The application provides only a file tree (path, size) as a context for the first request, and the AI ​​agent through OpenAI Function calling / Anthropic tools, itself request the necessary files (the whole or just a header, for example) for the full context.

The shipped features can be found in the release note of 16x Prompt: https://prompt.16x.engineer/release

Ability to directly edit the code with preview and approval (accept / reject). Need to identify if the task is simple and possible.

The current response textarea is too small, especially when inserting large code snippets or writing detailed prompts. It quickly becomes difficult to maintain an overview. I would like the option to resize the textarea significantly, ideally up to 50% of the screen height. Even better, it would be great to have a setting where I can adjust the textarea height according to my personal needs. This would make handling larger inputs much easier.

We need to be able to clear history after a task is completed I can't find anywhere and it's chaotic

I'm on Windows 10, v0.0.82. The file selection is not very eyes-friendly at the moment. The checkmarks are light grey on white background and selected checkboxes are just slightly darker and file names bold. Instead, it would be great if the the whole "row" with the file got highlighted when the file is selected, to quickly tell which files are selected at a glance and not have to rely on checkmarks. This would also greatly help with directories, which don't get bolded at the moment when all files inside one are selected.

I'm on Windows 10, v0.0.82. If all files in all subdirectories are selected for inclusion in the prompt, the top-level directory does not have its checkmark. For example, there is currently no way to tell if a collapsed "src/components" directory has anything in it selected, even if every single sub-directory has every single file included. I would suggest to apply the same selection logic highlight to mid- and top-level directories as to the the lowest-level ones.

Make the custom instructions a toggleable feature so we can have a library of instructions and select specific ones to append to current prompt this way we can keep instructions shorter and mix whatever we need for a specific prompt

Windows displays a warning when installing the software (some windows defender smartscreen filter). It would be nice to get rid of this, as this gives a bad impression that the app could do malicious things.

Package your solution up to support installation via Brew.

Could you make the message textarea field fixed either at the top or bottom of the screen, so that users can scroll through the responses without having to scroll down to type their reply?

I would like to have a search function above the response textarea field, allowing me to search through previous responses in the history. The search results should be clickable, enabling me to jump to the relevant responses, and there should also be an option to quickly return to the textarea after viewing the results.

Is it possible to adjust the temperature for a third-party API?

This will be a great feature similat to claude.ai website chat feature which allows to paste images or upload images directly in chat box instead of uploading manually if it's saved in clipboard.

Say, I want to keep going with the same task but without previous history, from scratch. Presently, follow-up questions are added to the previous history, no matter what. Would be great to be able to clear the history.

My workflow is as follows: I create two-three workspaces that represent absolutely unrelated tasks and work with them. History is a great feature, it allows asking follow-up questions, but history is not stored along with the workspace. So when I switch the workspace - I have history the was valid for the previous workspace, so the follow-up opportunity is missing as soon as the workspace is switched, I have to re-start the conversation. Is it technically possible to keep the history along with the workspace? So the follow-ups will be workspace-dependent.