Add swe-bench test - Not to run in regular PRs#100
Conversation
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #100 +/- ##
==========================================
- Coverage 92.27% 91.84% -0.43%
==========================================
Files 21 21
Lines 880 834 -46
==========================================
- Hits 812 766 -46
Misses 68 68
Flags with carried forward coverage won't be shown. Click here to find out more.
🚀 New features to boost your workflow:
|
| command: str = "" | ||
|
|
||
| class LLMInterface(ABC): | ||
| def __init__(self, system_prompt: str, max_retries: int = 3): |
There was a problem hiding this comment.
LLMInterface.__init__ was removed but _validate_llm_response still relies on self.retries, self.max_retries, and self.messages. This means any new subclass that forgets to manually initialize these will get an error at runtime, nothing in the interface signals they're required.
Maybe its better to restore a base __init__ and have subclasses call super().__init__(). The only subclass-specific part is whether messages includes a system prompt entry that can be handled after the super call by appending to messges in the subclass.This also removes the duplicated 3 lines in each subclass.
No description provided.