NEWS

geokit 0.0.2

Added three vignettes: case studies for diabetic nephropathy and urothelial cancer meta-analysis workflows, and a geo_chat() guide for LLM-assisted exploration of GEO metadata.
Added a public Agent Skills file for AI agents, published with the pkgdown site at https://wanglabcsu.github.io/geokit/skills/geokit/SKILL.md.
Fixed parse_sample_data() ignoring its sep argument when parsing characteristic key-value pairs, silently skipping non-contiguous characteristics_ch columns, and crashing on custom fields names without a ch<N> suffix (they now prefix parsed columns with the field name).
Fixed plain-FTP downloads (ftp_over_https = FALSE) being deleted and reported as failures after a success-detection regression.
Fixed a guaranteed crash in geo_matrix() when a platform annotation download fails, and clarified the error when sample metadata lacks platform_id.
CI resolves querychat from the posit-dev R-universe after its CRAN archive; workflows and the skill file keep working unchanged.

This is the first release of geokit on CRAN.

Added a unified interface for resolving, downloading, and parsing NCBI GEO resources, including SOFT files, Series Matrix files, supplementary files, and metadata.
Added vectorized GEO accession handling and parallel downloads with progress reporting.
Added geo_search() for searching GEO records through NCBI Entrez.
Added S4 classes and accessors for working with GEO DataSets, Series, Samples, and Platforms.
Added geo_chat(), geo_console(), and geo_shiny() for exploring GEO metadata with natural-language queries.
Added a Rust-based backend for fast GEO URL resolution and SOFT parsing.