Abstract: Audio streams can be produced according to sequences of select content items. A first content item can be identified based on user input indicating tags associated with it. A second content item, and a content sequence for ordering the first content item and second item, are identified based on scores determined for various candidate content items. An audio stream is produced according to the content sequence and may then be transmitted to a client device for playback. The tags include metadata usable to selectively identify content items that can be combined to produce an optimal audio stream, such as based on the types of content items, the number of content items, the purpose of the content items to the audio stream, or more. Missing content items can be detected and signaled for production. An audio stream and content sequence therefor can be updated after initial production occurs.