1)I have my code of 140k lines, I want to generate more code with it, Is there any way I can do it? Also, what is the token limit? how can I overcome it in the free version of chatgpt?
- Chatgpt loses context sometimes, for instance I am giving a larger code and working with it, it gives good answers initially but then forgets what it was doing. Sometimes it even happens on the second prompt. Why does that happen? is there a way to overcome this shortcoming?
You’d probably be better off using a tool built for this purpose like GitHub’s Copilot, or Cursor.
2 Likes
I think, they are are good for short codes, but for long codes, these tools do not understand the logic or produce desirable goals
I can understand why you would think this, but there’s no reason to assume these things when you can try them easily for free.
140k lines of code is a massive code base. There’s something seriously concerning when you feel it’s required to send it ALL to generate more code.
4 Likes
One trick you can do is send it all with a big context model to have it summarize a list of all functions and variables. Then when you’re asking for new code or changed code, provide the list and tell the LLM to let you know if it will need access to other functions. It’s clunky (and what Cursor and others do in the background, I believe), but it can be useful until the context windows are a lot larger.
Good luck!
2 Likes