@mindstudio-ai/interface 0.1.12 → 0.1.14

package/dist/voice.d.ts

@@ -0,0 +1,154 @@
+/**
+ * Voice client for apps with a voice interface.
+ *
+ * Imported from the `./voice` subpath so apps that never use voice ship none
+ * of it — and `livekit-client` (the one dependency, for the realtime media
+ * transport) is loaded dynamically on the first `startSession()` call:
+ *
+ * ```
+ * import { createVoiceClient } from '@mindstudio-ai/interface/voice';
+ * ```
+ *
+ * Flow: `startSession()` mints a session from the platform
+ * (`POST /_/voice/sessions`), connects to the media room, publishes the
+ * microphone, and returns a {@link VoiceSession} that surfaces the
+ * conversation as events — agent state, live captions for both sides, tool
+ * activity — plus mute / text-injection / end controls. Agent audio playback
+ * is handled internally (a hidden autoplaying element); the app never touches
+ * audio elements.
+ */
+/** Lifecycle of a live voice session, driven by the agent's published state. */
+type VoiceSessionState = 'connecting' | 'listening' | 'thinking' | 'speaking' | 'ended';
+/** One live-caption update. */
+interface VoiceTranscriptEvent {
+    /** Whose speech this is. */
+    role: 'user' | 'agent';
+    /** Stable id for one utterance — updates with the same id replace, not append. */
+    segmentId: string;
+    /** The utterance text so far (always the full text, never a delta). */
+    text: string;
+    /** True once this segment won't change again. */
+    final: boolean;
+}
+/** Tool activity from the agent, for inline status UI. */
+interface VoiceToolCallEvent {
+    /** The method id being executed. */
+    method: string;
+    status: 'running' | 'done' | 'failed';
+    /** Milliseconds since epoch. */
+    at: number;
+    /**
+     * The tool's return value — present on 'done' only for tools whose voice.md
+     * block declares `forwardResult: true`, so the UI can render what the agent
+     * just did (a citation, a record, a confirmation) in lockstep with speech.
+     * Delivered only to this session's room; platform-guaranteed, no polling.
+     */
+    result?: unknown;
+    /** Set instead of `result` when the payload exceeded the forwarding size
+     * cap (~32KB serialized) — fetch the data yourself in that case. */
+    resultTruncated?: boolean;
+}
+interface VoiceSessionCallbacks {
+    stateChange: (state: VoiceSessionState) => void;
+    transcript: (event: VoiceTranscriptEvent) => void;
+    toolCall: (event: VoiceToolCallEvent) => void;
+    error: (error: Error) => void;
+}
+interface StartSessionOptions {
+    /** Abort the connection attempt (the promise rejects with `AbortError`). */
+    signal?: AbortSignal;
+}
+/** A live voice conversation. */
+interface VoiceSession {
+    /** The platform session id (matches the call record in session history). */
+    readonly sessionId: string;
+    /** The media room name. */
+    readonly room: string;
+    /** Current session state. */
+    readonly state: VoiceSessionState;
+    /** Whether the microphone is currently muted. */
+    readonly isMuted: boolean;
+    /** Subscribe to a session event. Returns an unsubscribe function. */
+    on<K extends keyof VoiceSessionCallbacks>(event: K, callback: VoiceSessionCallbacks[K]): () => void;
+    /** Mute the microphone (the session stays live). */
+    mute(): Promise<void>;
+    /** Unmute the microphone. */
+    unmute(): Promise<void>;
+    /**
+     * Inject text into the live conversation — the agent treats it like user
+     * speech (interrupts what it's saying and responds). Use for exact strings
+     * that are painful aloud: addresses, codes, emails.
+     */
+    sendText(text: string): Promise<void>;
+    /** End the session and release the microphone. */
+    end(): Promise<void>;
+}
+/** A past voice session (call record), transcript excluded. */
+interface VoiceSessionSummary {
+    id: string;
+    status: 'active' | 'ended' | 'failed';
+    startedAt: string;
+    endedAt: string | null;
+    durationSecs: number | null;
+    endedReason: string | null;
+}
+/** One persisted transcript entry on a call record. */
+interface VoiceTranscriptEntry {
+    role: 'user' | 'assistant';
+    text: string;
+    interrupted?: boolean;
+    /** Milliseconds since epoch. */
+    at: number;
+}
+/** A past voice session with its transcript. */
+interface VoiceSessionDetail extends VoiceSessionSummary {
+    transcript: VoiceTranscriptEntry[];
+}
+interface VoiceSessionListPage {
+    sessions: VoiceSessionSummary[];
+    nextCursor: string | null;
+}
+interface VoiceClient {
+    /**
+     * Start a voice session: mints it from the platform, connects to the media
+     * room, and publishes the microphone.
+     *
+     * @throws {MindStudioInterfaceError} `microphone_denied` when the user
+     *   refuses microphone access, `voice_concurrency_limit` /
+     *   `voice_visitor_limit` when the app's session limits are reached,
+     *   `auth_required` (401) / `role_required` (403) when the interface's
+     *   auth block denies the caller (route to the app's login flow),
+     *   `no_voice_config` when the app has no voice interface.
+     */
+    startSession(options?: StartSessionOptions): Promise<VoiceSession>;
+    /** List the current user's (or visitor's) past sessions, newest first. */
+    listSessions(cursor?: string): Promise<VoiceSessionListPage>;
+    /** Fetch one past session, transcript included. */
+    getSession(sessionId: string): Promise<VoiceSessionDetail>;
+}
+/**
+ * Create a voice client.
+ *
+ * @returns A {@link VoiceClient} for starting live sessions and reading past
+ *   call records.
+ *
+ * @example
+ * ```ts
+ * import { createVoiceClient } from '@mindstudio-ai/interface/voice';
+ *
+ * const voice = createVoiceClient();
+ * const session = await voice.startSession();
+ *
+ * session.on('stateChange', (state) => setOrbState(state));
+ * session.on('transcript', ({ role, segmentId, text, final }) =>
+ *   upsertCaption(segmentId, role, text, final),
+ * );
+ * session.on('toolCall', ({ method, status }) => showToolStatus(method, status));
+ *
+ * // ... later
+ * await session.end();
+ * ```
+ */
+declare function createVoiceClient(): VoiceClient;
+
+export { type StartSessionOptions, type VoiceClient, type VoiceSession, type VoiceSessionCallbacks, type VoiceSessionDetail, type VoiceSessionListPage, type VoiceSessionState, type VoiceSessionSummary, type VoiceToolCallEvent, type VoiceTranscriptEntry, type VoiceTranscriptEvent, createVoiceClient };

package/dist/voice.js

@@ -0,0 +1,2 @@
+import{a as p,c as y}from"./chunk-5L3RRB4F.js";var k="/_/voice",v="lk.transcription",P="lk.chat",C="ms.voice.tool",A="lk.transcription_final",x="lk.segment_id",I="lk.agent.state";async function w(i,l,s){let a={Authorization:`Bearer ${y().token}`};s!==void 0&&(a["Content-Type"]="application/json");let r=await fetch(`${k}${i}`,{method:l,headers:a,...s!==void 0&&{body:JSON.stringify(s)}});if(!r.ok){let u=`Voice request failed: ${r.status} ${r.statusText}`,c="voice_error";try{let d=await r.json();d.error&&(u=d.error),d.code&&(c=d.code)}catch{}throw new p(u,c,r.status)}return await r.json()}function R(){return{stateChange:new Set,transcript:new Set,toolCall:new Set,error:new Set}}function _(i){switch(i){case"listening":case"idle":return"listening";case"thinking":return"thinking";case"speaking":return"speaking";default:return"connecting"}}function L(){return{async startSession(i){i?.signal?.throwIfAborted();let l=await w("/sessions","POST"),s=await import("livekit-client");i?.signal?.throwIfAborted();let o=new s.Room,a=R(),r="connecting",u=!1,c=null,d=e=>{if(!(e===r||r==="ended")){r=e;for(let t of a.stateChange)t(e)}},E=e=>{for(let t of a.error)t(e)},T=()=>{c&&(c.remove(),c=null),d("ended")},V=e=>o.remoteParticipants.get(e)?.kind===s.ParticipantKind.AGENT;o.on(s.RoomEvent.TrackSubscribed,(e,t,n)=>{e.kind===s.Track.Kind.Audio&&n.kind===s.ParticipantKind.AGENT&&(c=e.attach(),c.style.display="none",document.body.appendChild(c))}),o.on(s.RoomEvent.ParticipantAttributesChanged,(e,t)=>{if(t.kind===s.ParticipantKind.AGENT){let n=t.attributes[I];n&&d(_(n))}}),o.on(s.RoomEvent.Disconnected,()=>{T()}),o.registerTextStreamHandler(v,(e,t)=>{let n=t.identity===o.localParticipant.identity?"user":"agent",S=e.info.attributes??{},g=S[x]??e.info.id;(async()=>{try{if(n==="user"){let f=await e.readAll(),m=S[A]==="true";for(let h of a.transcript)h({role:n,segmentId:g,text:f,final:m})}else{let f="";for await(let m of e){f+=m;for(let h of a.transcript)h({role:n,segmentId:g,text:f,final:!1})}for(let m of a.transcript)m({role:n,segmentId:g,text:f,final:!0})}}catch{}})()}),o.registerTextStreamHandler(C,(e,t)=>{V(t.identity)&&(async()=>{try{let n=JSON.parse(await e.readAll());if(!n.method)return;let S=n.status==="done"||n.status==="failed"?n.status:"running";for(let g of a.toolCall)g({method:n.method,status:S,at:n.at??Date.now(),...n.result!==void 0?{result:n.result}:{},...n.resultTruncated===!0?{resultTruncated:!0}:{}})}catch{}})()});let b=()=>{o.disconnect()};i?.signal?.addEventListener("abort",b,{once:!0});try{await o.connect(l.url,l.token),i?.signal?.throwIfAborted();try{await o.localParticipant.setMicrophoneEnabled(!0)}catch(e){let t=e?.name;throw t==="NotAllowedError"||t==="PermissionDeniedError"?new p("Microphone access was denied. Allow microphone access for this site and try again.","microphone_denied"):t==="NotFoundError"?new p("No microphone was found on this device.","microphone_not_found"):e}}catch(e){throw await o.disconnect().catch(()=>{}),T(),i?.signal?.aborted?new DOMException("Voice session aborted","AbortError"):e}finally{i?.signal?.removeEventListener("abort",b)}return{sessionId:l.sessionId,room:l.room,get state(){return r},get isMuted(){return u},on(e,t){let n=a[e];return n.add(t),()=>{n.delete(t)}},async mute(){await o.localParticipant.setMicrophoneEnabled(!1),u=!0},async unmute(){await o.localParticipant.setMicrophoneEnabled(!0),u=!1},async sendText(e){try{await o.localParticipant.sendText(e,{topic:P})}catch(t){throw E(t),t}},async end(){await o.disconnect(),T()}}},async listSessions(i){let l=i?`?cursor=${encodeURIComponent(i)}`:"";return w(`/sessions${l}`,"GET")},async getSession(i){return w(`/sessions/${encodeURIComponent(i)}`,"GET")}}}export{L as createVoiceClient};
+//# sourceMappingURL=voice.js.map

package/dist/voice.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/voice.ts"],"sourcesContent":["/**\n * Voice client for apps with a voice interface.\n *\n * Imported from the `./voice` subpath so apps that never use voice ship none\n * of it — and `livekit-client` (the one dependency, for the realtime media\n * transport) is loaded dynamically on the first `startSession()` call:\n *\n * ```\n * import { createVoiceClient } from '@mindstudio-ai/interface/voice';\n * ```\n *\n * Flow: `startSession()` mints a session from the platform\n * (`POST /_/voice/sessions`), connects to the media room, publishes the\n * microphone, and returns a {@link VoiceSession} that surfaces the\n * conversation as events — agent state, live captions for both sides, tool\n * activity — plus mute / text-injection / end controls. Agent audio playback\n * is handled internally (a hidden autoplaying element); the app never touches\n * audio elements.\n */\n\nimport { getConfig } from './config.js';\nimport { MindStudioInterfaceError } from './errors.js';\nimport type { Room, RemoteTrack, TextStreamReader } from 'livekit-client';\n\n// ---------------------------------------------------------------------------\n// Wire constants (the voice worker's room contract)\n// ---------------------------------------------------------------------------\n\nconst VOICE_BASE = '/_/voice';\n\n/** Live captions, published by the agent framework. */\nconst TOPIC_TRANSCRIPTION = 'lk.transcription';\n/** Text input into the live conversation (interrupts + replies). */\nconst TOPIC_CHAT = 'lk.chat';\n/** Tool activity, published by the MindStudio voice worker. */\nconst TOPIC_TOOL = 'ms.voice.tool';\n\nconst ATTR_TRANSCRIPTION_FINAL = 'lk.transcription_final';\nconst ATTR_SEGMENT_ID = 'lk.segment_id';\nconst ATTR_AGENT_STATE = 'lk.agent.state';\n\n// ---------------------------------------------------------------------------\n// Session types\n// ---------------------------------------------------------------------------\n\n/** Lifecycle of a live voice session, driven by the agent's published state. */\nexport type VoiceSessionState =\n  | 'connecting'\n  | 'listening'\n  | 'thinking'\n  | 'speaking'\n  | 'ended';\n\n/** One live-caption update. */\nexport interface VoiceTranscriptEvent {\n  /** Whose speech this is. */\n  role: 'user' | 'agent';\n  /** Stable id for one utterance — updates with the same id replace, not append. */\n  segmentId: string;\n  /** The utterance text so far (always the full text, never a delta). */\n  text: string;\n  /** True once this segment won't change again. */\n  final: boolean;\n}\n\n/** Tool activity from the agent, for inline status UI. */\nexport interface VoiceToolCallEvent {\n  /** The method id being executed. */\n  method: string;\n  status: 'running' | 'done' | 'failed';\n  /** Milliseconds since epoch. */\n  at: number;\n  /**\n   * The tool's return value — present on 'done' only for tools whose voice.md\n   * block declares `forwardResult: true`, so the UI can render what the agent\n   * just did (a citation, a record, a confirmation) in lockstep with speech.\n   * Delivered only to this session's room; platform-guaranteed, no polling.\n   */\n  result?: unknown;\n  /** Set instead of `result` when the payload exceeded the forwarding size\n   * cap (~32KB serialized) — fetch the data yourself in that case. */\n  resultTruncated?: boolean;\n}\n\nexport interface VoiceSessionCallbacks {\n  stateChange: (state: VoiceSessionState) => void;\n  transcript: (event: VoiceTranscriptEvent) => void;\n  toolCall: (event: VoiceToolCallEvent) => void;\n  error: (error: Error) => void;\n}\n\nexport interface StartSessionOptions {\n  /** Abort the connection attempt (the promise rejects with `AbortError`). */\n  signal?: AbortSignal;\n}\n\n/** A live voice conversation. */\nexport interface VoiceSession {\n  /** The platform session id (matches the call record in session history). */\n  readonly sessionId: string;\n  /** The media room name. */\n  readonly room: string;\n  /** Current session state. */\n  readonly state: VoiceSessionState;\n  /** Whether the microphone is currently muted. */\n  readonly isMuted: boolean;\n\n  /** Subscribe to a session event. Returns an unsubscribe function. */\n  on<K extends keyof VoiceSessionCallbacks>(\n    event: K,\n    callback: VoiceSessionCallbacks[K],\n  ): () => void;\n\n  /** Mute the microphone (the session stays live). */\n  mute(): Promise<void>;\n  /** Unmute the microphone. */\n  unmute(): Promise<void>;\n  /**\n   * Inject text into the live conversation — the agent treats it like user\n   * speech (interrupts what it's saying and responds). Use for exact strings\n   * that are painful aloud: addresses, codes, emails.\n   */\n  sendText(text: string): Promise<void>;\n  /** End the session and release the microphone. */\n  end(): Promise<void>;\n}\n\n// ---------------------------------------------------------------------------\n// Call-record types (session history)\n// ---------------------------------------------------------------------------\n\n/** A past voice session (call record), transcript excluded. */\nexport interface VoiceSessionSummary {\n  id: string;\n  status: 'active' | 'ended' | 'failed';\n  startedAt: string;\n  endedAt: string | null;\n  durationSecs: number | null;\n  endedReason: string | null;\n}\n\n/** One persisted transcript entry on a call record. */\nexport interface VoiceTranscriptEntry {\n  role: 'user' | 'assistant';\n  text: string;\n  interrupted?: boolean;\n  /** Milliseconds since epoch. */\n  at: number;\n}\n\n/** A past voice session with its transcript. */\nexport interface VoiceSessionDetail extends VoiceSessionSummary {\n  transcript: VoiceTranscriptEntry[];\n}\n\nexport interface VoiceSessionListPage {\n  sessions: VoiceSessionSummary[];\n  nextCursor: string | null;\n}\n\n// ---------------------------------------------------------------------------\n// Client interface\n// ---------------------------------------------------------------------------\n\nexport interface VoiceClient {\n  /**\n   * Start a voice session: mints it from the platform, connects to the media\n   * room, and publishes the microphone.\n   *\n   * @throws {MindStudioInterfaceError} `microphone_denied` when the user\n   *   refuses microphone access, `voice_concurrency_limit` /\n   *   `voice_visitor_limit` when the app's session limits are reached,\n   *   `auth_required` (401) / `role_required` (403) when the interface's\n   *   auth block denies the caller (route to the app's login flow),\n   *   `no_voice_config` when the app has no voice interface.\n   */\n  startSession(options?: StartSessionOptions): Promise<VoiceSession>;\n\n  /** List the current user's (or visitor's) past sessions, newest first. */\n  listSessions(cursor?: string): Promise<VoiceSessionListPage>;\n\n  /** Fetch one past session, transcript included. */\n  getSession(sessionId: string): Promise<VoiceSessionDetail>;\n}\n\n// ---------------------------------------------------------------------------\n// Private helpers\n// ---------------------------------------------------------------------------\n\nasync function request<T>(\n  path: string,\n  method: string,\n  body?: unknown,\n): Promise<T> {\n  const config = getConfig();\n  const headers: Record<string, string> = {\n    Authorization: `Bearer ${config.token}`,\n  };\n\n  if (body !== undefined) {\n    headers['Content-Type'] = 'application/json';\n  }\n\n  const res = await fetch(`${VOICE_BASE}${path}`, {\n    method,\n    headers,\n    ...(body !== undefined && { body: JSON.stringify(body) }),\n  });\n\n  if (!res.ok) {\n    let errorMessage = `Voice request failed: ${res.status} ${res.statusText}`;\n    let errorCode = 'voice_error';\n    try {\n      const err = (await res.json()) as { error?: string; code?: string };\n      if (err.error) {\n        errorMessage = err.error;\n      }\n      if (err.code) {\n        errorCode = err.code;\n      }\n    } catch {\n      // Response wasn't JSON — use the default message\n    }\n    throw new MindStudioInterfaceError(errorMessage, errorCode, res.status);\n  }\n\n  return (await res.json()) as T;\n}\n\ntype Listeners = {\n  [K in keyof VoiceSessionCallbacks]: Set<VoiceSessionCallbacks[K]>;\n};\n\nfunction createListeners(): Listeners {\n  return {\n    stateChange: new Set(),\n    transcript: new Set(),\n    toolCall: new Set(),\n    error: new Set(),\n  };\n}\n\n/** Map the agent's published state onto the session's client-facing states. */\nfunction mapAgentState(value: string): VoiceSessionState {\n  switch (value) {\n    case 'listening':\n    case 'idle':\n      return 'listening';\n    case 'thinking':\n      return 'thinking';\n    case 'speaking':\n      return 'speaking';\n    default:\n      return 'connecting';\n  }\n}\n\n// ---------------------------------------------------------------------------\n// Factory\n// ---------------------------------------------------------------------------\n\n/**\n * Create a voice client.\n *\n * @returns A {@link VoiceClient} for starting live sessions and reading past\n *   call records.\n *\n * @example\n * ```ts\n * import { createVoiceClient } from '@mindstudio-ai/interface/voice';\n *\n * const voice = createVoiceClient();\n * const session = await voice.startSession();\n *\n * session.on('stateChange', (state) => setOrbState(state));\n * session.on('transcript', ({ role, segmentId, text, final }) =>\n *   upsertCaption(segmentId, role, text, final),\n * );\n * session.on('toolCall', ({ method, status }) => showToolStatus(method, status));\n *\n * // ... later\n * await session.end();\n * ```\n */\nexport function createVoiceClient(): VoiceClient {\n  return {\n    async startSession(options?: StartSessionOptions): Promise<VoiceSession> {\n      options?.signal?.throwIfAborted();\n\n      const minted = await request<{\n        sessionId: string;\n        url: string;\n        token: string;\n        room: string;\n      }>('/sessions', 'POST');\n\n      // The media transport loads on first use — apps that never start a\n      // voice session never fetch it.\n      const lk = await import('livekit-client');\n\n      options?.signal?.throwIfAborted();\n\n      const room: Room = new lk.Room();\n      const listeners = createListeners();\n      let state: VoiceSessionState = 'connecting';\n      let muted = false;\n      let audioEl: HTMLMediaElement | null = null;\n\n      const setState = (next: VoiceSessionState) => {\n        if (next === state || state === 'ended') {\n          return;\n        }\n        state = next;\n        for (const cb of listeners.stateChange) {\n          cb(next);\n        }\n      };\n\n      const emitError = (error: Error) => {\n        for (const cb of listeners.error) {\n          cb(error);\n        }\n      };\n\n      const cleanup = () => {\n        if (audioEl) {\n          audioEl.remove();\n          audioEl = null;\n        }\n        setState('ended');\n      };\n\n      const isAgent = (identity: string) => {\n        const participant = room.remoteParticipants.get(identity);\n        return participant?.kind === lk.ParticipantKind.AGENT;\n      };\n\n      // Agent audio: attach to an SDK-managed hidden element. The agent\n      // waits for this subscription before speaking.\n      room.on(\n        lk.RoomEvent.TrackSubscribed,\n        (track: RemoteTrack, _pub, participant) => {\n          if (\n            track.kind === lk.Track.Kind.Audio &&\n            participant.kind === lk.ParticipantKind.AGENT\n          ) {\n            audioEl = track.attach();\n            audioEl.style.display = 'none';\n            document.body.appendChild(audioEl);\n          }\n        },\n      );\n\n      // Session state from the agent participant's published attribute.\n      room.on(\n        lk.RoomEvent.ParticipantAttributesChanged,\n        (_changed, participant) => {\n          if (participant.kind === lk.ParticipantKind.AGENT) {\n            const value = participant.attributes[ATTR_AGENT_STATE];\n            if (value) {\n              setState(mapAgentState(value));\n            }\n          }\n        },\n      );\n\n      room.on(lk.RoomEvent.Disconnected, () => {\n        cleanup();\n      });\n\n      // Live captions. User transcripts arrive attributed as the local\n      // participant (full-replacement streams per segment, final flagged in\n      // the attributes); agent transcripts are one delta stream per segment,\n      // final when the stream ends.\n      room.registerTextStreamHandler(\n        TOPIC_TRANSCRIPTION,\n        (reader: TextStreamReader, participantInfo) => {\n          const role =\n            participantInfo.identity === room.localParticipant.identity\n              ? 'user'\n              : 'agent';\n          const attributes = reader.info.attributes ?? {};\n          const segmentId = attributes[ATTR_SEGMENT_ID] ?? reader.info.id;\n\n          void (async () => {\n            try {\n              if (role === 'user') {\n                const text = await reader.readAll();\n                const final = attributes[ATTR_TRANSCRIPTION_FINAL] === 'true';\n                for (const cb of listeners.transcript) {\n                  cb({ role, segmentId, text, final });\n                }\n              } else {\n                let text = '';\n                for await (const chunk of reader) {\n                  text += chunk;\n                  for (const cb of listeners.transcript) {\n                    cb({ role, segmentId, text, final: false });\n                  }\n                }\n                for (const cb of listeners.transcript) {\n                  cb({ role, segmentId, text, final: true });\n                }\n              }\n            } catch {\n              // A transcript stream dropping mid-read is cosmetic — skip it.\n            }\n          })();\n        },\n      );\n\n      // Tool activity from the voice worker.\n      room.registerTextStreamHandler(\n        TOPIC_TOOL,\n        (reader: TextStreamReader, participantInfo) => {\n          if (!isAgent(participantInfo.identity)) {\n            return;\n          }\n          void (async () => {\n            try {\n              const parsed = JSON.parse(await reader.readAll()) as {\n                method?: string;\n                status?: string;\n                at?: number;\n                result?: unknown;\n                resultTruncated?: boolean;\n              };\n              if (!parsed.method) {\n                return;\n              }\n              const status =\n                parsed.status === 'done' || parsed.status === 'failed'\n                  ? parsed.status\n                  : 'running';\n              for (const cb of listeners.toolCall) {\n                cb({\n                  method: parsed.method,\n                  status,\n                  at: parsed.at ?? Date.now(),\n                  ...(parsed.result !== undefined\n                    ? { result: parsed.result }\n                    : {}),\n                  ...(parsed.resultTruncated === true\n                    ? { resultTruncated: true }\n                    : {}),\n                });\n              }\n            } catch {\n              // Malformed tool event — skip it.\n            }\n          })();\n        },\n      );\n\n      const abortHandler = () => {\n        void room.disconnect();\n      };\n      options?.signal?.addEventListener('abort', abortHandler, { once: true });\n\n      try {\n        await room.connect(minted.url, minted.token);\n        options?.signal?.throwIfAborted();\n\n        try {\n          await room.localParticipant.setMicrophoneEnabled(true);\n        } catch (err) {\n          const name = (err as Error)?.name;\n          if (name === 'NotAllowedError' || name === 'PermissionDeniedError') {\n            throw new MindStudioInterfaceError(\n              'Microphone access was denied. Allow microphone access for this site and try again.',\n              'microphone_denied',\n            );\n          }\n          if (name === 'NotFoundError') {\n            throw new MindStudioInterfaceError(\n              'No microphone was found on this device.',\n              'microphone_not_found',\n            );\n          }\n          throw err;\n        }\n      } catch (err) {\n        await room.disconnect().catch(() => undefined);\n        cleanup();\n        if (options?.signal?.aborted) {\n          throw new DOMException('Voice session aborted', 'AbortError');\n        }\n        throw err;\n      } finally {\n        options?.signal?.removeEventListener('abort', abortHandler);\n      }\n\n      return {\n        sessionId: minted.sessionId,\n        room: minted.room,\n        get state() {\n          return state;\n        },\n        get isMuted() {\n          return muted;\n        },\n        on(event, callback) {\n          const set = listeners[event] as Set<typeof callback>;\n          set.add(callback);\n          return () => {\n            set.delete(callback);\n          };\n        },\n        async mute() {\n          await room.localParticipant.setMicrophoneEnabled(false);\n          muted = true;\n        },\n        async unmute() {\n          await room.localParticipant.setMicrophoneEnabled(true);\n          muted = false;\n        },\n        async sendText(text: string) {\n          try {\n            await room.localParticipant.sendText(text, { topic: TOPIC_CHAT });\n          } catch (err) {\n            emitError(err as Error);\n            throw err;\n          }\n        },\n        async end() {\n          await room.disconnect();\n          cleanup();\n        },\n      };\n    },\n\n    async listSessions(cursor?: string): Promise<VoiceSessionListPage> {\n      const query = cursor ? `?cursor=${encodeURIComponent(cursor)}` : '';\n      return request<VoiceSessionListPage>(`/sessions${query}`, 'GET');\n    },\n\n    async getSession(sessionId: string): Promise<VoiceSessionDetail> {\n      return request<VoiceSessionDetail>(\n        `/sessions/${encodeURIComponent(sessionId)}`,\n        'GET',\n      );\n    },\n  };\n}\n"],"mappings":"+CA4BA,IAAMA,EAAa,WAGbC,EAAsB,mBAEtBC,EAAa,UAEbC,EAAa,gBAEbC,EAA2B,yBAC3BC,EAAkB,gBAClBC,EAAmB,iBAsJzB,eAAeC,EACbC,EACAC,EACAC,EACY,CAEZ,IAAMC,EAAkC,CACtC,cAAe,UAFFC,EAAU,EAES,KAAK,EACvC,EAEIF,IAAS,SACXC,EAAQ,cAAc,EAAI,oBAG5B,IAAME,EAAM,MAAM,MAAM,GAAGb,CAAU,GAAGQ,CAAI,GAAI,CAC9C,OAAAC,EACA,QAAAE,EACA,GAAID,IAAS,QAAa,CAAE,KAAM,KAAK,UAAUA,CAAI,CAAE,CACzD,CAAC,EAED,GAAI,CAACG,EAAI,GAAI,CACX,IAAIC,EAAe,yBAAyBD,EAAI,MAAM,IAAIA,EAAI,UAAU,GACpEE,EAAY,cAChB,GAAI,CACF,IAAMC,EAAO,MAAMH,EAAI,KAAK,EACxBG,EAAI,QACNF,EAAeE,EAAI,OAEjBA,EAAI,OACND,EAAYC,EAAI,KAEpB,MAAQ,CAER,CACA,MAAM,IAAIC,EAAyBH,EAAcC,EAAWF,EAAI,MAAM,CACxE,CAEA,OAAQ,MAAMA,EAAI,KAAK,CACzB,CAMA,SAASK,GAA6B,CACpC,MAAO,CACL,YAAa,IAAI,IACjB,WAAY,IAAI,IAChB,SAAU,IAAI,IACd,MAAO,IAAI,GACb,CACF,CAGA,SAASC,EAAcC,EAAkC,CACvD,OAAQA,EAAO,CACb,IAAK,YACL,IAAK,OACH,MAAO,YACT,IAAK,WACH,MAAO,WACT,IAAK,WACH,MAAO,WACT,QACE,MAAO,YACX,CACF,CA6BO,SAASC,GAAiC,CAC/C,MAAO,CACL,MAAM,aAAaC,EAAsD,CACvEA,GAAS,QAAQ,eAAe,EAEhC,IAAMC,EAAS,MAAMhB,EAKlB,YAAa,MAAM,EAIhBiB,EAAK,KAAM,QAAO,gBAAgB,EAExCF,GAAS,QAAQ,eAAe,EAEhC,IAAMG,EAAa,IAAID,EAAG,KACpBE,EAAYR,EAAgB,EAC9BS,EAA2B,aAC3BC,EAAQ,GACRC,EAAmC,KAEjCC,EAAYC,GAA4B,CAC5C,GAAI,EAAAA,IAASJ,GAASA,IAAU,SAGhC,CAAAA,EAAQI,EACR,QAAWC,KAAMN,EAAU,YACzBM,EAAGD,CAAI,EAEX,EAEME,EAAaC,GAAiB,CAClC,QAAWF,KAAMN,EAAU,MACzBM,EAAGE,CAAK,CAEZ,EAEMC,EAAU,IAAM,CAChBN,IACFA,EAAQ,OAAO,EACfA,EAAU,MAEZC,EAAS,OAAO,CAClB,EAEMM,EAAWC,GACKZ,EAAK,mBAAmB,IAAIY,CAAQ,GACpC,OAASb,EAAG,gBAAgB,MAKlDC,EAAK,GACHD,EAAG,UAAU,gBACb,CAACc,EAAoBC,EAAMC,IAAgB,CAEvCF,EAAM,OAASd,EAAG,MAAM,KAAK,OAC7BgB,EAAY,OAAShB,EAAG,gBAAgB,QAExCK,EAAUS,EAAM,OAAO,EACvBT,EAAQ,MAAM,QAAU,OACxB,SAAS,KAAK,YAAYA,CAAO,EAErC,CACF,EAGAJ,EAAK,GACHD,EAAG,UAAU,6BACb,CAACiB,EAAUD,IAAgB,CACzB,GAAIA,EAAY,OAAShB,EAAG,gBAAgB,MAAO,CACjD,IAAMJ,EAAQoB,EAAY,WAAWlC,CAAgB,EACjDc,GACFU,EAASX,EAAcC,CAAK,CAAC,CAEjC,CACF,CACF,EAEAK,EAAK,GAAGD,EAAG,UAAU,aAAc,IAAM,CACvCW,EAAQ,CACV,CAAC,EAMDV,EAAK,0BACHxB,EACA,CAACyC,EAA0BC,IAAoB,CAC7C,IAAMC,EACJD,EAAgB,WAAalB,EAAK,iBAAiB,SAC/C,OACA,QACAoB,EAAaH,EAAO,KAAK,YAAc,CAAC,EACxCI,EAAYD,EAAWxC,CAAe,GAAKqC,EAAO,KAAK,IAEvD,SAAY,CAChB,GAAI,CACF,GAAIE,IAAS,OAAQ,CACnB,IAAMG,EAAO,MAAML,EAAO,QAAQ,EAC5BM,EAAQH,EAAWzC,CAAwB,IAAM,OACvD,QAAW4B,KAAMN,EAAU,WACzBM,EAAG,CAAE,KAAAY,EAAM,UAAAE,EAAW,KAAAC,EAAM,MAAAC,CAAM,CAAC,CAEvC,KAAO,CACL,IAAID,EAAO,GACX,cAAiBE,KAASP,EAAQ,CAChCK,GAAQE,EACR,QAAWjB,KAAMN,EAAU,WACzBM,EAAG,CAAE,KAAAY,EAAM,UAAAE,EAAW,KAAAC,EAAM,MAAO,EAAM,CAAC,CAE9C,CACA,QAAWf,KAAMN,EAAU,WACzBM,EAAG,CAAE,KAAAY,EAAM,UAAAE,EAAW,KAAAC,EAAM,MAAO,EAAK,CAAC,CAE7C,CACF,MAAQ,CAER,CACF,GAAG,CACL,CACF,EAGAtB,EAAK,0BACHtB,EACA,CAACuC,EAA0BC,IAAoB,CACxCP,EAAQO,EAAgB,QAAQ,IAG/B,SAAY,CAChB,GAAI,CACF,IAAMO,EAAS,KAAK,MAAM,MAAMR,EAAO,QAAQ,CAAC,EAOhD,GAAI,CAACQ,EAAO,OACV,OAEF,IAAMC,EACJD,EAAO,SAAW,QAAUA,EAAO,SAAW,SAC1CA,EAAO,OACP,UACN,QAAWlB,KAAMN,EAAU,SACzBM,EAAG,CACD,OAAQkB,EAAO,OACf,OAAAC,EACA,GAAID,EAAO,IAAM,KAAK,IAAI,EAC1B,GAAIA,EAAO,SAAW,OAClB,CAAE,OAAQA,EAAO,MAAO,EACxB,CAAC,EACL,GAAIA,EAAO,kBAAoB,GAC3B,CAAE,gBAAiB,EAAK,EACxB,CAAC,CACP,CAAC,CAEL,MAAQ,CAER,CACF,GAAG,CACL,CACF,EAEA,IAAME,EAAe,IAAM,CACpB3B,EAAK,WAAW,CACvB,EACAH,GAAS,QAAQ,iBAAiB,QAAS8B,EAAc,CAAE,KAAM,EAAK,CAAC,EAEvE,GAAI,CACF,MAAM3B,EAAK,QAAQF,EAAO,IAAKA,EAAO,KAAK,EAC3CD,GAAS,QAAQ,eAAe,EAEhC,GAAI,CACF,MAAMG,EAAK,iBAAiB,qBAAqB,EAAI,CACvD,OAAST,EAAK,CACZ,IAAMqC,EAAQrC,GAAe,KAC7B,MAAIqC,IAAS,mBAAqBA,IAAS,wBACnC,IAAIpC,EACR,qFACA,mBACF,EAEEoC,IAAS,gBACL,IAAIpC,EACR,0CACA,sBACF,EAEID,CACR,CACF,OAASA,EAAK,CAGZ,MAFA,MAAMS,EAAK,WAAW,EAAE,MAAM,IAAG,EAAY,EAC7CU,EAAQ,EACJb,GAAS,QAAQ,QACb,IAAI,aAAa,wBAAyB,YAAY,EAExDN,CACR,QAAE,CACAM,GAAS,QAAQ,oBAAoB,QAAS8B,CAAY,CAC5D,CAEA,MAAO,CACL,UAAW7B,EAAO,UAClB,KAAMA,EAAO,KACb,IAAI,OAAQ,CACV,OAAOI,CACT,EACA,IAAI,SAAU,CACZ,OAAOC,CACT,EACA,GAAG0B,EAAOC,EAAU,CAClB,IAAMC,EAAM9B,EAAU4B,CAAK,EAC3B,OAAAE,EAAI,IAAID,CAAQ,EACT,IAAM,CACXC,EAAI,OAAOD,CAAQ,CACrB,CACF,EACA,MAAM,MAAO,CACX,MAAM9B,EAAK,iBAAiB,qBAAqB,EAAK,EACtDG,EAAQ,EACV,EACA,MAAM,QAAS,CACb,MAAMH,EAAK,iBAAiB,qBAAqB,EAAI,EACrDG,EAAQ,EACV,EACA,MAAM,SAASmB,EAAc,CAC3B,GAAI,CACF,MAAMtB,EAAK,iBAAiB,SAASsB,EAAM,CAAE,MAAO7C,CAAW,CAAC,CAClE,OAASc,EAAK,CACZ,MAAAiB,EAAUjB,CAAY,EAChBA,CACR,CACF,EACA,MAAM,KAAM,CACV,MAAMS,EAAK,WAAW,EACtBU,EAAQ,CACV,CACF,CACF,EAEA,MAAM,aAAasB,EAAgD,CACjE,IAAMC,EAAQD,EAAS,WAAW,mBAAmBA,CAAM,CAAC,GAAK,GACjE,OAAOlD,EAA8B,YAAYmD,CAAK,GAAI,KAAK,CACjE,EAEA,MAAM,WAAWC,EAAgD,CAC/D,OAAOpD,EACL,aAAa,mBAAmBoD,CAAS,CAAC,GAC1C,KACF,CACF,CACF,CACF","names":["VOICE_BASE","TOPIC_TRANSCRIPTION","TOPIC_CHAT","TOPIC_TOOL","ATTR_TRANSCRIPTION_FINAL","ATTR_SEGMENT_ID","ATTR_AGENT_STATE","request","path","method","body","headers","getConfig","res","errorMessage","errorCode","err","MindStudioInterfaceError","createListeners","mapAgentState","value","createVoiceClient","options","minted","lk","room","listeners","state","muted","audioEl","setState","next","cb","emitError","error","cleanup","isAgent","identity","track","_pub","participant","_changed","reader","participantInfo","role","attributes","segmentId","text","final","chunk","parsed","status","abortHandler","name","event","callback","set","cursor","query","sessionId"]}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mindstudio-ai/interface",
-  "version": "0.1.12",
+  "version": "0.1.14",
   "description": "Frontend SDK for MindStudio v2 app interfaces",
   "type": "module",
   "main": "./dist/index.js",
@@ -10,6 +10,11 @@
       "types": "./dist/index.d.ts",
       "import": "./dist/index.js",
       "default": "./dist/index.js"
+    },
+    "./voice": {
+      "types": "./dist/voice.d.ts",
+      "import": "./dist/voice.js",
+      "default": "./dist/voice.js"
     }
   },
   "files": [
@@ -55,5 +60,8 @@
   },
   "lint-staged": {
     "**/*": "prettier --write --ignore-unknown"
+  },
+  "dependencies": {
+    "livekit-client": "^2.22.0"
   }
 }