"API space_import failed with status 500: Internal server error"
I tried importing a space .zip file that seemed to be broken. I had been working on it for days and now I can't seem to get it fixed.
Despite lots of issues getting an LM Studio model working semi-reliably in Space Agent, I did manage to do it. However, it has been a major hassle to get it to consistently understand how to work within spaces. I have plenty of vram, but context settings seem to be hardcoded at exactly 120000 as a maximum.
My agents kept complaining about the middle of my widget's code being trunicated and they kept trying to read and patch line-by-line as it didn't seem reliable. I tried downloading the .onnx model, but could never get that to complete either. It has just been a hassle all the way through.
--
As a side note:
Has this project been abandoned? - I haven't seen any updates for a bit and despite enjoying the heck out of the concept, the practical reality of using it reliably screams vibe-code. I suggest going back to GPT 5.4 or Sonnet 4.6 and runing with that model for coding. It is the best in my experience. I really want this project to succeed, but if I can't even save/edit/restore my widgets reliably without errors, it makes my heart break that I might have to abandon this project too. It has so much potential. I think you should keep going - the vibe and usefulness of this tool is out of the world (pun not intended).
Once we can get it polished and functional on the more obscure space and file/widgets safety/restorability (without 500 errors) and local model reliability features (the major parts of the user experience), this should be amazing.
"API space_import failed with status 500: Internal server error"
I tried importing a space .zip file that seemed to be broken. I had been working on it for days and now I can't seem to get it fixed.
Despite lots of issues getting an LM Studio model working semi-reliably in Space Agent, I did manage to do it. However, it has been a major hassle to get it to consistently understand how to work within spaces. I have plenty of vram, but context settings seem to be hardcoded at exactly 120000 as a maximum.
My agents kept complaining about the middle of my widget's code being trunicated and they kept trying to read and patch line-by-line as it didn't seem reliable. I tried downloading the .onnx model, but could never get that to complete either. It has just been a hassle all the way through.
--
As a side note:
Has this project been abandoned? - I haven't seen any updates for a bit and despite enjoying the heck out of the concept, the practical reality of using it reliably screams vibe-code. I suggest going back to GPT 5.4 or Sonnet 4.6 and runing with that model for coding. It is the best in my experience. I really want this project to succeed, but if I can't even save/edit/restore my widgets reliably without errors, it makes my heart break that I might have to abandon this project too. It has so much potential. I think you should keep going - the vibe and usefulness of this tool is out of the world (pun not intended).
Once we can get it polished and functional on the more obscure space and file/widgets safety/restorability (without 500 errors) and local model reliability features (the major parts of the user experience), this should be amazing.