Changelog

Attendee now supports bidirectional realtime audio streaming through websockets! You can receive audio from meetings and have your bot output audio in real-time.

This allows you to send voice agents into the meeting and have them respond to questions and requests in real-time.

Configure via the websocket_settings.audio parameter when creating a bot. Check out our example integration with Deepgram's Voice Agent API and the API docs for implementation details.

Attendee now supports the ability to retrieve events for each participant in the meeting. This is useful for products that want to track participant activity and behavior in the meeting.

Check out the docs for more details.

Attendee now supports the ability to pause and resume the recording. This feature may be useful for products where there's a chance of recording something sensitive and you want to let the user prevent that. Pausing will not change the recording's length - the paused part will be completely black and silent.

Check out the API reference for more details.

Attendee now supports scheduled bots! Just include a join_at parameter in your API call to create a bot and the bot will join at precisely that time.

This feature is perfect for automating bot attendance for calendar meetings or ensuring your bot joins exactly when needed without manual intervention.

See the scheduled bots documentation for more details.

Attendee bots are now capable of sending chat messages in the meeting!

There are two ways to send a chat message:

• Specify the message to send when the bot joins, via the bot_chat_message param in the create bot endpoint.
• Send a chat message at any point in the meeting via the send_chat_message endpoint.

Checkout the API reference for more details.

You can now retrieve the chat messages exchanged during a meeting. If you're trying to have an LLM analyze meeting information, the chat messages often contain useful info that can complement the information in the transcript. This functionality is working for all 3 meeting platforms. Check out the documentation here.

Released the ability to receive transcripts in real-time without polling the API, currently only for Deepgram transcription. Provide a callback URL in the API call, and Deepgram will send transcript results directly to that URL via HTTP POST requests or websockets. This method offers low latency but requires handling Deepgram's format directly, and transcripts won't be stored on Attendee's server. Slack thread with more details

Users can now create bots directly through the Attendee web UI. Initially, options to set the bot name and meeting URL are available, with more parameters to be exposed later. Slack thread with more details.

Released a new method for recording the Google Meet UI. Instead of rendering a custom UI, Attendee is now recording a portion of Google Meet directly. This aims to improve active speaker highlighting, display of participants without video, and layout accuracy.

Users running the Attendee app locally and utilizing Zoom meeting bots need to rebuild their Docker image. The Docker image has been updated to use a new version of the Zoom Meeting SDK.

Added the ability to transcribe Google Meet audio via the Deepgram API. This enables retrieval of per-participant audio streams, allowing for perfect diarization. Slack thread with more details.

Audio output for Google Meet bots is now in production. Two endpoints are available: /output_audio to play a base64 encoded MP3 file, and /speech for text-to-speech (currently supporting Google's TTS API). Slack thread with more details.

Webhooks for bot state changes are now available to all users. Refer to the documentation for details.

Bots can now display a custom image in Zoom and Google Meet, allowing for company logo display. The image can be assigned in the POST /bots endpoint using the bot_image parameter (Base64 encoded PNG). Teams support will follow shortly. Optimal aspect ratios are 4:3 or 2:1.

Added support for transcribing Google Meet captions in various languages. Pass the language code to the API using the transcription_settings.meeting_closed_captions.google_meet_language argument. A list of available codes can be found here.

MS Teams meetings are now supported in production. Provide a Teams meeting URL (e.g., https://teams.live.com/meet/xxx?p=xxx) and the bot will join and record.