Bespoke Curator is an open-source Python library designed for post-training data curation and structured data extraction in AI workflows.
The library offers flexibility, performance, and usability, supporting batch processing, multimodal generation, and real-time viewer support.
It focuses on quality by enabling structured outputs, fine-tuned caching, asynchronous execution, and integration with leading language model providers.
Curator allows users to define custom prompting and parsing logic using Python classes and pydantic models, ensuring conformity to expected schemas.