Pixi is a conda environment manager that creates reproducible, lock-file-backed environments with cross-platform support and multi-language dependency resolution. It combines package management from both conda-forge and PyPI simultaneously, coordinating their dependency graphs to prevent conflicts, while pinning every transitive dependency to exact versions in a cross-platform lock file for bit-for-bit identical environments across machines.
The project distinguishes itself by also functioning as a cross-platform task runner that executes user-defined commands and multi-step pipelines inside isolated project environments on Linux, macOS, and Windows. It provides a workspace manager for handling multiple sub-projects with their own manifests and dependencies within a single repository, and supports installing command-line tools in isolated environments so they are available system-wide without conflicting with project dependencies. A dedicated GitHub Actions action caches and restores conda environments using lock-file hashes to speed up CI workflows.
The tool manages multiple isolated environments within a single manifest, each with its own dependency set for different workflows or tools. It supports initializing new project directories with a manifest file and directory structure from a single command, and allows defining cross-platform commands with arguments, dependencies, working directory, and environment variables that execute in the workspace. Authentication against package registries is supported using tokens, username and password, conda tokens, S3 keys, or keyring.
Documentation covers installation, project initialization, dependency management, environment composition, task execution, CI integration, and global tool installation across all supported platforms.
