Chore: Make sure vision gets data from camera, mic and screen and it works #35
Description
We adapted our plugin-vision to take a data stream from the frontend. However, it hasn't really been tested in container, during gameplay. Go through the end to end, make sure it works, the agent receives the feed and gets the data. Video will be ongoing. Transcription of audio will need to be figured out-- are we handling this like a message? Or just ambient audio? What's interesting and fun for the game?