The recent changes annoyed me for a bit when shift enter added the message to the queue instead of going to a new line but now im wondering what is the value or point or use case or benefit to load up multiple user messages before submitting?
I never done that but feel like maybe im missing out on something potentially useful. How does it work, read each message and respond and in that case whats the point, does splitting up longer messages have some kind of benefit or performance increase?
When you send data to Gemini, you are basically using this API : Method: models.generateContent | Google AI for Developers | Google for Developers. Now, looking at the details, you see it really uses https://ai.google.dev/api/rest/v1beta/Content, which is an array of Part and a role, which is mostly the string “user”.
Having multiple Part allows for interleaving of text and media content, and is useful even if you are submitting multiple text-only Part one after the other. When preparing content, you can have fixed template text that you want mixed in with variable user provided text, and the API design makes this easier.
On the backend, all the Part are converted into a token stream, it makes no difference if it was all one long string or if you had segmented it into Parts. No cost difference, no performance difference. The API design using Parts is to make programming easier.
The AI Studio helps with a similar approach when designing a longer prompt. Hope that helps.