Conda Package Manager Guide¶

Prism uses conda as the primary package manager for research environments, providing reliable, cross-platform package management for Python, R, and scientific computing.

🎯 Why Conda for Research?¶

✅ Research-Optimized¶

Scientific Packages: Comprehensive ecosystem (conda-forge, bioconda)
Reproducibility: Environment isolation and dependency management
Cross-Platform: Consistent across Windows, macOS, Linux, ARM64
GPU Support: Native CUDA, PyTorch, TensorFlow integration

✅ Prism Integration¶

Smart Defaults: Automatically selected for Python/R templates
Manual Override: --with conda for explicit control
Optimized Installation: Miniforge for fast, reliable setup
Multi-Architecture: Native ARM64 and x86_64 support

🚀 Usage Examples¶

Basic Usage (Automatic)¶

# Conda automatically selected for Python/R templates
prism workspace launch python-research my-analysis
prism workspace launch r-research stats-project

# Templates detect scientific packages and choose conda
prism workspace launch neuroimaging brain-study

Explicit Conda Selection¶

# Force conda package manager
prism workspace launch python-research my-project --with conda

# Combine with other options
prism workspace launch python-research gpu-training --with conda --size GPU-L --volume shared-data

Advanced Usage¶

# Dry run to see conda installation script
prism workspace launch python-research test --with conda --dry-run

# Launch with specific conda environment
prism workspace launch r-research stats-work --with conda --storage L

📦 Supported Package Types¶

Python Packages¶

packages:
  conda:
    - python=3.11
    - jupyter
    - numpy=1.24.3
    - pandas=2.0.3
    - matplotlib=3.7.1
    - scikit-learn=1.3.0
    - pytorch=2.0.1
    - tensorflow=2.13.0

R Packages¶

packages:
  conda:
    - r-base=4.3.0
    - rstudio
    - r-tidyverse
    - r-ggplot2
    - r-dplyr
    - r-shiny

Scientific Computing¶

packages:
  conda:
    - numpy
    - scipy  
    - matplotlib
    - jupyter
    - pandas
    - seaborn
    - plotly

🔧 How Conda Integration Works¶

1. Template Detection¶

Prism automatically selects conda when templates contain: - Python data science packages (numpy, pandas, jupyter)
- R packages (r-base, tidyverse, rstudio) - Scientific computing libraries (scipy, matplotlib)

2. Installation Process¶

# 1. Download and install Miniforge  
wget -O /tmp/miniforge.sh "$MINIFORGE_URL"
bash /tmp/miniforge.sh -b -p /opt/miniforge

# 2. Install packages via conda
/opt/miniforge/bin/conda install -y python=3.11 jupyter numpy pandas

# 3. Configure environment for users
echo 'export PATH="/opt/miniforge/bin:$PATH"' >> ~/.bashrc
/opt/miniforge/bin/conda init bash

3. Service Integration¶

Jupyter: Automatically configured with conda environment
RStudio: R packages available through conda integration
Custom Services: Access to conda-installed packages

🎛️ Environment Configuration¶

Multi-User Setup¶

# Each user gets conda access
sudo -u researcher /opt/miniforge/bin/conda init bash
echo 'export PATH="/opt/miniforge/bin:$PATH"' >> /home/researcher/.bashrc

# Shared conda installation at /opt/miniforge
# User-specific environments in ~/.conda/envs/

Package Management¶

# Install additional packages
conda install package-name

# Create custom environments  
conda create -n myproject python=3.11 pandas numpy
conda activate myproject

# Export environment for reproducibility
conda env export > environment.yml

📊 Performance Benefits¶

✅ Optimized for Research¶

Fast Solving: Miniforge uses libmamba for faster dependency resolution
Pre-compiled: Binary packages avoid compilation time
GPU Acceleration: Native CUDA toolkit integration
ARM64 Native: Apple Silicon optimization

✅ Prism Optimizations¶

Multi-Architecture: Smart ARM64/x86_64 detection
Package Caching: Reduced installation time for common packages
Environment Reuse: Efficient environment setup across instances

🛠️ Troubleshooting¶

Common Issues¶

Package Installation Fails¶

# Update conda first
conda update -n base -c defaults conda

# Clear cache if needed
conda clean --all

# Use conda-forge channel
conda install -c conda-forge package-name

Environment Issues¶

# Reinitialize conda
conda init bash
source ~/.bashrc

# Fix PATH issues
export PATH="/opt/miniforge/bin:$PATH"

GPU Package Issues¶

# Install GPU packages explicitly
conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia

# Verify GPU access
python -c "import torch; print(torch.cuda.is_available())"

🔮 Future Enhancements¶

Planned Improvements¶

Mamba Integration: Even faster package solving
Environment Templates: Pre-configured research environments
Package Caching: Instance-level package cache optimization
GPU Optimization: Enhanced CUDA/PyTorch conda integration

Specialized Conda Support¶

Bioconda: Bioinformatics package ecosystem
Conda-Forge: Community-maintained packages
PyPI Integration: Seamless pip package fallback
R Integration: Enhanced R + conda workflow

📚 Resources¶

Conda Documentation¶

Prism Resources¶

Template examples with conda integration
Best practices for research environments
Multi-user conda configuration guides

Summary: Conda provides Prism users with world-class package management for research computing, combining reliability, performance, and comprehensive scientific package ecosystems in a research-optimized platform.