# This CITATION.cff file was generated for the oxo-call project.
# See https://citation-file-format.github.io/ for the specification.
cff-version: 1.2.0
title: "oxo-call: Documentation-grounded LLM command generation for bioinformatics"
message: >-
If you use this software in your research, please cite it using
the metadata from this file.
type: software
authors:
- family-names: Wang
given-names: Shixiang
affiliation: Traitome
orcid: "https://orcid.org/0000-0001-9855-7357"
- name: Traitome
version: 0.11.0
date-released: "2026-04-12"
license: LicenseRef-oxo-call-dual
url: https://github.com/Traitome/oxo-call
repository-code: https://github.com/Traitome/oxo-call
keywords:
- bioinformatics
- command-line
- LLM
- natural-language
- workflow
- genomics
- documentation-grounding
- benchmark
- skill-augmented-prompting
abstract: >-
oxo-call is a documentation-grounded command-generation system for
bioinformatics that augments large language models (LLMs) with structured
tool documentation and curated domain-specific skill files. The system
implements a docs-first grounding strategy that loads complete tool
documentation and expert-curated examples into the LLM prompt before
command generation. Systematic evaluation across 286,200 trials spanning
159 tools and 44 analytical domains demonstrates 25-47 percentage-point
improvements in exact-match accuracy over bare LLM baselines, with all
models exceeding 99.5% exact match. The benchmark framework includes
per-category analysis with 95% confidence intervals, a seven-category
error taxonomy, and Cohen's h effect sizes.