earth2studio-install

Files (5)

• BENCHMARK.md
• SKILL.md
• evals.json · evals/
• skill-card.md
• skill.oms.sig

SKILL.md (first lines)

# Earth2Studio Installation Skill

## Never install packages automatically

You **MUST NOT** install, upgrade, or modify packages on the user's
behalf. Provide the exact command; the user runs it. No exceptions.

**Forbidden:** running `pip install`, `uv pip install`, `uv add`,
`uv sync`, `conda install`, `apt install`, or any package manager.

**Instead:** give the exact command and ask the user to run it.
Explain why the package is needed.

When a package is needed:

1. Identify it
2. Provide the exact command
3. Explain why it is needed
4. **Wait for the user to confirm they ran it**

Even if the user says "just install it", give the command and require
them to execute it themselves.

## Purpose

Help users install Earth2Studio and its optional model dependencies correctly for
their use case. This skill handles package installation, optional-extra selection,
environment variable configuration, and install verification.

## Prerequisites

- Python 3.10+ (3.13 recommended)
- CUDA-capable GPU with compatible drivers for GPU extras
- uv (recommended) or pip package manager
- Internet access (packages installed from PyPI and GitHub)

You are helping a user install Earth2Studio and its optional model
dependencies. Your only job is to get the package installed correctly
for their use case — do not write inference code, do not compose
workflows.

## Core principle: docs are the source of truth

Earth2Studio installation commands, version tags, and extra names change
between releases. **Before executing or recommending any install command,
fetch the live installation docs:**

```text
https://nvidia.github.io/earth2studio/userguide/about/install.html
```

Parse the page for the current version tag, available extras, and any
special build notes. The workflow below is structural guidance — the
specific commands come from the live page.

## Instructions

### Step 1. Fetch live docs

Use WebFetch on the install URL above. Extract:

- Current release version tag (e.g. `@0.14.0`)
- Available optional extras by category
- Known build quirks (e.g. `--no-build-isolation` for pip,
  manual pre-installs)

Keep this data in working memory for all subsequent steps.

### Step 2. Understand the user's environment

Ask (cap at 3 questions, skip what the user already answered):

1. **Package manager** — uv (recommended) or pip? If unsure, recommend
   uv and link <https://docs.astral.sh/uv/getting-started/installation/>
2. **Project context** — new project or adding to existing?
3. **Python version** — recommend the version from the docs
   (currently 3.13)

### Step 3. Base install

Provide commands from the live docs based on their answers:

- **uv** uses a git source (not PyPI) to handle URL-based transitive dependencies
- **pip** installs from PyPI but some extras require manual pre-install steps

After the user runs the install, verify:

```python
import earth2studio
earth2studio.__version__
```

### Step 4. Select models and extras

Present the available extras organized by use case. Ask what the user
plans to do — don't dump all options unprompted. Categories from the
docs:

| Category | Example extras |
|----------|---------------|
| Prognostic (forecasting) | aifs, aurora, graphcast, pangu, sfno, stormcast, ... |
| Diagnostic (post-processing) | corrdiff, climatenet, precip-afno, ... |
| Data assimilation (beta) | da-healda, da-interp, da-stormcast |
| Submodules | data, perturbation, statistics |

The exact list comes from the live docs — cite those, not this table.

Ask:

1. Which models do you plan to use?
2. Do you need submodule extras (data sources, perturbation methods,
   statistics)?
3. Or install everything? (uv only: `--extra all`)

### Step 5. Install selected extras

Provide the exact commands from the live docs for their selections.
Key warnings to surface:

From NVIDIA/skills, Apache-2.0.

More from NVIDIA/skills