: (issue: IssueRef) => void
+  data?: MilestoneView
+}) {
   const cols = COLUMNS
   const byState = (ids: number[]) =>
     cols.flatMap((c) => c.issues.map((i) => ({ ...i, col: c.label }))).filter((i) => ids.includes(i.id))
 
-  const groups = [
+  const fixtureGroups = [
     { label: 'Steeping', issues: byState([87, 84]) },
     { label: 'In review', issues: byState([92]) },
     { label: 'Queued', issues: byState([102, 103, 99, 96, 78]) },
     { label: 'Done', issues: byState([71, 69, 65]), muted: true },
   ]
 
+  // real or demo, in one shape the render loop understands
+  const groups = data
+    ? data.groups.map((g) => ({ label: g.label, issues: g.issues, muted: g.label === 'Done' }))
+    : fixtureGroups
+  const name = data ? data.name : 'Beta'
+  const dueLine = data
+    ? `milestone · due ${data.due} · ${data.soft ? 'soft — scope may flex' : 'hard deadline'}`
+    : 'milestone · due Mar 15 · soft — scope may flex'
+  const forecastLabel = data ? (data.forecastRange ? `80% ${data.forecastRange}` : 'all shipped') : '80% Mar 3–12'
+  const scopeStat = data ? `${data.scopeCount} issues · est ${data.scopeEstDays}d` : '42 issues · est 61d'
+  const doneStat = data ? `${data.doneCount} · ${data.donePct}%` : '24 · 57%'
+
   const Stat = ({ label, value, tone }: { label: string; value: string; tone?: string }) => (
     <div style={{ flex: 1, padding: '12px 18px', borderRight: '1px solid var(--line-1)' }}>
       <div style={{ font: 'var(--text-overline)', letterSpacing: 'var(--letter-spacing-wide)', textTransform: 'uppercase', color: 'var(--ink-3)', marginBottom: 5 }}>{label}</div>
@@ -36,12 +58,12 @@ export function MilestoneScreen({ onBack, onOpenIssue }: { onBack: () => void; o
         <header style={{ borderBottom: 'var(--rule-double)', paddingBottom: 14, display: 'flex', alignItems: 'flex-start', gap: 16 }}>
           <div style={{ flex: 1, minWidth: 0 }}>
             <p style={{ font: '400 12px var(--font-mono)', color: 'var(--ink-3)', margin: '0 0 6px', display: 'inline-flex', alignItems: 'center', gap: 6 }}>
-              <Icon name="milestone" size={13} /> milestone · due Mar 15 · soft — scope may flex
+              <Icon name="milestone" size={13} /> {dueLine}
             </p>
-            <h1 style={{ font: 'var(--text-display)', color: 'var(--ink-1)', margin: 0 }}>Beta</h1>
+            <h1 style={{ font: 'var(--text-display)', color: 'var(--ink-1)', margin: 0 }}>{name}</h1>
             <div style={{ display: 'flex', alignItems: 'center', gap: 8, marginTop: 10 }}>
-              <Badge tone="ok" dot>ahead of forecast</Badge>
-              <span style={{ font: '400 12px var(--font-mono)', color: 'var(--ink-2)', whiteSpace: 'nowrap' }}>80% Mar 3–12</span>
+              <Badge tone="ok" dot>{data ? `${data.doneCount}/${data.scopeCount} done` : 'ahead of forecast'}</Badge>
+              <span style={{ font: '400 12px var(--font-mono)', color: 'var(--ink-2)', whiteSpace: 'nowrap' }}>{forecastLabel}</span>
             </div>
           </div>
           <Button variant="secondary" icon="arrow-up-right">Open in Gitea</Button>
@@ -51,21 +73,25 @@ export function MilestoneScreen({ onBack, onOpenIssue }: { onBack: () => void; o
       {/* stats strip */}
       <Card flush>
         <div style={{ display: 'flex' }}>
-          <Stat label="Scope" value="42 issues · est 61d" />
-          <Stat label="Done" value="24 · 57%" />
-          <Stat label="Forecast" value="80% Mar 3–12" />
+          <Stat label="Scope" value={scopeStat} />
+          <Stat label="Done" value={doneStat} />
+          <Stat label="Forecast" value={forecastLabel} />
           <div style={{ flex: 1, padding: '12px 18px' }}>
-            <div style={{ font: 'var(--text-overline)', letterSpacing: 'var(--letter-spacing-wide)', textTransform: 'uppercase', color: 'var(--ink-3)', marginBottom: 5 }}>Drift · 7d</div>
-            <div style={{ font: '500 14px var(--font-mono)', color: 'var(--ok)', whiteSpace: 'nowrap' }}>−2d · cone narrowed</div>
+            <div style={{ font: 'var(--text-overline)', letterSpacing: 'var(--letter-spacing-wide)', textTransform: 'uppercase', color: 'var(--ink-3)', marginBottom: 5 }}>Remaining</div>
+            <div style={{ font: '500 14px var(--font-mono)', color: 'var(--ink-1)', whiteSpace: 'nowrap' }}>{data ? `${data.scopeCount - data.doneCount} open` : '−2d · cone narrowed'}</div>
           </div>
         </div>
       </Card>
 
       <div style={{ display: 'grid', gridTemplateColumns: '1.2fr 1fr', gap: 14, alignItems: 'start' }}>
-        <Card overline="Burn-up" title={<>80% this lands <span style={{ whiteSpace: 'nowrap' }}>Mar 3–12</span></>} jade>
-          <BurnUpCone />
+        <Card overline="Burn-up" title={<>80% this lands <span style={{ whiteSpace: 'nowrap' }}>{data ? (data.forecastRange ?? 'shipped') : 'Mar 3–12'}</span></>} jade>
+          <BurnUpCone data={data?.cone ?? undefined} />
           <p style={{ font: 'var(--text-agent)', color: 'var(--ink-2)', margin: '10px 0 0' }}>
-            Comfortably ahead. Beta needs #87 more than it needs my commentary.
+            {data
+              ? data.cone
+                ? `${data.scopeCount - data.doneCount} open of ${data.scopeCount}. Cone over what's left.`
+                : 'Everything here has shipped.'
+              : 'Comfortably ahead. Beta needs #87 more than it needs my commentary.'}
           </p>
         </Card>
 

apps/desktop/src/renderer/src/components/screens/runway-screen.tsx

@@ -2,18 +2,22 @@ import React from 'react'
 
 import { RunwayBar } from '../charts/chart.js'
 import { Card, Badge, Tag, Icon, IconButton } from '../ui/index.js'
-import { RUNWAY, CAPACITY } from '../../data/fixtures.js'
+import { RUNWAY, CAPACITY, type RunwayMilestone } from '../../data/fixtures.js'
 
-// Runway — capacity vs milestone dates; ranges, never points
+// Runway — capacity vs milestone dates; ranges, never points.
+// `milestones` (real per-milestone forecasts) overrides the demo when present.
 export function RunwayScreen({
   onOpenCalibration,
   onOpenMilestone,
   calibration,
+  milestones,
 }: {
   onOpenCalibration: () => void
-  onOpenMilestone: () => void
+  onOpenMilestone: (id?: number) => void
   calibration?: { n: number; coldStart: boolean }
+  milestones?: RunwayMilestone[]
 }) {
+  const rows = milestones && milestones.length ? milestones : RUNWAY
   const calibNote = calibration
     ? calibration.coldStart
       ? `cold-start priors · ${calibration.n}/20 closed issues estimated`
@@ -28,8 +32,8 @@ export function RunwayScreen({
 
       <Card overline="Milestones" flush>
         <div>
-          {RUNWAY.map((m, i) => (
-            <div key={m.name} onClick={onOpenMilestone} style={{
+          {rows.map((m, i) => (
+            <div key={m.name} onClick={() => onOpenMilestone(m.id)} style={{
               display: 'grid', gridTemplateColumns: '160px 1fr 150px 90px', gap: 16, alignItems: 'center', cursor: 'pointer',
               padding: '14px 20px', borderTop: i === 0 ? 'none' : '1px solid var(--line-1)',
             }}

apps/desktop/src/renderer/src/components/shell/app-shell.tsx

@@ -4,7 +4,14 @@ import logoIcon from '../../design/assets/logo-icon.png'
 import type { IssueChange } from '@commitea/core'
 
 import type { IssueRef } from '../../data/fixtures.js'
-import { backlogCalibration, forecastBacklog, issuesToBoardColumns, scheduleFocus } from '../../lib/backlog.js'
+import {
+  backlogCalibration,
+  forecastBacklog,
+  issuesToBoardColumns,
+  milestoneView,
+  runwayView,
+  scheduleFocus,
+} from '../../lib/backlog.js'
 import { useBacklog } from '../../lib/use-backlog.js'
 import { PrimitivesGallery } from '../gallery.js'
 import { BoardScreen } from '../screens/board-screen.js'
@@ -87,6 +94,7 @@ export function AppShell() {
   const [offline, setOffline] = useState(false)
   const [issue, setIssue] = useState<IssueRef | null>(null)
   const [readIds, setReadIds] = useState<number[]>([])
+  const [milestoneId, setMilestoneId] = useState<number | null>(null)
   const [backlog, refetchBacklog] = useBacklog()
   const boardColumns =
     backlog.status === 'ready' ? issuesToBoardColumns(backlog.issues, backlog.timelines) : undefined
@@ -98,6 +106,19 @@ export function AppShell() {
     backlog.status === 'ready'
       ? (forecastBacklog(backlog.issues, backlog.deps, new Date(), calibration?.model) ?? undefined)
       : undefined
+  const runwayMilestones =
+    backlog.status === 'ready' ? runwayView(backlog.issues, backlog.milestones, backlog.deps) : undefined
+  const milestone =
+    backlog.status === 'ready' && milestoneId != null
+      ? (milestoneView(
+          milestoneId,
+          backlog.issues,
+          backlog.milestones,
+          backlog.deps,
+          backlog.timelines,
+          calibration?.model,
+        ) ?? undefined)
+      : undefined
 
   useEffect(() => {
     document.documentElement.setAttribute('data-theme', dark ? 'dark' : 'light')
@@ -194,14 +215,18 @@ export function AppShell() {
         return (
           <RunwayScreen
             onOpenCalibration={() => setView('calibration')}
-            onOpenMilestone={() => setView('milestone')}
+            onOpenMilestone={(id) => {
+              setMilestoneId(id ?? null)
+              setView('milestone')
+            }}
             calibration={calibration ? { n: calibration.model.n, coldStart: calibration.model.coldStart } : undefined}
+            milestones={runwayMilestones}
           />
         )
       case 'calibration':
         return <CalibrationScreen onBack={() => setView('runway')} data={calibration?.data} />
       case 'milestone':
-        return <MilestoneScreen onBack={() => setView('runway')} onOpenIssue={openIssue} />
+        return <MilestoneScreen onBack={() => setView('runway')} onOpenIssue={openIssue} data={milestone} />
       case 'inbox':
         return (
           <InboxScreen

apps/desktop/src/renderer/src/components/shell/chat-panel.tsx

@@ -19,7 +19,7 @@ export interface ChatPanelProps {
 }
 
 export function ChatPanel({ onOpenDirectives, offline, onApplyChange }: ChatPanelProps) {
-  const { msgs, thinking, live, model, steps, proposals, send: sendChat, approve, dismiss } = useChat(onApplyChange)
+  const { msgs, thinking, live, model, steps, proposals, streaming, send: sendChat, approve, dismiss } = useChat(onApplyChange)
   // shorten "google/gemma-4-26b-a4b-qat" → "gemma-4-26b" for the header chip
   const modelLabel = model ? (model.split('/').pop() ?? model).replace(/-(qat|instruct|it|gguf)$/i, '') : 'gemma-4'
   const [text, setText] = useState('')
@@ -28,7 +28,7 @@ export function ChatPanel({ onOpenDirectives, offline, onApplyChange }: ChatPane
   useEffect(() => {
     const el = scrollRef.current
     if (el) el.scrollTop = el.scrollHeight
-  }, [msgs, thinking])
+  }, [msgs, thinking, streaming])
 
   const send = () => {
     const t = text.trim()
@@ -98,7 +98,14 @@ export function ChatPanel({ onOpenDirectives, offline, onApplyChange }: ChatPane
             The model is away from its desk. Reads still work; writes will wait their turn.
           </div>
         ) : null}
-        {thinking ? <div style={{ font: 'var(--text-agent)', color: 'var(--ink-3)' }}>considering…</div> : null}
+        {streaming ? (
+          <div style={{ font: 'var(--text-agent)', color: 'var(--ink-1)', lineHeight: 1.55 }}>
+            {streaming}
+            <span style={{ opacity: 0.5 }}>▊</span>
+          </div>
+        ) : thinking ? (
+          <div style={{ font: 'var(--text-agent)', color: 'var(--ink-3)' }}>considering…</div>
+        ) : null}
         {!thinking && steps.length ? (
           <div style={{ font: 'var(--text-caption)', color: 'var(--ink-3)', display: 'flex', alignItems: 'center', gap: 5 }}>
             <Icon name="eye" size={11} /> consulted {Array.from(new Set(steps.map((s) => s.replace('query_project', 'the project').replace('propose_change', 'the labels').replace('record_directive', 'the directive ledger')))).join(', ')}

apps/desktop/src/renderer/src/data/fixtures.ts

@@ -315,6 +315,8 @@ export interface RunwayMilestone {
   spread: number
   tone: 'ok' | 'warn'
   note: string
+  /** gitea milestone id, for the drill-in (absent on the demo fixture). */
+  id?: number
 }
 
 export const RUNWAY: RunwayMilestone[] = [

apps/desktop/src/renderer/src/global.d.ts

@@ -68,6 +68,8 @@ export interface ModelBridge {
   status(): Promise<{ configured: boolean; model: string | null }>
   chat(messages: ChatMessage[]): Promise<ChatResult>
   capture(braindump: string): Promise<CaptureResult>
+  /** Subscribe to streamed prose tokens; returns an unsubscribe fn. */
+  onToken(cb: (delta: string) => void): () => void
 }
 
 /** The result of reading the directive ledger. */

apps/desktop/src/renderer/src/lib/backlog.ts

@@ -7,6 +7,7 @@ import {
   fitCalibration,
   forecast,
   type GiteaIssue,
+  type GiteaMilestone,
   inferLifecycle,
   type LifecycleColumn,
   type LifecycleEvent,
@@ -16,10 +17,17 @@ import {
   type ScheduledItem,
   selectFocus,
   toDurationModel,
+  workingDaysBetween,
 } from '@commitea/core'
 
-import { type BoardColumn, type BoardIssue, type CalibrationData, type FocusIssue } from '../data/fixtures.js'
-import { type BurnUpData, buildBurnUpData } from './dates.js'
+import {
+  type BoardColumn,
+  type BoardIssue,
+  type CalibrationData,
+  type FocusIssue,
+  type RunwayMilestone,
+} from '../data/fixtures.js'
+import { addWorkingDays, type BurnUpData, buildBurnUpData, formatRange, formatShort } from './dates.js'
 
 type Timelines = Record<number, LifecycleEvent[]>
 
@@ -232,3 +240,133 @@ export function scheduleFocus(
     later: toFocusIssue(f.later, inf),
   }
 }
+
+/**
+ * Per-milestone Monte Carlo forecast for the Runway screen — each open milestone's
+ * open scope gets its own cone, and the p80 landing is compared to the milestone's
+ * due date (ok/at-risk). Ranges, never point dates. Milestones with no open scope
+ * (already shipped) are omitted. Falls back to the demo when there's nothing real.
+ */
+export function runwayView(
+  issues: GiteaIssue[],
+  milestones: GiteaMilestone[],
+  deps: DependencyEdge[],
+  today: Date = new Date(),
+): RunwayMilestone[] {
+  const open = issues.filter((i) => i.state === 'open')
+  const rows = milestones
+    .filter((m) => m.state === 'open')
+    .map((m) => ({ m, scope: open.filter((i) => i.milestone?.id === m.id) }))
+    .filter((x) => x.scope.length > 0)
+    .map(({ m, scope }) => {
+      const f = forecast(
+        scope.map((i) => ({
+          number: i.number,
+          title: i.title,
+          labels: i.labels,
+          estimateDays: i.facts.estimateDays,
+          priority: i.facts.priority,
+        })),
+        deps,
+      )
+      const p90 = f.curve.length ? f.curve[f.curve.length - 1].p90Day : f.p95Day
+      const dueDay = m.dueOn ? workingDaysBetween(today, new Date(m.dueOn)) : null
+      return { m, p50: f.p50Day, p80: f.p80Day, p90, dueDay }
+    })
+
+  const horizon = Math.max(1, ...rows.map((r) => Math.max(r.p90, r.dueDay ?? 0)))
+
+  return rows.map(({ m, p50, p80, p90, dueDay }) => {
+    const onTrack = dueDay == null || p80 <= dueDay
+    return {
+      id: m.id,
+      name: m.title,
+      due: m.dueOn ? formatShort(new Date(m.dueOn)) : 'no date',
+      hard: false,
+      p80: formatRange(addWorkingDays(today, p50), addWorkingDays(today, p90)),
+      pos: p80 / horizon,
+      spread: Math.min(0.6, (p90 - p50) / horizon),
+      tone: onTrack ? ('ok' as const) : ('warn' as const),
+      note: dueDay == null ? `${Math.ceil(p80)}d of work` : onTrack ? 'on track' : 'at risk',
+    }
+  })
+}
+
+export interface MilestoneGroup {
+  label: string
+  issues: { id: number; title: string; labels: string[]; days?: string }[]
+}
+
+export interface MilestoneView {
+  name: string
+  due: string
+  soft: boolean
+  scopeCount: number
+  scopeEstDays: number
+  doneCount: number
+  donePct: number
+  /** p50..p90 landing range for the remaining open scope; null when nothing's open. */
+  forecastRange: string | null
+  cone: BurnUpData | null
+  groups: MilestoneGroup[]
+}
+
+/**
+ * The Milestone drill-in: real scope, done %, a Monte Carlo cone over the
+ * milestone's remaining open work, and its issues grouped by lifecycle column.
+ * Returns null for an unknown id — the screen then shows the demo fixture.
+ */
+export function milestoneView(
+  id: number,
+  issues: GiteaIssue[],
+  milestones: GiteaMilestone[],
+  deps: DependencyEdge[],
+  timelines: Timelines = {},
+  calibration?: CalibrationModel,
+  today: Date = new Date(),
+): MilestoneView | null {
+  const m = milestones.find((x) => x.id === id)
+  if (!m) return null
+  const all = issues.filter((i) => i.milestone?.id === id)
+  const open = all.filter((i) => i.state === 'open')
+  const done = all.filter((i) => i.state === 'closed')
+  const scopeEstDays = all.reduce((sum, i) => sum + (i.facts.estimateDays ?? 2), 0)
+
+  const model = calibration ? toDurationModel(calibration) : undefined
+  const f = forecast(
+    open.map((i) => ({
+      number: i.number,
+      title: i.title,
+      labels: i.labels,
+      estimateDays: i.facts.estimateDays,
+      priority: i.facts.priority,
+    })),
+    deps,
+    model ? { model } : {},
+  )
+  const cone = buildBurnUpData(f, today)
+
+  const inf = inferAll(all, timelines, today)
+  const groups: MilestoneGroup[] = COLUMN_ORDER.map((key) => ({
+    label: COLUMN_LABELS[key],
+    issues: all
+      .filter((i) => inf.get(i.number)!.column === key)
+      .map((i) => {
+        const days = inf.get(i.number)!.steepingDays
+        return { id: i.number, title: i.title, labels: i.labels, days: days != null ? `${days}d` : undefined }
+      }),
+  })).filter((g) => g.issues.length > 0)
+
+  return {
+    name: m.title,
+    due: m.dueOn ? formatShort(new Date(m.dueOn)) : 'no date',
+    soft: true,
+    scopeCount: all.length,
+    scopeEstDays,
+    doneCount: done.length,
+    donePct: all.length ? Math.round((done.length / all.length) * 100) : 0,
+    forecastRange: cone ? cone.rangeLabel : null,
+    cone,
+    groups,
+  }
+}

apps/desktop/src/renderer/src/lib/use-chat.ts

@@ -20,6 +20,8 @@ export interface ChatState {
   steps: string[]
   /** Changes Reginald has proposed and is awaiting approval on. */
   proposals: ChangeProposal[]
+  /** The in-flight streamed prose (grows token-by-token) before the turn finalizes. */
+  streaming: string
   send: (text: string) => void
   approve: (p: ChangeProposal) => void
   dismiss: (p: ChangeProposal) => void
@@ -41,6 +43,7 @@ export function useChat(onApplyChange?: (change: IssueChange) => Promise<{ ok: b
   const [model, setModel] = useState<string | null>(null)
   const [steps, setSteps] = useState<string[]>([])
   const [proposals, setProposals] = useState<ChangeProposal[]>([])
+  const [streaming, setStreaming] = useState('')
   const convoRef = useRef(convo)
   convoRef.current = convo
 
@@ -86,10 +89,17 @@ export function useChat(onApplyChange?: (change: IssueChange) => Promise<{ ok: b
         role: m.from === 'user' ? 'user' : 'assistant',
         content: m.text,
       }))
+      setStreaming('')
+      const unsubscribe = window.commitea.model.onToken((delta) => setStreaming((s) => s + delta))
+      const finish = () => {
+        unsubscribe()
+        setThinking(false)
+        setStreaming('')
+      }
       window.commitea.model
         .chat(wire)
         .then((res) => {
-          setThinking(false)
+          finish()
           if (res.ok) {
             setSteps(res.steps.map((s) => s.tool))
             setProposals(res.proposals)
@@ -105,7 +115,7 @@ export function useChat(onApplyChange?: (change: IssueChange) => Promise<{ ok: b
           }
         })
         .catch(() => {
-          setThinking(false)
+          finish()
           setConvo((c) => [...c, { from: 'agent', text: 'I could not reach the model.' }])
         })
     },
@@ -137,5 +147,5 @@ export function useChat(onApplyChange?: (change: IssueChange) => Promise<{ ok: b
     setConvo((c) => [...c, { from: 'agent', text: `Left #${p.change.issue} as it was.` }])
   }, [])
 
-  return { msgs: [...seed, ...convo], thinking, live, model, steps, proposals, send, approve, dismiss }
+  return { msgs: [...seed, ...convo], thinking, live, model, steps, proposals, streaming, send, approve, dismiss }
 }

packages/core/src/agent/agent-loop.ts

@@ -31,18 +31,20 @@ function stringify(result: unknown): string {
 }
 
 export async function runAgentTurn(opts: {
-  complete: (messages: ChatMessage[], tools?: ToolDecl[]) => Promise<CompletionResult>
+  complete: (messages: ChatMessage[], tools?: ToolDecl[], onToken?: (delta: string) => void) => Promise<CompletionResult>
   messages: ChatMessage[]
   tools: ToolDecl[]
   execute: ToolExecutor
   maxSteps?: number
+  /** Streams content deltas as the model produces prose (final-answer streaming). */
+  onToken?: (delta: string) => void
 }): Promise<AgentTurn> {
   const maxSteps = opts.maxSteps ?? DEFAULT_MAX_STEPS
   const convo: ChatMessage[] = [...opts.messages]
   const steps: AgentStep[] = []
 
   for (let step = 0; step < maxSteps; step++) {
-    const { content, toolCalls } = await opts.complete(convo, opts.tools)
+    const { content, toolCalls } = await opts.complete(convo, opts.tools, opts.onToken)
     if (toolCalls.length === 0) {
       convo.push({ role: 'assistant', content })
       return { content, steps, messages: convo }
@@ -63,7 +65,7 @@ export async function runAgentTurn(opts: {
   }
 
   // Out of tool budget — force a final prose answer with tools withheld.
-  const final = await opts.complete(convo, [])
+  const final = await opts.complete(convo, [], opts.onToken)
   convo.push({ role: 'assistant', content: final.content })
   return { content: final.content, steps, messages: convo }
 }

packages/core/src/agent/agent.test.ts

@@ -25,6 +25,48 @@ describe('createChatClient', () => {
     return { fetch, calls }
   }
 
+  function sseStub(chunks: string[]) {
+    const calls: { url: string; body: unknown }[] = []
+    const fetch: FetchLike = (url, init) => {
+      calls.push({ url, body: init?.body ? JSON.parse(init.body) : undefined })
+      const enc = new TextEncoder()
+      const body = new ReadableStream<Uint8Array>({
+        start(c) {
+          for (const ch of chunks) c.enqueue(enc.encode(ch))
+          c.close()
+        },
+      })
+      return Promise.resolve({ ok: true, status: 200, body, json: () => Promise.resolve({}), text: () => Promise.resolve('') })
+    }
+    return { fetch, calls }
+  }
+
+  it('streams content deltas via onToken and returns the assembled result', async () => {
+    const { fetch, calls } = sseStub([
+      'data: {"choices":[{"delta":{"content":"Right "}}]}\n\n',
+      'data: {"choices":[{"delta":{"content":"now: #2."}}]}\n\n',
+      'data: [DONE]\n\n',
+    ])
+    const client = createChatClient({ baseUrl: 'http://x/v1', model: 'm' }, fetch)
+    const tokens: string[] = []
+    const res = await client.complete([{ role: 'user', content: 'now?' }], undefined, (d) => tokens.push(d))
+
+    expect(tokens).toEqual(['Right ', 'now: #2.'])
+    expect(res.content).toBe('Right now: #2.')
+    expect((calls[0].body as { stream?: boolean }).stream).toBe(true)
+  })
+
+  it('assembles a streamed tool call from argument deltas', async () => {
+    const { fetch } = sseStub([
+      'data: {"choices":[{"delta":{"tool_calls":[{"index":0,"id":"c1","function":{"name":"query_project","arguments":"{\\"view\\""}}]}}]}\n\n',
+      'data: {"choices":[{"delta":{"tool_calls":[{"index":0,"function":{"arguments":":\\"focus\\"}"}}]}}]}\n\n',
+      'data: [DONE]\n\n',
+    ])
+    const client = createChatClient({ baseUrl: 'http://x/v1', model: 'm' }, fetch)
+    const res = await client.complete([{ role: 'user', content: 'x' }], [{ name: 'query_project', description: '', parameters: {} }], () => {})
+    expect(res.toolCalls).toEqual([{ id: 'c1', name: 'query_project', arguments: '{"view":"focus"}' }])
+  })
+
   it('POSTs to /chat/completions and parses content', async () => {
     const { fetch, calls } = stub({ choices: [{ message: { content: 'the focus is #2' } }] })
     const client = createChatClient({ baseUrl: 'http://localhost:1234/v1', model: 'gemma' }, fetch)

packages/core/src/agent/chat-client.ts

@@ -48,8 +48,12 @@ export interface CompletionResult {
   toolCalls: ToolCall[]
 }
 
+/** Called with each streamed content delta (final-prose streaming). */
+export type OnToken = (delta: string) => void
+
 export interface ChatClient {
-  complete(messages: ChatMessage[], tools?: ToolDecl[]): Promise<CompletionResult>
+  /** When `onToken` is given, the response streams (SSE) and each content delta is emitted. */
+  complete(messages: ChatMessage[], tools?: ToolDecl[], onToken?: OnToken): Promise<CompletionResult>
 }
 
 /** Map our message shape to the OpenAI wire shape. */
@@ -85,7 +89,7 @@ export function createChatClient(config: ModelConfig, fetchImpl: FetchLike): Cha
   const url = `${config.baseUrl.replace(/\/+$/, '')}/chat/completions`
 
   return {
-    async complete(messages, tools) {
+    async complete(messages, tools, onToken) {
       const body: Record<string, unknown> = {
         model: config.model,
         messages: messages.map(toWireMessage),
@@ -95,11 +99,13 @@ export function createChatClient(config: ModelConfig, fetchImpl: FetchLike): Cha
         body.tools = tools.map(toWireTool)
         body.tool_choice = 'auto'
       }
+      const stream = !!onToken
+      if (stream) body.stream = true
       const res = await fetchImpl(url, {
         method: 'POST',
         headers: {
           'Content-Type': 'application/json',
-          Accept: 'application/json',
+          Accept: stream ? 'text/event-stream' : 'application/json',
           ...(config.apiKey ? { Authorization: `Bearer ${config.apiKey}` } : {}),
         },
         body: JSON.stringify(body),
@@ -108,6 +114,8 @@ export function createChatClient(config: ModelConfig, fetchImpl: FetchLike): Cha
         const text = await res.text().catch(() => '')
         throw new Error(`model completion failed (${res.status}): ${text.slice(0, 200)}`)
       }
+      if (stream && res.body) return readStream(res.body, onToken!)
+
       const json = (await res.json()) as { choices?: { message: RawChoiceMessage }[] }
       const msg = json.choices?.[0]?.message
       return {
@@ -121,3 +129,56 @@ export function createChatClient(config: ModelConfig, fetchImpl: FetchLike): Cha
     },
   }
 }
+
+/** Streamed tool-call delta: name arrives first, arguments accumulate across chunks. */
+interface RawToolDelta {
+  index: number
+  id?: string
+  function?: { name?: string; arguments?: string }
+}
+
+/** Parse an OpenAI SSE stream: emit content deltas via onToken, accumulate the final result. */
+async function readStream(body: ReadableStream<Uint8Array>, onToken: OnToken): Promise<CompletionResult> {
+  const reader = body.getReader()
+  const decoder = new TextDecoder()
+  let buffer = ''
+  let content = ''
+  const toolAcc: { id: string; name: string; arguments: string }[] = []
+
+  const handle = (data: string) => {
+    if (data === '[DONE]') return
+    let chunk: { choices?: { delta?: { content?: string; tool_calls?: RawToolDelta[] } }[] }
+    try {
+      chunk = JSON.parse(data)
+    } catch {
+      return
+    }
+    const delta = chunk.choices?.[0]?.delta
+    if (!delta) return
+    if (delta.content) {
+      content += delta.content
+      onToken(delta.content)
+    }
+    for (const tc of delta.tool_calls ?? []) {
+      const slot = (toolAcc[tc.index] ??= { id: '', name: '', arguments: '' })
+      if (tc.id) slot.id = tc.id
+      if (tc.function?.name) slot.name = tc.function.name
+      if (tc.function?.arguments) slot.arguments += tc.function.arguments
+    }
+  }
+
+  for (;;) {
+    const { done, value } = await reader.read()
+    if (done) break
+    buffer += decoder.decode(value, { stream: true })
+    const lines = buffer.split('\n')
+    buffer = lines.pop() ?? ''
+    for (const line of lines) {
+      const t = line.trim()
+      if (t.startsWith('data:')) handle(t.slice(5).trim())
+    }
+  }
+  if (buffer.trim().startsWith('data:')) handle(buffer.trim().slice(5).trim())
+
+  return { content, toolCalls: toolAcc.filter((t) => t.name).map((t) => ({ id: t.id, name: t.name, arguments: t.arguments })) }
+}

packages/core/src/gitea/types.ts

@@ -31,6 +31,8 @@ export interface GiteaHttpResponse {
   status: number
   json(): Promise<unknown>
   text(): Promise<string>
+  /** Present on the real fetch Response; used for SSE streaming (chat). */
+  body?: ReadableStream<Uint8Array> | null
 }
 
 export type FetchLike = (url: string, init?: GiteaRequestInit) => Promise<GiteaHttpResponse>

packages/core/src/index.ts

@@ -80,7 +80,7 @@ export type {
 } from './calibration/calibration-v0.js'
 
 export { createChatClient } from './agent/chat-client.js'
-export type { ChatClient, ChatMessage, CompletionResult, ModelConfig, ToolCall, ToolDecl } from './agent/chat-client.js'
+export type { ChatClient, ChatMessage, CompletionResult, ModelConfig, OnToken, ToolCall, ToolDecl } from './agent/chat-client.js'
 export { pickModel } from './agent/model-router.js'
 export type { ModelRouter, TaskKind } from './agent/model-router.js'
 export { runAgentTurn } from './agent/agent-loop.js'