Hello, I don’t know if this is the right place to report this issue, but we have absolutely no business use for Gemini at all without the ability to continue raw text or unscoped textual patterns in general.
In docs, all we see is chat and prompts; which is very limiting and not how we use language models at all. We need actual Text Completion with a stop sequence. Just telling the fine-tuned AI to “continue” introduces a lot of logistical biases that we certainly want none of.
We’ve been wanting to use and experiment with Gemini for the literal last whole year but we’ve been sticking with hosting our own models for now since at least they allow us to get the raw outputs.
And we would have certainly continued using OpenAI’s API if they still supported text completion with their latest models. So, Google, maybe be different and support it.
Or at least, add the ability to continue the last message in the history parameter. This would actually allow people to continue last incomplete message sent by Gemini; maybe interrupted by end users or due to Google’s servers being down, or the max tokens just being reached; Very useful for everyone.