Skip to main content
🤖 LLM Friendly: This page is available in raw Markdown format for LLM consumption:changelog.md|Get full documentation:llms.txt/llms-full.txt

Changelog

All notable changes to the @firetell/firetell-client-sdk package are documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.2.4] - 2026-09-09​

Added​

  • Camera Mute/Unmute (Toggle Camera):
    • Added call.muteVideo(), call.unmuteVideo(), and call.toggleCamera() methods on Call.
    • Added call.isCameraOff: boolean property tracking local camera state.
    • Added ECallEventName.CAMERA ("camera") event emitted locally and on remote camera state updates via signaling.
  • Native Screen Sharing (getDisplayMedia + replaceTrack):
    • Added call.startScreenShare(), call.stopScreenShare(), and call.toggleScreenShare() on Call.
    • Seamlessly replaces video track on the active RTCRtpSender without SDP renegotiation or call interruption.
    • Automatically restores the original camera video track when the user stops sharing screen.
    • Added call.isScreenSharing: boolean property tracking active screen share state.
    • Added ECallEventName.SCREEN_SHARE ("screenShare") event emitted locally and on remote screen share updates.
  • SDP Video Auto-Detection in call.offer & accept():
    • Call.handleWsMessage: Automatically falls back to inspecting the remote SDP offer for active video media sections (/m=video [1-9]/) if the explicit is_video boolean flag is omitted in the call.offer payload, ensuring call.isVideo is reliably set on the callee side.
    • Call.accept(): Evaluates this.remoteDescription.sdp to guarantee this.isVideo is true before setting up local media, prompting the browser for both Camera and Microphone permissions ({ video: true, audio: true }).
  • Resilient Remote Stream Track Assembly:
    • Improved RTCPeerConnection.ontrack in _setupWebrtcMedia: Automatically creates a MediaStream and attaches tracks incrementally if event.streams[0] is missing, ensuring both remote audio and video tracks are emitted via ECallEventName.REMOTE_STREAM.
  • Caller Metadata Continuity:
    • Added support for from_avatar and transfer_reason fields in call.offer event processing so callee avatar and transfer context remain intact through WebRTC session negotiation.

Fixed​

  • Callee Video Negotiation:
    • Fixed an issue where incoming video calls defaulted to audio-only on the callee side due to missing is_video flag in signaling WebSocket payloads.
  • Example Demo Web App:
    • Fixed hideIncomingBanner reference error on call dismissal.
    • Added dynamic updateVideoUI() helper to automatically bind and display local/remote video elements when tracks arrive.

[1.2.3] - 2026-09-04​

Changed​

  • call.transfer() simplified (Breaking):
    • Signature changed from transfer(targetUsername, teamId?, reason?) to transfer(target, reason?).
    • target now accepts: agent username, extension number (e.g. "100"), team ID (te_...), or SIP account ID (si_...).
    • Removed teamId parameter — target type is auto-resolved server-side via prefix-based resolution.
  • FiretellClient.sendTransfer() simplified (Breaking):
    • Signature changed from sendTransfer(callId, targetUsername, teamId?, reason?) to sendTransfer(callId, target, reason?).
    • HTTP fallback body changed from { target_username, team_id, reason } to { target, reason }.
  • Per-call WebSocket payloads cleaned up:
    • Removed redundant call_id from all WS event payloads (call.hangup, call.answer, call.reject, call.transfer, call.dtmf, call.mute, call.hold, call.unhold). The per-call WS connection inherently binds the socket to a specific call, making call_id unnecessary.
    • call.transfer payload field renamed from to to target for consistency with the HTTP API.

[1.2.2] - 2026-09-03​

Added​

  • getPhoneNumbers() REST API on FiretellClient:
    • Added public async getPhoneNumbers(options?: { page?: number; limit?: number }): Promise<IClientPhoneNumbersResponse> to query phone numbers (DIDs) accessible by the authenticated agent and their shared teams.
    • Automatically sends the agent's JWT to GET /api/v1/call-center/phone-numbers.
    • Added API_ENDPOINTS.PHONE_NUMBERS ("/api/v1/call-center/phone-numbers").
    • Added IClientPhoneNumber and IClientPhoneNumbersResponse interfaces exporting phone number details (id, number, title, country_code, dial_code, status, capabilities, enable_outbound, shared_teams_id).
  • Static & Named SDK Version Exports:
    • Added static FiretellClient.VERSION property on the class for direct, synchronous version access without instantiation.
    • Exported SDK_VERSION constant directly from package root (import { SDK_VERSION } from '@firetell/firetell-client-sdk').
    • Added "./package.json": "./package.json" to the package exports map for direct version inspection by package managers and bundlers.

[1.2.1] - 2026-08-22​

Added​

  • transfer_reason in call.ring Event & Call Instance:
    • Added transfer_reason?: string to ICallRingParams payload in the call.ring SSE event.
    • Exposed call.transferReason?: string on the Call instance when a transferred call session is initialized.
    • Allows frontends, Webphones, and Console agents to display the contextual transfer reason provided by Voice AI agents or human agents during ringing.

[1.2.0] - 2026-08-20​

Added​

  • Dedicated Per-Call SSE Event Stream (Call.connectCallEventStream):
    • Automatically establishes a dedicated SSE connection (/stream?call_id=<callId>) for each active call session (call.start(), call.accept(), call.joinSession(), and client.createCallSession()).
    • Subscribes to the Per-Call channel for real-time speech-to-text transcriptions and recording events without flooding the global workspace event stream.
    • Automatically disconnects and cleans up the stream upon call termination (call.destroy()).
  • Call-Scoped Real-Time Transcription Events & APIs:
    • Added ECallEventName.TRANSCRIPTION, TRANSCRIPTION_STARTED, TRANSCRIPTION_DIALOGUE, and TRANSCRIPTION_COMPLETED.
    • Added ergonomic helper listener methods on Call: call.onDialogue((dialogue) => ...) and call.onTranscription((event) => ...).
    • Exported TypeScript interfaces: ITranscriptionStartedEvent, ITranscriptionDialogueEvent, ITranscriptionCompletedEvent, and TranscriptionEvent.
  • Call-Scoped Real-Time Recording Events & APIs:
    • Added ECallEventName.RECORDING, RECORDING_STARTED, RECORDING_COMPLETED, and RECORDING_READY.
    • Added ergonomic helper listener method on Call: call.onRecording((event) => ...).
    • Exported TypeScript interfaces: ICallRecordingStartedEvent, ICallRecordingCompletedEvent, ICallRecordingReadyEvent, and CallRecordingEvent.
  • Public Getters on FiretellClient:
    • Added client.getBaseUrl() and client.getJwt() getters.

Changed & Improved​

  • Strict Single-Emission & Call-Scoped Event Architecture:
    • Removed duplicate event emissions across FiretellClient._initEventStream() and Call._handleSseMessage().
    • Decoupled per-call transcription events from client.events, strictly routing all live subtitles, chunks, and summaries directly to the owning Call instance (call.on(...)).

[1.1.7] - 2026-08-17​

Added​

  • FiretellClient.stopSupervision(): Added helper method to terminate an ongoing call supervision session (cleans up WebRTC call session and calls REST API DELETE /v1/call-center/calls/:id/supervision).
  • Call State Monotonicity Guard: Added automatic state ranking guard in FiretellClient to prevent late/out-of-order call.created events from overriding or regressing active call state machine.

[1.1.6] - 2026-08-17​

Added​

  • SseStreamClient (Header-Based SSE Streaming): Integrated lightweight, zero-dependency SSE stream client with Authorization: Bearer <jwt> HTTP header support via fetch and ReadableStream, completely eliminating JWT token exposure in SSE stream URLs and server access logs.
  • SSE Stream Exports: Exported SseStreamClient, SseMessageEvent, and SseStreamConfig from SDK entry point.

[1.1.5] - 2026-08-17​

Added​

  • FiretellClient.startSupervision(): Added high-level helper method to initiate call supervision (listen, whisper, barge) and automatically establish WebRTC audio session in one call.
  • Call.joinSession(): Added method to connect signaling and send WebRTC offer using pre-generated call_token and ws_url.
  • ISupervisionResponse.ws_url: Added required ws_url field to ISupervisionResponse interface to support dedicated per-call WebSocket connections.

[1.1.4] - 2026-08-15​

Fixed​

  • Call Creation: Renamed _createCallSession to createCallSession.

[1.1.3] - 2026-08-14​

Added​

  • Real-Time Call Lifecycle Event Enums: Added EClientEventName.CALL_CREATED, EClientEventName.CALL_STARTED, and EClientEventName.CALL_ANSWERED.
  • Active Call State Synchronization: Added automatic transition of active calls to ECallState.ACTIVE / ANSWERED when receiving call.answered over the SSE stream.

[1.1.2] - 2026-08-14​

Added​

  • Client Event Enum Values (CALL_ENDED & CALL_CANCELED): Added EClientEventName.CALL_ENDED and EClientEventName.CALL_CANCELED.
  • SSE Stream Listeners for Early Call Cancellation: Automatically cleans up Call instances and dismisses incoming call ringing UI modals when calls are answered by another agent or canceled before WebSocket connection completion.

[1.1.1] - 2026-08-12​

Fixed​

  • Inbound Call DTLS Setup Role Validation: Fixed Answerer must use either active or passive value for setup attribute error during callee-initiated unhold (Call.unhold()).
  • Optimized Full ICE Gathering: Reduced hard safety timeout to 6s, and implemented a 3s fallback timer verifying presence of STUN Public IP (typ srflx) or Relay (typ relay) candidates.

[1.1.0] - 2026-08-12​

Fixed​

  • DTLS SSL Role Preservation on Renegotiation: Preserves established DTLS a=setup role (active / passive) in setRemoteDescription() across renegotiation answers from Media Server.
  • Incoming Call Accept Error Safety: Added try...catch block around Call.accept() to guarantee proper media cleanup and ECallState.ERROR state emission if microphone access or WebRTC answer creation fails.

[1.0.9] - 2026-08-12​

Fixed​

  • Call Unhold Event Signaling: Fixed WS event name emitted during Call.unhold() from "call.hold" to "call.unhold".
  • WebRTC Remote SDP Parsing: Added extractSdpInit() helper in Call.ts to safely parse SDP payloads received in WebSocket messages.
  • ICE Server Fallback: Configured DEFAULT_ICE_SERVERS (Google STUN + Cloudflare STUN) as default fallback.

[1.0.0] - 2026-01-01​

Added​

  • Initial release of @firetell/firetell-client-sdk.
  • WebRTC audio/video calls via native per-call WebSocket signaling.
  • JWT-based session authentication (session.connect).
  • Outbound call support (Call.start()) and incoming call handling (Call.accept(), Call.reject()).
  • Call controls: hold/unhold, mute/unmute, DTMF, transfer.
  • Auto-reconnect with exponential backoff.
  • Agent presence notifications (workspace.agent.state).
  • Full ICE gathering (non-Trickle) for Media Server compatibility.
  • Multi-format builds: CJS, ESM, IIFE with TypeScript declarations.