Merge pull request 'feat: capacity-aware scheduling (#8) — real focus factors drive every forecast' (#50) from feat/capacity into main

Reviewed-on: #50

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

@@ -44,6 +44,9 @@ test.describe('live backlog', () => {
     // 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()
+    // Real capacity config from pm-state (christian/stephen), not the fixture (Stephen/Ana K.)
+    await expect(win.getByText('christian', { exact: true })).toBeVisible()
+    await expect(win.getByText(/pd\/day/).first()).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

apps/desktop/src/main/gitea.ts

@@ -20,6 +20,7 @@ import {
   type IssueChange,
   type LifecycleEvent,
   makeDirectiveEntry,
+  parseCapacityConfig,
   parseDirectiveLog,
   planIssueChange,
   type ProjectSnapshot,
@@ -109,6 +110,19 @@ export async function readDirectives(client: GiteaClient) {
   return parseDirectiveLog(text)
 }
 
+const CAPACITY_PATH = 'capacity/members.json'
+
+/** Read the capacity config from the pm-state repo (empty when absent). */
+export async function readCapacity(client: GiteaClient) {
+  const file = await client.getFile(CAPACITY_PATH)
+  if (!file) return []
+  try {
+    return parseCapacityConfig(JSON.parse(Buffer.from(file.contentBase64, 'base64').toString('utf8')))
+  } catch {
+    return []
+  }
+}
+
 /** Full reconcile: issues + milestones + native deps + lifecycle timelines. */
 export async function reconcileSnapshot(
   client: GiteaClient,
@@ -266,4 +280,15 @@ export function registerGiteaIpc(): void {
       return { ok: false as const, reason: 'error' as const, message: e instanceof Error ? e.message : String(e) }
     }
   })
+
+  // Read the capacity config from the pm-state repo (for capacity-aware forecasts).
+  ipcMain.handle('pmstate:capacity', async () => {
+    const pm = getPmStateClient()
+    if (!pm) return { ok: false as const, reason: 'unconfigured' as const, members: [] }
+    try {
+      return { ok: true as const, members: await readCapacity(pm) }
+    } catch {
+      return { ok: true as const, members: [] }
+    }
+  })
 }

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

@@ -19,12 +19,20 @@ const api = {
   pmstate: {
     /** Read the directive ledger from the pm-state repo. */
     directives: () => ipcRenderer.invoke('pmstate:directives'),
+    /** Read the capacity config from the pm-state repo. */
+    capacity: () => ipcRenderer.invoke('pmstate:capacity'),
   },
   model: {
     /** Whether a model endpoint is configured (else the UI keeps the scripted Reginald). */
     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/runway-screen.tsx

@@ -1,23 +1,35 @@
 import React from 'react'
 
+import { type CapacityMember, capacityPerWorkday } from '@commitea/core'
+
 import { RunwayBar } from '../charts/chart.js'
 import { Card, Badge, Tag, Icon, IconButton } from '../ui/index.js'
 import { RUNWAY, CAPACITY, type RunwayMilestone } from '../../data/fixtures.js'
 
 // Runway — capacity vs milestone dates; ranges, never points.
-// `milestones` (real per-milestone forecasts) overrides the demo when present.
+// `milestones` (real per-milestone forecasts) + `capacity` (real config) override the demo.
 export function RunwayScreen({
   onOpenCalibration,
   onOpenMilestone,
   calibration,
   milestones,
+  capacity,
 }: {
   onOpenCalibration: () => void
   onOpenMilestone: (id?: number) => void
   calibration?: { n: number; coldStart: boolean }
   milestones?: RunwayMilestone[]
+  capacity?: CapacityMember[]
 }) {
   const rows = milestones && milestones.length ? milestones : RUNWAY
+  const capacityRows =
+    capacity && capacity.length
+      ? capacity.map((m) => ({
+          who: m.person,
+          slices: `focus ${m.focusFactor} · alloc ${Math.round(m.allocation * 100)}%`,
+          hours: `${capacityPerWorkday(m).toFixed(2)} pd/day`,
+        }))
+      : CAPACITY
   const calibNote = calibration
     ? calibration.coldStart
       ? `cold-start priors · ${calibration.n}/20 closed issues estimated`
@@ -59,7 +71,7 @@ export function RunwayScreen({
       <div style={{ display: 'grid', gridTemplateColumns: '1fr 1fr', gap: 14, alignItems: 'start' }}>
         <Card overline="Capacity" flush>
           <div>
-            {CAPACITY.map((p, i) => (
+            {capacityRows.map((p, i) => (
               <div key={p.who} style={{
                 display: 'flex', alignItems: 'center', gap: 12, padding: '12px 20px',
                 borderTop: i === 0 ? 'none' : '1px solid var(--line-1)',

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

@@ -6,6 +6,7 @@ import type { IssueChange } from '@commitea/core'
 import type { IssueRef } from '../../data/fixtures.js'
 import {
   backlogCalibration,
+  capacityWorkers,
   forecastBacklog,
   issuesToBoardColumns,
   milestoneView,
@@ -13,6 +14,7 @@ import {
   scheduleFocus,
 } from '../../lib/backlog.js'
 import { useBacklog } from '../../lib/use-backlog.js'
+import { useCapacity } from '../../lib/use-capacity.js'
 import { PrimitivesGallery } from '../gallery.js'
 import { BoardScreen } from '../screens/board-screen.js'
 import { CalibrationScreen } from '../screens/calibration-screen.js'
@@ -96,6 +98,8 @@ export function AppShell() {
   const [readIds, setReadIds] = useState<number[]>([])
   const [milestoneId, setMilestoneId] = useState<number | null>(null)
   const [backlog, refetchBacklog] = useBacklog()
+  const capacityMembers = useCapacity()
+  const workers = capacityWorkers(capacityMembers)
   const boardColumns =
     backlog.status === 'ready' ? issuesToBoardColumns(backlog.issues, backlog.timelines) : undefined
   const focus =
@@ -104,10 +108,12 @@ export function AppShell() {
     backlog.status === 'ready' ? backlogCalibration(backlog.issues, backlog.timelines) : undefined
   const forecast =
     backlog.status === 'ready'
-      ? (forecastBacklog(backlog.issues, backlog.deps, new Date(), calibration?.model) ?? undefined)
+      ? (forecastBacklog(backlog.issues, backlog.deps, new Date(), calibration?.model, workers) ?? undefined)
       : undefined
   const runwayMilestones =
-    backlog.status === 'ready' ? runwayView(backlog.issues, backlog.milestones, backlog.deps) : undefined
+    backlog.status === 'ready'
+      ? runwayView(backlog.issues, backlog.milestones, backlog.deps, new Date(), workers)
+      : undefined
   const milestone =
     backlog.status === 'ready' && milestoneId != null
       ? (milestoneView(
@@ -117,6 +123,8 @@ export function AppShell() {
           backlog.deps,
           backlog.timelines,
           calibration?.model,
+          new Date(),
+          workers,
         ) ?? undefined)
       : undefined
 
@@ -221,6 +229,7 @@ export function AppShell() {
             }}
             calibration={calibration ? { n: calibration.model.n, coldStart: calibration.model.coldStart } : undefined}
             milestones={runwayMilestones}
+            capacity={capacityMembers}
           />
         )
       case 'calibration':

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/global.d.ts

@@ -1,5 +1,6 @@
 import type {
   AgentStep,
+  CapacityMember,
   ChangeProposal,
   ChatMessage,
   CaptureProposal,
@@ -68,6 +69,8 @@ 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. */
@@ -75,9 +78,15 @@ export type DirectivesResult =
   | { ok: false; reason: 'unconfigured' | 'error'; message?: string }
   | { ok: true; directives: DirectiveRecord[] }
 
+/** The result of reading the capacity config. */
+export type CapacityResult =
+  | { ok: false; reason: 'unconfigured'; members: CapacityMember[] }
+  | { ok: true; members: CapacityMember[] }
+
 /** The pm-state bridge (machine-derived state) exposed by the preload over IPC. */
 export interface PmStateBridge {
   directives(): Promise<DirectivesResult>
+  capacity(): Promise<CapacityResult>
 }
 
 declare global {

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

@@ -2,10 +2,13 @@ import {
   type CalibrationModel,
   type CalibrationSample,
   calibrationSamples,
+  type CapacityMember,
+  capacityPerWorkday,
   COLD_START_THRESHOLD,
   type DependencyEdge,
   fitCalibration,
   forecast,
+  type Worker,
   type GiteaIssue,
   type GiteaMilestone,
   inferLifecycle,
@@ -107,9 +110,15 @@ function toSchedulable(issues: GiteaIssue[]) {
       labels: i.labels,
       estimateDays: i.facts.estimateDays,
       priority: i.facts.priority,
+      assignee: i.assignee,
     }))
 }
 
+/** gitea CapacityMembers → scheduler lanes (person + throughput). */
+export function capacityWorkers(members: CapacityMember[]): Worker[] {
+  return members.map((m) => ({ person: m.person, speed: capacityPerWorkday(m) }))
+}
+
 export interface ForecastView {
   scope: number
   cone: BurnUpData
@@ -132,9 +141,10 @@ export function forecastBacklog(
   deps: DependencyEdge[],
   today: Date = new Date(),
   calibration?: CalibrationModel,
+  workers: Worker[] = [],
 ): ForecastView | null {
   const model = calibration ? toDurationModel(calibration) : undefined
-  const f = forecast(toSchedulable(issues), deps, model ? { model } : {})
+  const f = forecast(toSchedulable(issues), deps, { ...(model ? { model } : {}), workers })
   const cone = buildBurnUpData(f, today)
   if (!cone) return null
   return {
@@ -252,6 +262,7 @@ export function runwayView(
   milestones: GiteaMilestone[],
   deps: DependencyEdge[],
   today: Date = new Date(),
+  workers: Worker[] = [],
 ): RunwayMilestone[] {
   const open = issues.filter((i) => i.state === 'open')
   const rows = milestones
@@ -266,8 +277,10 @@ export function runwayView(
           labels: i.labels,
           estimateDays: i.facts.estimateDays,
           priority: i.facts.priority,
+          assignee: i.assignee,
         })),
         deps,
+        { workers },
       )
       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
@@ -324,6 +337,7 @@ export function milestoneView(
   timelines: Timelines = {},
   calibration?: CalibrationModel,
   today: Date = new Date(),
+  workers: Worker[] = [],
 ): MilestoneView | null {
   const m = milestones.find((x) => x.id === id)
   if (!m) return null
@@ -340,9 +354,10 @@ export function milestoneView(
       labels: i.labels,
       estimateDays: i.facts.estimateDays,
       priority: i.facts.priority,
+      assignee: i.assignee,
     })),
     deps,
-    model ? { model } : {},
+    { ...(model ? { model } : {}), workers },
   )
   const cone = buildBurnUpData(f, today)
 

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

@@ -0,0 +1,24 @@
+import { useEffect, useState } from 'react'
+
+import type { CapacityMember } from '@commitea/core'
+
+/**
+ * The team's capacity config from the pm-state repo. Empty when unconfigured —
+ * forecasts then fall back to a single serial worker. Read once on mount.
+ */
+export function useCapacity(): CapacityMember[] {
+  const [members, setMembers] = useState<CapacityMember[]>([])
+  useEffect(() => {
+    let alive = true
+    window.commitea.pmstate
+      .capacity()
+      .then((r) => {
+        if (alive) setMembers(r.members)
+      })
+      .catch(() => {})
+    return () => {
+      alive = false
+    }
+  }, [])
+  return members
+}

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/capacity/capacity-v0.test.ts

@@ -0,0 +1,88 @@
+import { describe, expect, it } from 'vitest'
+
+import { type DependencyEdge, type SchedulableIssue } from '../scheduler/scheduler-v0.js'
+import { makespan, scheduleWithCapacity, type Worker } from '../scheduler/scheduler-capacity-v0.js'
+import { capacityPerWorkday, parseCapacityConfig } from './capacity-v0.js'
+
+describe('capacity model', () => {
+  it('capacityPerWorkday = focusFactor × allocation', () => {
+    expect(capacityPerWorkday({ person: 'a', focusFactor: 0.8, allocation: 0.5 })).toBeCloseTo(0.4)
+  })
+
+  it('parses + clamps config, drops invalid members', () => {
+    const members = parseCapacityConfig({
+      members: [
+        { person: 'christian', focusFactor: 0.8, allocation: 1 },
+        { person: 'ak', focusFactor: 1.5, allocation: -1 }, // clamps to 1 / 0 → zero capacity → dropped
+        { focusFactor: 0.8 }, // no person → dropped
+      ],
+    })
+    expect(members).toEqual([{ person: 'christian', focusFactor: 0.8, allocation: 1 }])
+  })
+
+  it('returns [] for a non-array/absent members field', () => {
+    expect(parseCapacityConfig({})).toEqual([])
+    expect(parseCapacityConfig({ members: 'nope' })).toEqual([])
+  })
+})
+
+function issue(number: number, over: Partial<SchedulableIssue> = {}): SchedulableIssue {
+  return { number, title: `#${number}`, labels: [], estimateDays: 4, priority: 2, ...over }
+}
+
+describe('scheduleWithCapacity', () => {
+  it('with no workers, falls back to the single serial plan', () => {
+    const issues = [issue(1), issue(2)]
+    const plan = scheduleWithCapacity(issues, [], [])
+    expect(makespan(plan)).toBe(8) // 4 + 4 serial
+  })
+
+  it('parallelizes independent work across lanes (makespan shrinks)', () => {
+    const issues = [issue(1), issue(2), issue(3), issue(4)] // 4×4d = 16d serial
+    const one: Worker[] = [{ person: 'a', speed: 1 }]
+    const two: Worker[] = [
+      { person: 'a', speed: 1 },
+      { person: 'b', speed: 1 },
+    ]
+    expect(makespan(scheduleWithCapacity(issues, [], one))).toBe(16)
+    expect(makespan(scheduleWithCapacity(issues, [], two))).toBe(8) // two lanes → half
+  })
+
+  it('scales duration by a lane speed (slower lane takes longer)', () => {
+    const plan = scheduleWithCapacity([issue(1, { estimateDays: 4 })], [], [{ person: 'a', speed: 0.5 }])
+    expect(plan.items[0].durationDays).toBe(8) // 4 / 0.5
+    expect(plan.items[0].worker).toBe('a')
+  })
+
+  it('routes an issue to its assignee lane', () => {
+    const issues = [issue(1, { assignee: 'ak' }), issue(2, { assignee: 'sm' })]
+    const workers: Worker[] = [
+      { person: 'ak', speed: 1 },
+      { person: 'sm', speed: 1 },
+    ]
+    const plan = scheduleWithCapacity(issues, [], workers)
+    const byN = Object.fromEntries(plan.items.map((i) => [i.number, i]))
+    expect(byN[1].worker).toBe('ak')
+    expect(byN[2].worker).toBe('sm')
+    // both start at 0 (different lanes) → parallel
+    expect(byN[1].startDay).toBe(0)
+    expect(byN[2].startDay).toBe(0)
+  })
+
+  it('adds a lane for an assignee not in the config (mean speed)', () => {
+    const plan = scheduleWithCapacity([issue(1, { assignee: 'newbie' })], [], [{ person: 'a', speed: 0.5 }])
+    expect(plan.items[0].worker).toBe('newbie')
+    expect(plan.items[0].durationDays).toBe(8) // mean speed 0.5 → 4/0.5
+  })
+
+  it('respects dependencies across lanes (a blocker finishes before its dependent starts)', () => {
+    const edges: DependencyEdge[] = [{ issue: 2, dependsOn: 1 }]
+    const workers: Worker[] = [
+      { person: 'a', speed: 1 },
+      { person: 'b', speed: 1 },
+    ]
+    const plan = scheduleWithCapacity([issue(1), issue(2)], edges, workers)
+    const byN = Object.fromEntries(plan.items.map((i) => [i.number, i]))
+    expect(byN[2].startDay).toBeGreaterThanOrEqual(byN[1].endDay) // #2 waits for #1 even on another lane
+  })
+})

packages/core/src/capacity/capacity-v0.ts

@@ -0,0 +1,49 @@
+/**
+ * Capacity model (#8). A member's throughput is `focusFactor × allocation` =
+ * ideal person-days of project work delivered per calendar working day. The
+ * scheduler treats each member as a lane whose task durations are scaled by that
+ * rate (a 0.5-capacity person takes twice as long on an est/2d task). Config
+ * lives in the pm-state repo (`capacity/members.yaml/json`, D4); estimates are
+ * in ideal person-days (pm-state.md).
+ */
+
+export interface CapacityMember {
+  /** gitea login. */
+  person: string
+  /** Productive fraction of a working day (0..1). */
+  focusFactor: number
+  /** Fraction of that allocated to this project (0..1). */
+  allocation: number
+}
+
+/** Ideal person-days delivered per calendar working day. */
+export function capacityPerWorkday(m: CapacityMember): number {
+  return m.focusFactor * m.allocation
+}
+
+function clamp01(n: unknown, fallback: number): number {
+  return typeof n === 'number' && Number.isFinite(n) ? Math.min(1, Math.max(0, n)) : fallback
+}
+
+/**
+ * Validate a raw capacity config (`{ members: [...] }`) into members. Unknown
+ * shapes degrade to `[]` (→ the scheduler falls back to a single worker), never
+ * throw. focusFactor/allocation clamp to [0,1]; a member without a person is dropped.
+ */
+export function parseCapacityConfig(raw: unknown): CapacityMember[] {
+  const list = (raw as { members?: unknown })?.members
+  if (!Array.isArray(list)) return []
+  const out: CapacityMember[] = []
+  for (const m of list) {
+    const r = (m ?? {}) as Record<string, unknown>
+    const person = typeof r.person === 'string' ? r.person.trim() : ''
+    if (!person) continue
+    const member: CapacityMember = {
+      person,
+      focusFactor: clamp01(r.focusFactor, 0.8),
+      allocation: clamp01(r.allocation, 1),
+    }
+    if (capacityPerWorkday(member) > 0) out.push(member)
+  }
+  return out
+}

packages/core/src/forecast/forecast-v0.test.ts

@@ -120,4 +120,20 @@ describe('forecast', () => {
     expect(fitted.coldStart).toBe(false)
     expect(fitted.p50Day).toBeLessThan(priors.p50Day)
   })
+
+  it('capacity lanes parallelize the sim — the makespan shrinks with more workers', () => {
+    const independent = [issue(1), issue(2), issue(3), issue(4)] // no deps → fully parallelizable
+    const serial = forecast(independent, [], { trials: 2000, seed: 7 })
+    const twoLanes = forecast(independent, [], {
+      trials: 2000,
+      seed: 7,
+      workers: [
+        { person: 'a', speed: 1 },
+        { person: 'b', speed: 1 },
+      ],
+    })
+    // two equal lanes ≈ half the serial landing (independent work)
+    expect(twoLanes.p80Day).toBeLessThan(serial.p80Day)
+    expect(twoLanes.p80Day).toBeLessThan(serial.p80Day * 0.7)
+  })
 })

packages/core/src/forecast/forecast-v0.ts

@@ -18,6 +18,7 @@ import {
   schedule,
   type SchedulableIssue,
 } from '../scheduler/scheduler-v0.js'
+import { type LaneInputs, layoutOnLanes, resolveLanes, type Worker } from '../scheduler/scheduler-capacity-v0.js'
 
 export interface LognormalPrior {
   /** Median log-ratio: sampled median duration = estimate * e^mu. */
@@ -76,6 +77,12 @@ export interface ForecastOptions {
    * the sim; otherwise the code-resident cold-start priors do.
    */
   model?: DurationModel
+  /**
+   * Capacity lanes. When provided, each trial lays sampled durations across the
+   * lanes (parallel) instead of a single serial worker — the makespan shrinks
+   * toward the critical path. Empty/absent → single serial worker.
+   */
+  workers?: Worker[]
 }
 
 /** Resolve the lognormal params for an estimate, preferring a fitted model. */
@@ -158,15 +165,32 @@ export function forecast(
   const priors = order.map((it) => durationParams(it.durationDays, options.model))
   const rng = mulberry32(seed)
 
-  // endByRank[k][t] = working day the (k+1)-th scheduled issue completes on trial t.
+  // Capacity lanes, when configured — the per-trial layout goes parallel.
+  const laneInputs: LaneInputs = {
+    order: order.map((it) => it.number),
+    blockedBy: new Map(order.map((it) => [it.number, it.blockedBy])),
+    assignee: new Map(issues.map((i) => [i.number, i.assignee ?? null])),
+  }
+  const lanes =
+    options.workers && options.workers.length ? resolveLanes(options.workers, [...laneInputs.assignee.values()]) : null
+
+  // endByRank[k][t] = working day the (k+1)-th issue *to finish* completes on trial t.
   const endByRank: number[][] = Array.from({ length: n }, () => new Array<number>(trials))
   for (let t = 0; t < trials; t++) {
-    let cursor = 0
+    const sampled = order.map((it, k) => it.durationDays * Math.exp(priors[k].mu + priors[k].sigma * standardNormal(rng)))
-    for (let k = 0; k < n; k++) {
+    if (lanes) {
-      const p = priors[k]
+      // parallel: lay the sampled durations across lanes, then sort finish days
-      const sampled = order[k].durationDays * Math.exp(p.mu + p.sigma * standardNormal(rng))
+      const byNumber = new Map(order.map((it, k) => [it.number, sampled[k]]))
-      cursor += sampled
+      const { finishAt } = layoutOnLanes(laneInputs, lanes, (num) => byNumber.get(num)!)
-      endByRank[k][t] = cursor
+      const finishes = order.map((it) => finishAt.get(it.number)!).sort((a, b) => a - b)
+      for (let k = 0; k < n; k++) endByRank[k][t] = finishes[k]
+    } else {
+      // single serial worker: cumulative sum (already sorted ascending)
+      let cursor = 0
+      for (let k = 0; k < n; k++) {
+        cursor += sampled[k]
+        endByRank[k][t] = cursor
+      }
     }
   }
 

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

@@ -49,6 +49,12 @@ export type {
   SchedulePlan,
 } from './scheduler/scheduler-v0.js'
 
+export { makespan, scheduleWithCapacity } from './scheduler/scheduler-capacity-v0.js'
+export type { Worker } from './scheduler/scheduler-capacity-v0.js'
+
+export { capacityPerWorkday, parseCapacityConfig } from './capacity/capacity-v0.js'
+export type { CapacityMember } from './capacity/capacity-v0.js'
+
 export {
   COLD_START_PRIORS,
   durationParams,
@@ -80,7 +86,7 @@ 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'

packages/core/src/scheduler/scheduler-capacity-v0.ts

@@ -0,0 +1,119 @@
+/**
+ * Capacity-aware scheduler (#8). Reuses the single-worker scheduler's topological
+ * order + critical-path marking, then re-lays the work across lanes: an issue
+ * runs on its assignee's lane (or the earliest-free lane when unassigned), its
+ * duration scaled by that lane's speed (ideal person-days/workday). Makespan
+ * shrinks toward the critical path as lanes are added. Deterministic; falls back
+ * to the single serial worker when no capacity is configured. The lane layout is
+ * factored so the Monte Carlo forecast reuses it per trial with sampled durations.
+ */
+
+import {
+  type DependencyEdge,
+  schedule,
+  type SchedulableIssue,
+  type SchedulePlan,
+} from './scheduler-v0.js'
+
+/** A scheduling lane: a person and their throughput (ideal person-days / workday). */
+export interface Worker {
+  person: string
+  speed: number
+}
+
+/** The topological order + per-issue relations the layout needs (from the base plan). */
+export interface LaneInputs {
+  order: number[]
+  blockedBy: Map<number, number[]>
+  assignee: Map<number, string | null>
+}
+
+/** Ensure a lane exists for every assignee; unconfigured assignees get the mean speed. */
+export function resolveLanes(workers: Worker[], assignees: (string | null)[]): Worker[] {
+  const lanes = [...workers]
+  const known = new Set(lanes.map((w) => w.person))
+  const meanSpeed = lanes.length ? lanes.reduce((s, w) => s + w.speed, 0) / lanes.length : 1
+  for (const a of assignees) {
+    if (a && !known.has(a)) {
+      lanes.push({ person: a, speed: meanSpeed })
+      known.add(a)
+    }
+  }
+  return lanes
+}
+
+function pickWorker(lanes: Worker[], freeAt: Map<string, number>, assignee: string | null | undefined): Worker {
+  if (assignee) {
+    const own = lanes.find((w) => w.person === assignee)
+    if (own) return own
+  }
+  let best = lanes[0]
+  for (const w of lanes) if (freeAt.get(w.person)! < freeAt.get(best.person)!) best = w
+  return best
+}
+
+/**
+ * Lay a topologically-ordered set out across lanes. `duration(n)` supplies each
+ * issue's duration for this layout (estimate, or a sampled value in a MC trial).
+ * Returns each issue's finish day + the lane it ran on. Order guarantees a
+ * dependency is always laid out before its dependents.
+ */
+export function layoutOnLanes(
+  inputs: LaneInputs,
+  lanes: Worker[],
+  duration: (n: number) => number,
+): { finishAt: Map<number, number>; startAt: Map<number, number>; laneOf: Map<number, string> } {
+  const freeAt = new Map(lanes.map((w) => [w.person, 0]))
+  const finishAt = new Map<number, number>()
+  const startAt = new Map<number, number>()
+  const laneOf = new Map<number, string>()
+  for (const n of inputs.order) {
+    const worker = pickWorker(lanes, freeAt, inputs.assignee.get(n))
+    const blockers = inputs.blockedBy.get(n) ?? []
+    const depFinish = blockers.length ? Math.max(...blockers.map((d) => finishAt.get(d) ?? 0)) : 0
+    const start = Math.max(freeAt.get(worker.person)!, depFinish)
+    const end = start + duration(n) / worker.speed
+    startAt.set(n, start)
+    finishAt.set(n, end)
+    laneOf.set(n, worker.person)
+    freeAt.set(worker.person, end)
+  }
+  return { finishAt, startAt, laneOf }
+}
+
+/**
+ * Capacity-aware plan. An empty `workers` means no capacity is configured → the
+ * single-worker plan verbatim.
+ */
+export function scheduleWithCapacity(
+  issues: SchedulableIssue[],
+  edges: DependencyEdge[],
+  workers: Worker[],
+): SchedulePlan {
+  if (workers.length === 0) return schedule(issues, edges)
+  const base = schedule(issues, edges)
+  if (base.cycle || base.items.length === 0) return base
+
+  const inputs: LaneInputs = {
+    order: base.items.map((it) => it.number),
+    blockedBy: new Map(base.items.map((it) => [it.number, it.blockedBy])),
+    assignee: new Map(issues.map((i) => [i.number, i.assignee ?? null])),
+  }
+  const lanes = resolveLanes(workers, [...inputs.assignee.values()])
+  const durationOf = new Map(base.items.map((it) => [it.number, it.durationDays]))
+  const { finishAt, startAt, laneOf } = layoutOnLanes(inputs, lanes, (n) => durationOf.get(n)!)
+
+  const items = base.items.map((it) => ({
+    ...it,
+    startDay: startAt.get(it.number)!,
+    endDay: finishAt.get(it.number)!,
+    durationDays: finishAt.get(it.number)! - startAt.get(it.number)!,
+    worker: laneOf.get(it.number),
+  }))
+  return { items, cycle: null }
+}
+
+/** Makespan (last finish) of a plan — the project's landing day. */
+export function makespan(plan: SchedulePlan): number {
+  return plan.items.reduce((m, it) => Math.max(m, it.endDay), 0)
+}

packages/core/src/scheduler/scheduler-v0.ts

@@ -21,6 +21,8 @@ export interface SchedulableIssue {
   estimateDays: number | null
   /** 1 (most urgent) … 4; null when unset. */
   priority: number | null
+  /** gitea assignee login, for capacity-aware lane routing; optional. */
+  assignee?: string | null
 }
 
 export interface DependencyEdge {
@@ -45,6 +47,8 @@ export interface ScheduledItem {
   /** On a longest-duration dependency chain. */
   critical: boolean
   rationale: string
+  /** Lane it's scheduled on, when capacity-aware; absent for the single-worker plan. */
+  worker?: string
 }
 
 export interface SchedulePlan {