Feedback

I'm finding that about 40-50% of the in-place text replacement actions I invoke don't work the first time. I'll hit my hotkey, wait a couples of seconds, and the selected text will unselect as though the action were run successfully -- but without any changes to the text. Hitting undo (to-reselect the text) and then invoking the action a second time usually does the trick. I'm not sure whether in the background the API call is failing, or if it's something else.

When I select a text and run an action, the text highlight often disappears, and nothing else happens. When I select the text again and run the action again, it usually works on the second or third attempt. I haven't encountered this issue previously and am not sure when it started.

Cache available models and refresh the list once a day. Otherwise, sometimes I am not able to select any model for a few minutes. Also, maybe the option to deselect models we will probably never use and add other models like the moderation ones for testing purposes. Thank you

Add support for the DeepSeek AI, please. Thank you

Don’t erase the text inputted but not sent yet when hiding and reshowing the app. Thank you

for the last few days the application has stopped outputting answers to the GUI and needs to be restarted frequently. macos 15.2, m3 macbook pro.

To write a paragraph, sometimes you would need to do multiple recordings. Say two sentences at a time. Since you don't see the text until you hit the 'recording' button (which trigger the transcription), it would help if there's a way to minimize keystrokes when doing multiple separate recordings in a row. Would it be possible to add a button in the recording window, something like "Transcribe Now", which will send the recording over and start a new recording right after x seconds? And the current "Recording" button would become (technically) a "Transcribe and Close"? That would save a lot of keystrokes of reactivating the recording window. Also would help the flow of thoughts during dictation (you don't have to press shortcuts keys multiple times until you are done). BTW thanks for the wonderful work. I see a lot of potential using it with voice control apps like Talon voice and mac voice control. I have a talon voice command to trigger/stop the recording and the experience is quite smooth so far. With continuous recording support, it would make life much easier for those with RSI :)

Support of OpenRouter, the sole purpose of this is to allow up for up to date model use. This will also serve as an alternative to TogetherAI. And allow users to select openrouter/auto as a set and forget model of all actions.

Currently, actions are very "stiff." One action is attached to one model; it would be great to be able to switch models within the same action and edit the prompts on the spot so they can adjust to specific needs. Next to that, the ability to duplicate an action would also be very helpful.