Hey everyone, I wanted to present this question to see what your strategies are for enabling the LM to handle large code files.
Some of my files have upward to 2200 lines of code. I am finding it increasingly hard to get any kind of productivity out of the LM.
It seems, at a certain point, it simply can’t read the other code, thereby producing duplicate or wrong code.
I’m using Claude Sonnet. Have also tried the others, but as of now I am not getting anything of quality.
I also am using cursor rules.
I have indexed the codebase, etc…
Is there anything else I can do to help the LM see the entire code files?
ive started asking it to compartmentalize files into sections and its pretty nice. twice the files but way better structure and easy to go specific snipppets
Hello, yes, many times. I have found the composer to actually be great at doing very focused code implementations, but doesn’t seem effective in my case to interpret the bigger picture, due to the size of some of these files.
Which might take a few other threads of context (haha) to let you grok what we’ve all be kjetching about for the last weeks on this forum Much love and effort is being thrown at the exact problem you are facing.
Grab a [thing] and read a bunch of ideas on here and attempts at solving your issue.
I think we are all going to soon congeal on a .01 standard for Agentic Controls.
Much love and effort is being thrown at the exact problem you are facing.