batch_generate_json_retries in our LLM backend has tyoe signatures like this prompts: list[str] | list[dict[str, str]], which make the code and type inference complicated. This might be due to mixing prompts and messages in the same variables. Should be looked at and potentially refactored at some point in time.
batch_generate_json_retriesin our LLM backend has tyoe signatures like thisprompts: list[str] | list[dict[str, str]], which make the code and type inference complicated. This might be due to mixing prompts and messages in the same variables. Should be looked at and potentially refactored at some point in time.