“LLM evaluation”

Hobby project for LLM code editing in large codebases, run via Docker with a web UI and a SWE-bench harness.