theJayTea
Mac OS verison appears to not send selected text to Ollama Server
I primarily use writing tools as a form of fast and robust spellcheck across different apps on Mac OS, I recently took a break from using the tool, which worked previously, while I set up a dedicated home server for running llms locally on my network. It uses ollama to run models and is started with the Ollama serve command. This ollama instance is confirmed to work with other programs, however, upon reinstalling writing tools to my machine, and asking it to proof the selected text using my ollama instance, I encounter an error. When tried locally and with my home server, ollama does not receive the text that it is supposed to proof. It instead deletes the selected text and replaces it with a response that the llm spits back when provided the prompt that it has been given by the preset that I use, without any other information, such as my selected text. Usually I receive outputs like "this is exactly what i was looking for" or "REQUEST CANNOT BE PROCESSED WITH PROVIDED TEXT".
Hi, I'm sorry that you are facing issues with the app. I have just published a new version. Please give it a try and let me know if the issue is fixed.
Hello, I am so sorry for the late response, I had a personal issue come up and was away from my computers. It still does not work. I actually think the fault may lie more with ollama than with the code of your program. Through further analysis, ollama receives the prompt given by your program, but not the text that the program should send. It doesn't receive said text at all, even appended to the prompt. I suspect ollama may have changed aspects of how it communicates with other programs that may be negatively effecting the use of the Mac version of the app. I think this was likely done with the release of gpt-oss and their own app, as it broke at the same exact time. This, however is an amateurs guess and through trying to debug the program myself I gave myself a bit of a headache. If there is any more diagnostic logs or things on my system that would help or assist you, let me know what to grab and I will get them for you. I find this program to be really the only useful method for using llms for mass grammar correction, so it is very valuable to my current stack of programs that my home llm server uses.
@Starlightsysdeveoper btw, for now, LM studio should work too if you're not able to get Ollama working. Feel free to check it out!
Hi, sorry for the late response. Please check v5.3 and see if the issue is fixed.