@craftedxp/voice-js 0.5.4 → 0.6.0

package/dist/browser.d.mts

@@ -221,6 +221,61 @@ interface RoomSession {
 }
 declare const joinRoom: (opts: JoinRoomOptions) => Promise<RoomSession>;
 
+/**
+ * Browser-friendly text-channel chat session. Mint a `ct_` token with
+ * `channel: 'text'` on your backend, then call `startTextSession({...})`
+ * to open the SSE stream.
+ *
+ * Each `.send(text)` is a fresh POST; SSE-per-turn means the connection
+ * closes when each turn ends. Conversation state lives server-side on the
+ * underlying CallRecord.
+ */
+type ChatEvent = {
+    type: 'chat.started';
+    chatId: string;
+    callId: string;
+} | {
+    type: 'token';
+    text: string;
+} | {
+    type: 'tool.call';
+    name: string;
+    args: unknown;
+} | {
+    type: 'tool.result';
+    name: string;
+    ok?: boolean;
+    [key: string]: unknown;
+} | {
+    type: 'turn.end';
+    finishReason: 'stop' | 'aborted' | 'length' | 'tool_error';
+    committedText?: string;
+} | {
+    type: 'error';
+    code: string;
+    message: string;
+};
+interface StartTextSessionOpts {
+    baseUrl: string;
+    token: string;
+    agentId: string;
+    /** Optional inline first user message; otherwise the agent's greeting opens the stream. */
+    text?: string;
+    /** Override the global fetch (useful for tests; defaults to globalThis.fetch). */
+    fetch?: typeof fetch;
+}
+interface TextSession {
+    id: string;
+    callId: string;
+    /** Async iterable for the opening turn — greeting tokens / first reply if text was inlined. */
+    greeting: AsyncIterable<ChatEvent>;
+    /** Send a user message; returns an async iterable for the agent's reply. */
+    send(text: string): Promise<AsyncIterable<ChatEvent>>;
+    /** End the session — DELETE /v1/calls/:callId. */
+    end(): Promise<void>;
+}
+declare function startTextSession(opts: StartTextSessionOpts): Promise<TextSession>;
+
 interface FetchTokenArgs {
     /** The agent the SDK is about to call. */
     agentId: string;
@@ -380,6 +435,15 @@ interface VoiceClientFactory {
      * implement this — livekit-client is a browser-only WebRTC client.
      */
     joinRoom?: (options: Omit<JoinRoomOptions, 'apiBase'>) => Promise<RoomSession>;
+    /**
+     * Open a text-channel chat session (no microphone / audio required).
+     * Mint a `ct_` token with `channel: 'text'` server-side, then call
+     * this to connect. Returns a `TextSession` with:
+     *   - `.greeting` — async iterable for the opening turn
+     *   - `.send(text)` — send a user message; returns an async iterable for the reply
+     *   - `.end()` — close the session (DELETE /v1/calls/:callId)
+     */
+    startTextSession?: (opts: Omit<StartTextSessionOpts, 'baseUrl' | 'fetch'>) => Promise<TextSession>;
 }
 
 type OnChunk = (pcm: ArrayBuffer) => void;
@@ -511,4 +575,4 @@ declare const parseIncomingCall: (raw: unknown) => IncomingCallPayload;
  */
 declare function configureVoiceClient(config: VoiceClientConfig): VoiceClientFactory;
 
-export { type Call, type CallEndEvent, type CallEndReason, type CallError, type CallErrorCode, type CallState, type CaptureController, type CaptureOptions, type ClientTool, type ClientToolMap, type FetchToken, type FetchTokenArgs, type FetchTokenResult, type IncomingCallPayload, type JoinRoomOptions, type OnAgentSpeakingChange, type OnChunk, type OnError, type OnVolume$1 as OnVolume, type PlaybackController, type PlaybackOptions, type ProtocolCallbacks, type ProtocolState, type RWSEvent, type RWSOptions, type ReconnectingWebSocket, type RoomEventName, type RoomEventPayloads, type RoomParticipantInfo, type RoomSession, type ServerMessage, type StartCallOptions, type SystemMessage, type TranscriptEntry, type TranscriptMessage, type VoiceClientConfig, type VoiceClientFactory, type VolumeEvent, type WebSocketFactory, type WebSocketLike, buildWsUrl, configureVoiceClient, createAudioCapture, createAudioPlayback, createProtocolState, createReconnectingWebSocket, handleServerMessage, joinRoom, parseIncomingCall };
+export { type Call, type CallEndEvent, type CallEndReason, type CallError, type CallErrorCode, type CallState, type CaptureController, type CaptureOptions, type ChatEvent, type ClientTool, type ClientToolMap, type FetchToken, type FetchTokenArgs, type FetchTokenResult, type IncomingCallPayload, type JoinRoomOptions, type OnAgentSpeakingChange, type OnChunk, type OnError, type OnVolume$1 as OnVolume, type PlaybackController, type PlaybackOptions, type ProtocolCallbacks, type ProtocolState, type RWSEvent, type RWSOptions, type ReconnectingWebSocket, type RoomEventName, type RoomEventPayloads, type RoomParticipantInfo, type RoomSession, type ServerMessage, type StartCallOptions, type StartTextSessionOpts, type SystemMessage, type TextSession, type TranscriptEntry, type TranscriptMessage, type VoiceClientConfig, type VoiceClientFactory, type VolumeEvent, type WebSocketFactory, type WebSocketLike, buildWsUrl, configureVoiceClient, createAudioCapture, createAudioPlayback, createProtocolState, createReconnectingWebSocket, handleServerMessage, joinRoom, parseIncomingCall, startTextSession };