Learnware-LAMDA/Beimingwu

The system implements a four-layer architecture with a core engine handling learnware submission, search, and deployment; a heterogeneous specification engine that merges different specification formats and continuously improves through accumulated learnwares; and support for both semantic and statistical specifications across tables, images, and text. Specification generation and quality inspection occur locally without exposing raw data, while deployment supports both single learnware and multi-learnware combinations through unified APIs in the `learnware` Python package.

122 stars. No commits in the last 6 months.