Adaptation of Agentless for General SWE tasks

[lindafei01]

Hi, thanks for the great work! I’m curious if Agentless could be easily adapted for general software engineering tasks beyond repairing patches. For instance, could we provide Agentless with a prompt to write code?

[p81sunshine]

After running the localize.py, I think the current code relies on the specific data structure of the dataset used. For example, it should clone the repo according to the commit in the dataset, then capture the file content. For different hierarchies of localization steps, it uses different rules to capture the file content. Then, it passes the prompt and the relevant file content to the LLM.
So, if it is to be used for general problems, the code should be modified.