Hybrid Topic 0.1.0

Hybrid Topic 0.1.0 provides a Python/Jupyter workflow for discovering named topics with a language model, assigning documents through semantic embeddings and graph propagation, and reusing the fitted model for new texts.

Features include optional analytical guidance, supplied topic codebooks, bounded residual topic discovery, topic/document tables, CSV/JSON export, and model save/load. The default initial prompt asks for a compact set of recurring topics.

This release includes the English and Chinese manuscripts by Jiashuo Ren, University of Stirling, the compact/neutral prompt comparison, and text-free numerical prediction evidence with scoring scripts. The manuscript's arXiv link will be added after publication.

Validation: 121 tests completed with two environment/fixture skips; the wheel installed in a clean Python 3.12 environment; the offline lifecycle example passed; installed-package scoring reproduced 240 configurations and the 48 published summary groups. Provider evidence and platform scope are recorded in the README and docs/RELEASE_VALIDATION.md.

Install the wheel below for the base package, or clone the repository and follow its README for OpenAI/embedding extras and examples. Software is MIT licensed; article and dataset terms are described in NOTICE.md.