team-galaxy

Galaxy Project extensions for the team multi-agent LLM framework.

team-galaxy adds Galaxy-specific skills, personas, and ready-to-run scenario configs on top of team-core. Installing team-galaxy is all it takes — no environment variables or hardcoded paths required. It integrates with team-core ≥ 0.15.5 via Python entry points:

[!NOTE]

A significant portion of this code and documentation was written with the assistance of GitHub Copilot. All contributions have been reviewed and tested by the maintainers.

Scenarios

Installation

Basic (planemo scenarios, no Galaxy API)

pip install team-galaxy

Planemo must be installed separately and available on PATH:

pip install planemo

With BioBlend (Galaxy API scenarios)

pip install "team-galaxy[bioblend]"

Set your Galaxy credentials:

export GALAXY_URL=https://usegalaxy.org
export GALAXY_API_KEY=<your_api_key>

From source

git clone https://github.com/cumbof/team-galaxy
cd team-galaxy
pip install -e ".[dev,bioblend]"

Quick Start

1. Initialise a scenario

team galaxy init --scenario tool-wrapper-factory

This copies a ready-to-edit YAML config to the current directory. Skill references use short registered names — no paths to edit.

2. Prepare your workspace

Each scenario expects one or two input files in the workspace directory. The generated YAML contains a # Before running: comment describing exactly what to provide.

For the tool-wrapper-factory scenario, create tool_spec.md in the workspace:

# Tool: fastp

conda package: fastp
version: 0.23.4

## --help output

fastp: an ultra-fast all-in-one FASTQ preprocessor
...

## Intended Galaxy datatypes
- Input: fastqsanger, fastqsanger.gz (paired or single)
- Output: fastqsanger.gz (trimmed reads), html (QC report), json (QC metrics)

3. Run

team run tool-wrapper-factory.yaml

Skills

After pip install team-galaxy, use skill names directly in any team YAML:

defaults:
  skills:
    - bioblend        # Galaxy API tools (requires bioblend extra)
    - planemo         # planemo lint / test / autoupdate tools
    - toolshed        # Tool Shed search and metadata tools
    - iuc_standards   # IUC authoring standards (context injection)
    - iwc_checklist   # IWC workflow quality checklist (context injection)
    - gtn_format      # GTN tutorial format specification (context injection)
    - bioconda_guide  # Bioconda packaging guide (context injection)

Available skills

Personas

Galaxy-specific personas are automatically discovered when team-galaxy is installed. No TEAM_PERSONA_DIR needed.

Reference them in any team YAML:

members:
  - name: reviewer
    persona: "@iuc_reviewer"

How it extends team-core

team-galaxy uses the three entry-point groups added in team-core ≥ 0.15.5:

team-core plugin API
├── team.skills          ← bioblend, planemo, toolshed, iuc_standards, …
├── team.persona_dirs    ← team_galaxy/personas/ (auto-merged)
└── team.commands        ← team galaxy <subcommand>

This means:

• Skill names work in any team YAML, not just team-galaxy ones.
• Personas are available via @name without any env var.
• team galaxy --help is always accessible once team-galaxy is installed.

CLI reference

team galaxy --help

Commands:
  init       Copy a scenario template to a directory, ready to run.
  scenarios  List available scenario templates.
  skills     List available Galaxy skill names and descriptions.
  personas   List available Galaxy personas with descriptions.

# List scenarios
team galaxy scenarios

# Initialise a scenario in the current directory
team galaxy init --scenario bioblend-analysis

# Initialise in a specific directory
team galaxy init --scenario tool-wrapper-factory --output-dir ~/my-project/

Development

pip install -e ".[dev,bioblend]"
pytest -q

Integration tests (require a live Galaxy server and planemo):

pytest -m integration

Contributing

Pull requests are welcome. Please disclose AI assistance in the PR description (e.g. "co-authored with GitHub Copilot"). Before submitting:

• Run pytest -q and ensure all tests pass.
• Add or update tests for any behaviour changes.
• Follow the conventions described in AGENTS.md.