Sdrf:setup

SDRF Setup Workflow

You are guiding the user through installing SDRF skills dependencies. Follow these steps.

In Cursor: The user invokes this by asking "install SDRF dependencies" or similar (no /sdrf:setup slash command). Ensure environment.yml and requirements.txt exist at the workspace root; if not, suggest cloning the full sdrf-skills repo or copying those files.

Step 1: Detect Available Package Managers

Check which package managers are available (run these in the terminal or ask the user):

command -v conda && conda --version
command -v mamba && mamba --version
command -v uv && uv --version
command -v pip && pip --version

• Conda or mamba: Recommended — best for thermorawfileparser (Thermo .raw files) via bioconda
• Pip: Works for sdrf-pipelines and techsdrf; thermorawfileparser requires conda
• uv: Can install Python tools; same limitation as pip for thermorawfileparser

Step 2: Provide Installation Commands

Based on what's available, output the exact commands the user should run.

Option A — Conda (recommended)

# From the sdrf-skills project directory:
conda env create -f environment.yml
conda activate sdrf-skills

If using mamba (faster):

mamba env create -f environment.yml
conda activate sdrf-skills

Option B — Pip (venv)

# From the sdrf-skills project directory:
python -m venv .venv
source .venv/bin/activate   # On Windows: .venv\Scripts\activate
pip install -r requirements.txt

Note: With pip, thermorawfileparser is not available (not on PyPI). For Thermo .raw files, use conda.

Option C — uv

uv venv .venv
source .venv/bin/activate
uv pip install -r requirements.txt

Step 3: Verify Installation

After the user runs the commands, ask them to verify:

parse_sdrf --version
techsdrf --version

If both succeed, setup is complete.

Step 4: Optional — Spec Submodule

If the user cloned without submodules or wants the latest spec:

git submodule update --init --recursive
# To pull latest:
git submodule update --remote --recursive

Step 5: Optional — MCP Servers

For full SDRF annotation (PRIDE, OLS, PubMed), the user needs MCP servers configured. Tell them to check their host's MCP configuration:

• PRIDE MCP — project metadata, OLS, EuropePMC
• PubMed — literature, PMC full text
• bioRxiv — preprint search (optional)
• Consensus — evidence search (optional)

For Europe PMC full text, prefer the local normalizer over raw XML inspection: python scripts/europepmc_fulltext.py PMC_ID --format text This keeps methods/results/discussion easier for LLMs to interpret and preserves canonical links plus detected accessions in JSON mode.

Summary Output

Provide a clear summary:

1. Package manager detected: conda / pip / uv
2. Commands to run: (copy-paste block)
3. Verify: parse_sdrf --version, techsdrf --version
4. Next: Run /sdrf:annotate PXD###### or /sdrf:validate yourfile.sdrf.tsv

If User Passes "check"

When the user invokes /sdrf:setup check, run the verification step and report status:

• parse_sdrf: ✓ or ✗
• techsdrf: ✓ or ✗
• spec/ submodule: present and init'd or not
• Suggest fixes for any missing items

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: bigbio
• Source: bigbio/sdrf-skills
• License: MIT
• Homepage: https://sdrf.quantms.org

Install and usage instructions live in the source repository linked above.