How does kcpp launcher overwrites default Lite settings?

[alex-ie]

I have written a simple bash script to customize index.html - it changes fonts / windows positions and default settings for Lite. Today I have tried to run the script on klite.embd in --unpack folder. The fonts have changed but settings got reset during launcher run.

I have realized settings of Lite are populated by launcher, e.g. it needs to set context sizes and seems just populates everything else too. Where can I edit the settings in unpacked folder?

I suspect they are in Python files, but search for e.g. trimsentences returned only klite.embd. I have tried to find the code with help of one of free LLM models but have failed.

Also even kcpp_sdui.embd default are overwritten - that I did not expect, e.g. I want to enable preview by default (Fetch generation progres - typo as in the UI), changed it (when opened renamed as html - changed, if via launcher - as before).

TIA

[henk717]

It has an api endpoint for some settings, if you want a custom launch setting I don't recommend trying to hack this with a bash script, instead use our preloadstory feature in the launcher designed to set a default lite environment. All that variable needs is a copy of your Lite save.

The browser also remembers the settings so some of them can be for previous sessions.

[LostRuins]

Yup, what henk said, prepare your favorite settings, save it in lite as a .json save file, then add that to your bash script launcher as a --preloadstory and it shall launch with your desired "pre-built" settings.

[alex-ie]

Thanks, --preloadstory will do I think. What about SD UI settings?

[henk717]

To my knowledge sdui has no backend integration as far as settings go, its fully standalone.

[alex-ie]

See end of my post, changed SD UI emdb file's option gets overwritten in run via kcpp launcher. Per analysis by an LLM, browser "localStorage key" is involved (ln("fetchGenerationProgress", "Off")).

[alex-ie]

After I have posted this, on another attempt in a new conversation the model produced a workaround - renaming fetchGenerationProgress in ln("fetchGenerationProgress", "Off") and it seems to work for now.

ln("fetchGenerationProgress", "Off") -> ln("fetchGenerationProgress_dummy", "Image")