prime env install

Install an environment package

Overview

The prime env install command installs environment packages from three sources:

1. Environments Hub - Public and private environments from app.primeintellect.ai
2. Local directory - Environments in your ./environments folder
3. GitHub repository - Environments from the verifiers repo

Usage

prime env install <env> [OPTIONS]

Arguments

env · string · required

Environment identifier. Format depends on the source:

• Hub: owner/name or owner/name@version
• Local: env-name (looks in ./environments/)
• Repo: Use with --from-repo flag

Options

--path · string · default: "./environments"

Path to environments directory for local installations.

Aliases: -p

--from-repo · flag

Install from the verifiers GitHub repository instead of Hub or local.

Aliases: -r

--branch · string · default: "main"

Git branch to install from when using --from-repo.

Aliases: -b

Installation Sources

Environments Hub

Install public or private environments from the Hub:

# Install latest version
prime env install primeintellect/math-python

# Install specific version
prime env install primeintellect/math-python@1.0.0

The Hub provides:

• Versioned releases with semantic versioning
• Private environments (requires authentication)
• Automatic dependency resolution
• Simple index URLs for pip-compatible installation

Local Directory

Install environments from your local workspace:

# Install from ./environments/my_env
prime env install my-env

# Install from custom path
prime env install my-env --path ~/my-envs

Local installation uses editable mode (pip install -e), so changes to the environment code take effect immediately without reinstalling.

GitHub Repository

Install environments directly from the verifiers GitHub repo:

# Install from main branch
prime env install gsm8k --from-repo

# Install from specific branch
prime env install gsm8k --from-repo --branch dev

Examples

Install from Hub

prime env install primeintellect/gsm8k

Output:

Fetching environment details for primeintellect/gsm8k@latest...
Installing primeintellect/gsm8k...
Successfully installed gsm8k

Install Local Environment

# Create and install
prime env init my-env
prime env install my-env

Output:

Installing my-env from ./environments/my_env...
Successfully installed my-env

Install from Repository

prime env install math-python --from-repo

Output:

Installing math-python from verifiers repo (main)...
Successfully installed math-python

Install Specific Version

prime env install primeintellect/math-python@0.1.5

How It Works

Hub Installation

1. Queries the Environments Hub API for environment metadata
2. Retrieves the simple index URL or direct wheel URL
3. Installs using uv pip install with the appropriate index
4. Invalidates Python import caches

Local Installation

1. Locates the environment directory (e.g., ./environments/my_env)
2. Installs in editable mode using uv pip install -e
3. Environment becomes importable immediately

Repository Installation

1. Constructs a git+https URL pointing to the environment subdirectory
2. Installs directly from GitHub using uv pip install
3. Branch can be specified with --branch

Verifying Installation

Check if an environment is installed:

uv pip show my-env

Test the environment:

prime eval run my-env -m gpt-4.1-mini -n 1

Private Environments

For private Hub environments, authentication is required:

1. Option 1: Use prime env pull to download and install locally:

   prime env pull owner/private-env
   cd environments/private_env
   prime env install private-env
   

2. Option 2: Configure authentication for the simple index URL (contact support for details)

Troubleshooting

Environment Not Found

Local environment not found: ./environments/my_env

Solution: Check that the environment directory exists and contains pyproject.toml.

Hub Environment Not Available

Failed to fetch environment details: 404 Client Error

Solution:

• Verify the owner/name spelling
• Check if the environment is private (use prime env pull instead)
• Ensure you’re using the correct version

Installation Failed

Installation failed: ERROR: Could not find a version that satisfies...

Solution:

• Check that all dependencies are available
• Verify your Python version meets requirements (>=3.10)
• Try updating pip: uv pip install --upgrade pip

Module Import Errors

If the environment installs but can’t be imported:

# Invalidate caches
python -c "import importlib; importlib.invalidate_caches()"

# Verify installation
uv pip list | grep my-env

Next Steps

After installation:

1. Run an evaluation:

   prime eval run my-env -m gpt-4.1-mini
   

2. Check environment defaults:

   cat environments/my_env/pyproject.toml
   

3. Modify and test (for local installs):

   • Edit environment code
   • Re-run evaluation (changes take effect immediately)

4. Publish to Hub (for your own environments):

   prime env push my-env
   

Dependency Management

Environments can specify their own dependencies in pyproject.toml:

[project]
dependencies = [
    "verifiers>=0.8.0",
    "numpy",
    "requests",
]

These dependencies are installed automatically when you install the environment.

Uninstalling

To remove an installed environment:

uv pip uninstall my-env