Worked for two days attempting to get Gemini 2.0 to work with pydantic field definitions and json output schema for extraction from markdown documents. Unfortunately, the model performs very poorly relative other models. Just thought you should have the feedback. I am an alpha and beta tester of other foundation model providers and was optimistic that gemini 2.0 model would perform and it was disappointing.
2 Likes
Thank you for the feedback. Can you please provide more details and confirm if you are you referring to the model quality or something else? For more details on how to use Pydantic with the Gemini API, please refer to the “Generate JSON” section of this notebook.