feat(providers): add conda provider with micromamba, conda and mamba …

[flashpoint493]

Summary

This PR adds a new Conda provider to vx, enabling package, dependency, and environment management for Python and scientific computing (including CUDA, PyTorch, TensorFlow).

What does this PR change?

New Conda Provider (crates/vx-providers/conda/)

• micromamba: Minimal standalone mamba - single binary, no installer needed (recommended)
• conda: Full Conda installation via Miniforge
• mamba: Fast package manager bundled with Miniforge

Conda-tools Isolation Support

• New ~/.vx/conda-tools/ directory for isolated conda environments
• Follows the same pattern as existing pip-tools and npm-tools
• PathManager and PathResolver support for conda-tools

Why?

Conda/Mamba is essential for ML/AI development workflows that require:

• CUDA toolkit and GPU libraries
• PyTorch, TensorFlow with GPU support
• Scientific computing packages (NumPy, SciPy, etc.)
• Cross-platform binary package management

This fills a gap in vx's toolchain support for data science and machine learning developers.

Type

• feat
• fix
• docs
• refactor
• perf
• ci/chore

Checklist

• PR title follows Conventional Commits (e.g., feat: add X)
• Tests added/updated when necessary
• Docs updated (README/docs/CHANGELOG as needed)
• CI green

Breaking changes

• No breaking changes
• Breaking changes (describe):

Additional context

Directory Structure

~/.vx/
├── store/              # runtimes (node, go, micromamba...)
├── npm-tools/          # npm package tools
├── pip-tools/          # pip package tools
└── conda-tools/        # 🆕 conda package tools (isolated environments)
    └── <package>/<version>/env/

Usage Examples

Basic Usage

# Install micromamba (recommended)
vx install micromamba

# Or install specific version
vx install micromamba@2.5.0-1

# List available versions
vx list micromamba

Create & Use Conda Environments

# Create environment with PyTorch + CUDA
vx micromamba create -n ml python=3.11 pytorch pytorch-cuda=12.1 -c pytorch -c nvidia

# Run command in environment
vx micromamba run -n ml python train.py

# Activate environment (interactive)
vx micromamba shell -n ml

Project-level Configuration (vx.toml)

[tools]
micromamba = "latest"

[env]
# Project-level isolation - conda envs stored in project directory
MAMBA_ROOT_PREFIX = "${VX_PROJECT_ROOT}/.mamba"

[scripts]
setup = "micromamba create -n dev python=3.11 pytorch -c pytorch -y"
train = "micromamba run -n dev python train.py"

Why Micromamba?

Feature	micromamba	conda/mamba
Installation	Single binary	Requires installer
Size	~10MB	~400MB+
Speed	Very fast	Fast (mamba) / Slow (conda)
Compatibility	Full conda compat	Native

Files Changed

Path	Description
crates/vx-providers/conda/	New provider (6 files)
crates/vx-paths/src/	conda-tools path support
crates/vx-runtime/src/	conda-tools trait methods
crates/vx-cli/	Register conda provider

Platform Support

Platform	micromamba	conda/mamba
Linux x86_64	✅	✅
Linux aarch64	✅	✅
macOS x86_64	✅	✅
macOS arm64	✅	✅
Windows x86_64	✅	✅

[flashpoint493]

Why Micromamba?

Micromamba is the recommended choice because it fits vx's installation model perfectly:

✅ micromamba: Download .tar.bz2 → Extract → Ready to use
❌ conda/mamba: Download .sh/.exe installer → Run installer → Then use

vx's design philosophy is download → extract → use, and micromamba is the only conda-compatible tool that supports this pattern as a standalone binary.

Note: All three tools (micromamba, conda, mamba) are fully compatible - they use the same package repositories (conda-forge, anaconda) and can manage the same environments.

[flashpoint493]

fix updated