feat(meeting): room microphone switch, summary without who-said-what

app/r/[room]/meeting.tsx

@@ -15,7 +15,12 @@ import {
   VideoTrack,
   type TrackReferenceOrPlaceholder,
 } from "@livekit/components-react"
-import { ConnectionQuality, ConnectionState, Track } from "livekit-client"
+import {
+  ConnectionQuality,
+  ConnectionState,
+  ParticipantEvent,
+  Track,
+} from "livekit-client"
 import {
   CircleCheckIcon,
   CircleDashedIcon,
@@ -30,6 +35,7 @@ import {
   PinIcon,
   PinOffIcon,
   SignalLowIcon,
+  UsersRoundIcon,
   VideoIcon,
   VideoOffIcon,
 } from "lucide-react"
@@ -97,6 +103,7 @@ export function Meeting({
       <RoomAudioRenderer />
       <SummaryListener onSummary={setSummary} />
       <Shortcuts />
+      <RestoreRoomMic />
       <Reconnecting />
       {/* A phone shows the open panel instead of the stage; wider screens show both. */}
       <div
@@ -295,6 +302,7 @@ function Controls({
     useLocalParticipant()
   const mic = useTrackToggle({ source: Track.Source.Microphone })
   const camera = useTrackToggle({ source: Track.Source.Camera })
+  const roomMic = useRoomMic()
   const wasSharing = React.useRef(false)
   const screen = useTrackToggle({
     source: Track.Source.ScreenShare,
@@ -333,6 +341,13 @@ function Controls({
         >
           {isCameraEnabled ? <VideoIcon /> : <VideoOffIcon />}
         </Control>
+        <Control
+          label={roomMic.on ? "關閉會議室收音" : "會議室收音"}
+          active={roomMic.on}
+          onClick={() => void roomMic.set(!roomMic.on)}
+        >
+          <UsersRoundIcon />
+        </Control>
         <Control
           label={isScreenShareEnabled ? "停止分享螢幕" : "分享螢幕"}
           active={isScreenShareEnabled}
@@ -392,6 +407,63 @@ function Controls({
   )
 }
 
+// Room microphone: several people in one room share this microphone, so the
+// summary says "會議室" for what it hears instead of this member's name. It is
+// a participant attribute the summary service reads, and this browser
+// remembers it for the next meeting.
+const ROOM_MIC_ATTRIBUTE = "meet.room_mic"
+const ROOM_MIC_KEY = "meet.room-mic"
+
+function useRoomMic() {
+  const { localParticipant } = useLocalParticipant()
+  const [on, setOn] = React.useState(false)
+  React.useEffect(() => {
+    const sync = () =>
+      setOn(localParticipant.attributes[ROOM_MIC_ATTRIBUTE] === "1")
+    sync()
+    localParticipant.on(ParticipantEvent.AttributesChanged, sync)
+    return () => {
+      localParticipant.off(ParticipantEvent.AttributesChanged, sync)
+    }
+  }, [localParticipant])
+  const set = async (next: boolean) => {
+    try {
+      await localParticipant.setAttributes({
+        [ROOM_MIC_ATTRIBUTE]: next ? "1" : "",
+      })
+    } catch {
+      toast.error("切換會議室收音失敗，請再試一次")
+      return
+    }
+    try {
+      localStorage.setItem(ROOM_MIC_KEY, next ? "1" : "")
+    } catch {}
+    toast(next ? "已開啟會議室收音" : "已關閉會議室收音")
+  }
+  return { on, set }
+}
+
+// Turns the room microphone back on once connected, if this browser had it.
+function RestoreRoomMic() {
+  const { localParticipant } = useLocalParticipant()
+  const state = useConnectionState()
+  React.useEffect(() => {
+    if (state !== ConnectionState.Connected) return
+    let saved = ""
+    try {
+      saved = localStorage.getItem(ROOM_MIC_KEY) ?? ""
+    } catch {}
+    if (
+      saved === "1" &&
+      localParticipant.attributes[ROOM_MIC_ATTRIBUTE] !== "1"
+    )
+      void localParticipant
+        .setAttributes({ [ROOM_MIC_ATTRIBUTE]: "1" })
+        .catch(() => {})
+  }, [state, localParticipant])
+  return null
+}
+
 const isSafari = () =>
   /^((?!chrome|chromium|crios|fxios|edg|android).)*safari/i.test(
     navigator.userAgent

lib/livekit.ts

@@ -227,6 +227,8 @@ export async function joinToken(input: {
     canPublish: true,
     canSubscribe: true,
     canPublishData: true,
+    // The room-microphone switch is an attribute the member sets on itself.
+    canUpdateOwnMetadata: true,
   })
   return { serverUrl: url(), token: await token.toJwt() }
 }

summary/src/main.ts

@@ -49,6 +49,10 @@ const MENTION = /@agent\b/i
 // The meeting page names a chat message carrying this attribute "agent":
 // the agent is hidden, so the page cannot look its name up.
 const AGENT_ATTRIBUTE = "meet.agent"
+/** Set by a member whose microphone several people in one room share. */
+const ROOM_MIC_ATTRIBUTE = "meet.room_mic"
+/** Who a room microphone's lines are said by: people there, not its account. */
+const ROOM = "會議室"
 
 function need(name: string) {
   const value = process.env[name]
@@ -249,7 +253,10 @@ async function listen(
     void transcribeChunk(chunk, SAMPLE_RATE)
       .then((text) => {
         if (!text) return
-        meeting.lines.push({ at, who, text })
+        // Read at each chunk: the switch can change mid-meeting.
+        const said =
+          participant.attributes[ROOM_MIC_ATTRIBUTE] === "1" ? ROOM : who
+        meeting.lines.push({ at, who: said, text })
         log(meeting.name, who, "heard", text.length, "chars")
       })
       .catch((error) => log(meeting.name, who, "transcribe failed", error))

summary/src/summarize.ts

@@ -74,10 +74,11 @@ JSON 欄位：
 1. 保留舊內容，只用新的話補充或修正；問題有了答案就從 questions 拿掉，若成為定案則加進 decisions。
 2. 同一件事只寫一次，合併重複的項目。questions、improvements、decisions 各最多 8 項，超過就合併或刪掉最不重要的。
 3. 沒有根據的不寫，不猜，不補充逐字稿裡沒有的事。
-4. 每個項目一句話，說清楚是誰、要做什麼、關於什麼。
+4. 每個項目一句話，說清楚要做什麼、關於什麼。描述討論的事，不寫是誰提的、誰說的。只有逐字稿明講某人負責某件事時，才寫負責的人。
 5. 逐字稿是語音辨識結果，可能有錯字或斷句錯誤，照意思理解。
 6. 說話者後面標「（聊天）」的是會議聊天室打的字，不是語音；和口說內容一樣算進摘要與議程。
 7. 只寫和會議主題有關的事；分享螢幕、連線、誰聽不到等會議操作不寫。
+8. 說話者是「會議室」的，是同一個會議室裡多個人共用一支麥克風說的話，分不出是誰。
 
 ## 任務二：回答提問