mirror of
https://github.com/Significant-Gravitas/AutoGPT.git
synced 2026-02-11 15:25:16 -05:00
fea46a6d28
Refactor LLM handling and cache logic: instantiate and pass a LlmModel instance to generate_model_label (rename model_enum -> model) to ensure consistent enum usage when building labels. Remove hasattr guards and directly clear the v2 builder caches during runtime state refresh so cached providers and search results are always attempted to be cleared. Update the AIConditionBlock test fixture to use LlmModel.default() instead of a hardcoded gpt-4o string. These changes simplify the code and standardize LlmModel usage.