Large file updates can hit the model output limit

[rmulligan]

I'm receiving the following during the "build" step.

🚨 Server error
→ ListenStream - stream chunk missing function call. Reason
→ Stop, File

I am using openAI and the cloud implementation.

I can't find out where it's getting hung up as it's already generated the code changes.

Thanks in advance! This is great work so far.

BTW. I'm having it work on augmenting emacs-lisp.

[danenania]

Hey @rmulligan I replied to you in discord as well--this is most likely caused by a large file/complex change which hits the model's token output limit.

I have some ideas for how to handle this, but in the meantime a workaround could be to refactor large files into smaller files if possible.

[danenania]

For logging, if you're building from source, you could add your own logging statements--unfortunately there isn't anything fancier than that in place yet. All the requests happen here https://github.com/plandex-ai/plandex/tree/main/app/server/model/plan -- the files names *_exec.go are the ones that execute requests and file names like *_stream.go are where the response streams are handled.

It would be good to add logging levels or another kind of request tracking.

[rmulligan]

Sounds good. Thanks! I ended up doing just that. Plandex was able to refactor into smaller files and then didn't have any problems after that. (other than cost)