@craftedxp/voice-js 0.4.1 → 0.5.4

package/dist/browser.js

@@ -29,6 +29,8 @@ __export(browser_exports, {
   createProtocolState: () => createProtocolState,
   createReconnectingWebSocket: () => createReconnectingWebSocket,
   handleServerMessage: () => handleServerMessage,
+  joinRoom: () => joinRoom,
+  parseIncomingCall: () => parseIncomingCall,
 })
 module.exports = __toCommonJS(browser_exports)
 
@@ -958,8 +960,45 @@ async function createWebRtcCall(opts) {
   const iceUrl = gateway
     ? `${gateway}/webrtc/ice?token=${encodeURIComponent(opts.token)}`
     : `${opts.apiBase}/v1/agents/${encodeURIComponent(opts.agentId)}/webrtc/ice?token=${encodeURIComponent(opts.token)}`
+  const teardown = () => {
+    if (ended) return
+    ended = true
+    try {
+      mic.getTracks().forEach((t) => t.stop())
+    } catch {}
+    try {
+      pc.close()
+    } catch {}
+    try {
+      audioEl.remove()
+    } catch {}
+    fireState('ended')
+    opts.onEnd?.()
+  }
+  let callId = null
+  const pendingCandidates = []
+  const postCandidate = (candidate) => {
+    void fetch(iceUrl, {
+      method: 'POST',
+      headers: { 'content-type': 'application/json' },
+      body: JSON.stringify({ callId, candidate }),
+    }).catch(() => {})
+  }
+  pc.onicecandidate = (e) => {
+    if (!e.candidate) return
+    if (callId) postCandidate(e.candidate)
+    else pendingCandidates.push(e.candidate)
+  }
+  pc.onconnectionstatechange = () => {
+    const s = pc.connectionState
+    if (s === 'connected') fireState('listening')
+    if (s === 'failed' || s === 'disconnected') {
+      opts.onError?.({ code: 'socket_error', message: `webrtc connection ${s}` })
+      teardown()
+    }
+    if (s === 'closed' && !ended) teardown()
+  }
   await pc.setLocalDescription(await pc.createOffer())
-  let callId
   try {
     const offerRes = await fetch(offerUrl, {
       method: 'POST',
@@ -978,6 +1017,7 @@ async function createWebRtcCall(opts) {
     const body = await offerRes.json()
     callId = body.callId
     await pc.setRemoteDescription({ type: 'answer', sdp: body.sdp })
+    while (pendingCandidates.length > 0) postCandidate(pendingCandidates.shift())
   } catch (err) {
     if (!ended) {
       opts.onError?.({
@@ -991,38 +1031,6 @@ async function createWebRtcCall(opts) {
     }
     throw err
   }
-  pc.onicecandidate = (e) => {
-    if (!e.candidate) return
-    void fetch(iceUrl, {
-      method: 'POST',
-      headers: { 'content-type': 'application/json' },
-      body: JSON.stringify({ callId, candidate: e.candidate }),
-    }).catch(() => {})
-  }
-  pc.onconnectionstatechange = () => {
-    const s = pc.connectionState
-    if (s === 'connected') fireState('listening')
-    if (s === 'failed' || s === 'disconnected') {
-      opts.onError?.({ code: 'socket_error', message: `webrtc connection ${s}` })
-      teardown()
-    }
-    if (s === 'closed' && !ended) teardown()
-  }
-  const teardown = () => {
-    if (ended) return
-    ended = true
-    try {
-      mic.getTracks().forEach((t) => t.stop())
-    } catch {}
-    try {
-      pc.close()
-    } catch {}
-    try {
-      audioEl.remove()
-    } catch {}
-    fireState('ended')
-    opts.onEnd?.()
-  }
   return {
     get state() {
       return proto.state
@@ -1047,6 +1055,199 @@ async function createWebRtcCall(opts) {
   }
 }
 
+// src/room.ts
+var import_livekit_client = require('livekit-client')
+
+// src/roomProtocol.ts
+var SYSTEM_TOPIC = 'system'
+var TRANSCRIPT_TOPIC = 'transcript'
+var decodeSystem = (bytes) => {
+  try {
+    const v = JSON.parse(new TextDecoder().decode(bytes))
+    if (v && typeof v.kind === 'string') return v
+    return null
+  } catch {
+    return null
+  }
+}
+var decodeTranscript = (bytes) => {
+  try {
+    const v = JSON.parse(new TextDecoder().decode(bytes))
+    if (v && v.kind === 'partial') return v
+    return null
+  } catch {
+    return null
+  }
+}
+
+// src/room.ts
+var identityToPid = (identity) =>
+  identity.startsWith('guest:') ? identity.slice('guest:'.length) : identity
+var joinRoom = async (opts) => {
+  const exchangeUrl = `${opts.apiBase.replace(/\/+$/, '')}/v1/rooms/${encodeURIComponent(
+    opts.roomId,
+  )}/join`
+  const exchangeRes = await fetch(exchangeUrl, {
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify({ code: opts.joinCode, name: opts.name }),
+  })
+  if (!exchangeRes.ok) {
+    const err = await exchangeRes.json().catch(() => ({}))
+    throw new Error(err.error?.code ?? `join_failed_${exchangeRes.status}`)
+  }
+  const exchange = await exchangeRes.json()
+  const handlers = /* @__PURE__ */ new Map()
+  const emit = (e, payload) => {
+    handlers.get(e)?.forEach((h) => {
+      try {
+        h(payload)
+      } catch {}
+    })
+  }
+  const room = new import_livekit_client.Room({ adaptiveStream: true, dynacast: true })
+  room.on(import_livekit_client.RoomEvent.ParticipantConnected, (p) =>
+    emit('participant.joined', {
+      participantId: identityToPid(p.identity),
+      name: p.name ?? '',
+    }),
+  )
+  room.on(import_livekit_client.RoomEvent.ParticipantDisconnected, (p) =>
+    emit('participant.left', {
+      participantId: identityToPid(p.identity),
+      name: p.name ?? '',
+    }),
+  )
+  room.on(import_livekit_client.RoomEvent.Disconnected, () => emit('room.ended', void 0))
+  room.on(import_livekit_client.RoomEvent.DataReceived, (data, _participant, _kind, topic) => {
+    if (topic === SYSTEM_TOPIC) {
+      const m = decodeSystem(data)
+      if (m) emit('system.message', m)
+    } else if (topic === TRANSCRIPT_TOPIC) {
+      const m = decodeTranscript(data)
+      if (m) emit('transcript.partial', m)
+    }
+  })
+  const trackKind = (t) => (t.kind === import_livekit_client.Track.Kind.Video ? 'video' : 'audio')
+  const trackSource = (s) => {
+    switch (s) {
+      case import_livekit_client.Track.Source.Camera:
+        return 'camera'
+      case import_livekit_client.Track.Source.Microphone:
+        return 'microphone'
+      case import_livekit_client.Track.Source.ScreenShare:
+        return 'screen_share'
+      case import_livekit_client.Track.Source.ScreenShareAudio:
+        return 'screen_share_audio'
+      default:
+        return 'unknown'
+    }
+  }
+  room.on(import_livekit_client.RoomEvent.TrackSubscribed, (track, pub, participant) =>
+    emit('track.subscribed', {
+      participantId: identityToPid(participant.identity),
+      kind: trackKind(track),
+      source: trackSource(pub.source),
+      track,
+    }),
+  )
+  room.on(import_livekit_client.RoomEvent.TrackUnsubscribed, (track, pub, participant) =>
+    emit('track.unsubscribed', {
+      participantId: identityToPid(participant.identity),
+      kind: trackKind(track),
+      source: trackSource(pub.source),
+      track,
+    }),
+  )
+  room.on(import_livekit_client.RoomEvent.ActiveSpeakersChanged, (speakers) =>
+    emit(
+      'active.speakers',
+      speakers.map((p) => identityToPid(p.identity)),
+    ),
+  )
+  await room.connect(exchange.livekit.url, exchange.livekit.token)
+  return {
+    participantId: exchange.participantId,
+    get participants() {
+      return [...room.remoteParticipants.values()].map((p) => ({
+        participantId: identityToPid(p.identity),
+        name: p.name ?? '',
+      }))
+    },
+    on(event, handler) {
+      const set = handlers.get(event) ?? /* @__PURE__ */ new Set()
+      set.add(handler)
+      handlers.set(event, set)
+    },
+    publishMic: async () => {
+      await room.localParticipant.setMicrophoneEnabled(true)
+    },
+    publishCamera: async () => {
+      await room.localParticipant.setCameraEnabled(true)
+    },
+    setMicEnabled: async (on) => {
+      await room.localParticipant.setMicrophoneEnabled(on)
+    },
+    setCameraEnabled: async (on) => {
+      await room.localParticipant.setCameraEnabled(on)
+    },
+    isMicEnabled: () => room.localParticipant.isMicrophoneEnabled,
+    isCameraEnabled: () => room.localParticipant.isCameraEnabled,
+    getLocalCameraTrack: () =>
+      room.localParticipant.getTrackPublication(import_livekit_client.Track.Source.Camera)
+        ?.videoTrack ?? null,
+    getRemoteTracks: () => {
+      const out = []
+      for (const p of room.remoteParticipants.values()) {
+        for (const pub of p.trackPublications.values()) {
+          const track = pub.track
+          if (!track) continue
+          out.push({
+            participantId: identityToPid(p.identity),
+            kind: trackKind(track),
+            source: trackSource(pub.source),
+            track,
+          })
+        }
+      }
+      return out
+    },
+    setScreenShareEnabled: async (on, opts2) => {
+      await room.localParticipant.setScreenShareEnabled(on, { audio: opts2?.audio ?? false })
+    },
+    isScreenShareEnabled: () => room.localParticipant.isScreenShareEnabled,
+    getLocalScreenTrack: () =>
+      room.localParticipant.getTrackPublication(import_livekit_client.Track.Source.ScreenShare)
+        ?.videoTrack ?? null,
+    leave: async () => {
+      await room.disconnect()
+    },
+  }
+}
+
+// src/incomingCall.ts
+var parseIncomingCall = (raw) => {
+  if (typeof raw !== 'object' || raw === null) {
+    throw new Error('parseIncomingCall: payload must be an object')
+  }
+  const p = raw
+  if (typeof p.token !== 'string' || !p.token.startsWith('ct_')) {
+    throw new Error('parseIncomingCall: missing or invalid `token` (expected a ct_ string)')
+  }
+  if (typeof p.agentId !== 'string' || p.agentId.length === 0) {
+    throw new Error('parseIncomingCall: missing `agentId`')
+  }
+  const transport = p.transport === 'webrtc' ? 'webrtc' : 'ws'
+  const out = { token: p.token, agentId: p.agentId, transport }
+  if (transport === 'webrtc' && typeof p.webrtcGatewayBase === 'string') {
+    out.webrtcGatewayBase = p.webrtcGatewayBase
+  }
+  if (typeof p.expiresAt === 'number') out.expiresAt = p.expiresAt
+  if (typeof p.agentName === 'string') out.agentName = p.agentName
+  if (typeof p.agentAvatarUrl === 'string') out.agentAvatarUrl = p.agentAvatarUrl
+  return out
+}
+
 // src/browser.ts
 var browserWsFactory = (url) => new globalThis.WebSocket(url)
 var BrowserVoiceFactory = class {
@@ -1106,6 +1307,14 @@ var BrowserVoiceFactory = class {
       await client.start()
       return client
     }
+    // Multi-party rooms (Phase 7 video).
+    //
+    // The guest's browser calls this with the roomId + joinCode it parsed
+    // out of the invite link. The SDK exchanges the code for a LiveKit
+    // JWT against `${apiBase}/v1/rooms/:roomId/join` (an AUTH-EXEMPT
+    // endpoint — the opaque code is the only credential), then connects
+    // to LiveKit and returns a typed event surface.
+    this.joinRoom = (opts) => joinRoom({ apiBase: this.config.apiBase, ...opts })
     this.config = config
   }
 }
@@ -1122,5 +1331,7 @@ function configureVoiceClient(config) {
     createProtocolState,
     createReconnectingWebSocket,
     handleServerMessage,
+    joinRoom,
+    parseIncomingCall,
   })
 //# sourceMappingURL=browser.js.map