Аннотация
BibexPy v2 is a self-hosted bibliometric data-preparation platform for Web of Science and Scopus, distributed as a single pip-installable package that launches a local web application. Licensed exports are processed entirely on the user's machine, eliminating data-privacy concerns associated with cloud-based tools. The deployment process has been reduced to two commands, and a guided web workflow leads the user from raw exports to analysis-ready datasets. Smart Merge, the record-linkage engine, combines deterministic identifiers with Jaro–Winkler similarity scoring and an optional borderline-review queue to detect duplicates across both databases with high precision. Corpus filtering is treated as a transparent screening stage in which inclusion and exclusion criteria are applied explicitly, evaluated against a live record count, and saved as reusable presets, making the study corpus reproducible and reportable. The harmonization layer addresses the identity and naming fragmentation that distorts bibliometric networks: author-name disambiguation relies on ORCID identifiers as deterministic evidence, falling back to field-similarity measures only when identifier coverage is incomplete, while organization and country harmonization consolidates affiliation variants to canonical forms through Jaro–Winkler clustering and dictionary-based normalization. Metadata enrichment adopts a fetch-once-fill-all strategy across seven authoritative sources, writing only verifiable values to preserve data integrity. A bibliometrically weighted health score quantifies dataset completeness by emphasizing fields critical to downstream analysis. Structured exports in multiple formats ensure interoperability with VOSviewer, Biblioshiny, CiteSpace, Gephi, Zotero, and other science-mapping and reference-management tools. An append-only audit log, snapshot-based recovery, and an automatically generated methodology narrative provide end-to-end provenance aligned with computational reproducibility and FAIR data principles.