Conda Code

Conda Code provides conda environment and package management through the Python Environments extension. It defines its own conda provider for regular environments and uses conda-workspaces as a project-aware layer. It also publishes conda-workspaces tasks through the native VS Code task interface.

The registered environment manager and package manager ID is jezdez.conda-code:conda.

Read the Conda Code documentation for tutorials, how-to guides, reference material, and explanations.

Install

Install Conda Code from the Visual Studio Marketplace or run:

code --install-extension jezdez.conda-code

Python Environments is installed as an extension dependency.

Features

Regular conda environments

These are standard conda prefixes. Conda Code discovers them through conda info --json and manages them with core conda commands. The conda env command group refers to the same prefixes but is not used by Conda Code.

• Discover the base environment and registered named or prefix environments from conda info --json
• Read Python metadata directly from each prefix without starting its interpreter
• Keep environments without Python visible so they can be repaired or managed
• Create named environments
• Create named environments from project-root environment.yml, environment.yaml, and exact package inputs
• Create a project .conda prefix only when selected explicitly during interactive creation
• Remove named environments and project .conda prefixes while protecting base installations and other prefix environments
• List, install, update, and remove packages with the configured conda executable
• Show conda records, including conda-pypi packages, and omit raw pip-only distributions
• Resolve environments by prefix or Python executable
• Persist the manager-wide and project selections in the current VS Code workspace
• Run Python directly or activate it with hooks from the configured conda root

Conda workspaces

• Find conda.toml, pixi.toml, and compatible pyproject.toml manifests at registered Python project roots
• Validate candidates through conda workspace info --json
• Publish installed workspace environments, including environments that do not yet contain Python
• Install declared environments and clean their prefixes without deleting the manifest declaration
• Create a new conda.toml project through quick create
• List installed packages and distinguish direct from transitive dependencies
• Add conda dependencies for environments backed by zero or one feature
• Refuse workspace dependency removal, package upgrades, and additions to composite environments
• Expose declared workspace tasks through the native VS Code task interface while delegating discovery and execution to conda task
• Run a task in the selected Conda Code workspace environment when it belongs to the same manifest
• Refresh when a manifest or conda.lock changes

Conda Code publishes each prefix once. When exactly one workspace reports a prefix, environment and package changes use the workspace commands instead of the regular conda commands.

When multiple workspace manifests report the same prefix, Conda Code hides that prefix. Removal and package operations refuse stale references to that prefix until only one manifest reports it.

Other tools

conda-pypi

conda-pypi packages have ordinary conda records, so Conda Code lists those records with other conda packages. Regular package changes continue to use conda. Conda Code does not list raw pip-only distributions.

conda-workspaces and conda-pypi handle workspace [pypi-dependencies]. Package changes from the Python Environments view apply to conda dependencies.

Workspace PyPI dependencies require conda 26.5 or newer, the Rattler solver, the conda-pypi channel, and flexible channel priority. See the conda-pypi setup.

conda-lockfiles

With conda-lockfiles installed in the configured conda base environment, Conda Code can create exact named environments from project-root conda-lock.yml and conda-lock.yaml files. Conda-lockfiles also lets conda consume pixi.lock. Conda Code leaves that file with its Pixi project instead of offering it as a standalone environment input. Follow the conda-lockfiles installation instructions using the channels configured for your conda distribution.

conda-global

conda-global manages isolated tool environments and PATH trampolines. Conda Code skips prefixes below its active environment root:

1. $CONDA_GLOBAL_HOME/envs when configured
2. ~/.conda/global/envs after migration or for a new installation
3. ~/.cg/envs for an existing legacy installation

Pixi Code

Conda Code and Pixi Code can be installed together. Conda Code excludes .pixi/envs from regular conda discovery. When Pixi Code is installed, Conda Code ignores pixi.toml and [tool.pixi] projects during workspace discovery and does not create conda.toml in those projects. It continues to handle conda.toml and other conda-workspaces projects.

Python Environments built-in conda provider

Python Environments currently registers ms-python.python:conda unconditionally. Its public API does not provide a supported way for another extension to replace, hide, or disable that manager, and it does not deduplicate prefixes across managers.

Conda Code does not call or delegate to that implementation. Until Python Environments allows another extension to replace the built-in provider, both conda branches can appear in the environment view. Select jezdez.conda-code:conda as python-envs.defaultEnvManager or as the envManager for a python-envs.pythonProjects entry to use Conda Code for project operations.

Creation behavior

• An existing discovered conda workspace takes precedence over project creation inputs and offers its uninstalled declarations
• A project-root environment.yml or environment.yaml is a one-time CEP 24 input to a regular named environment
• A project-root explicit.txt is an exact CEP 23 input
• With conda-lockfiles installed, conda-lock.yml and conda-lock.yaml are exact inputs
• Quick Create uses the single recognized input and derives an available name from the project directory
• Quick Create fails when several recognized inputs exist
• Interactive creation lets the user choose an input and asks for the environment name
• Interactive project creation also offers a conda workspace, an explicit project .conda prefix, or a named environment
• Without an input, Quick Create creates conda.toml and installs its default workspace environment
• Quick Create refuses to add conda.toml to a Pixi project when Pixi Code is installed
• Global and multi-project creation creates a named environment
• Quick global and multi-project creation generates an available name without a prompt
• Regular environments created from package selections and new workspaces include Python unless the requested package list already contains a Python specification
• Adding packages while installing an existing workspace declaration requires zero or one feature
• Exact inputs refuse additional creation packages and disable configured default packages

For file-based creation, Conda Code runs conda create --name NAME --file FILE from the project root. The selected name overrides a YAML name or prefix. The input is not watched or synchronized after creation.

Execution and activation

Python execution uses the absolute interpreter inside the prefix. For Bash, Zsh, POSIX sh, Fish, PowerShell Core (pwsh), Git Bash, and Command Prompt, regular environments advertise shell-specific activation commands that load hooks from the configured conda root before activation. Other shells use direct interpreter execution.

Workspace environments use direct interpreter execution. Conda Code does not advertise conda workspace shell as terminal activation because that command starts a nested blocking shell.

Requirements

• Visual Studio Code 1.118 or newer
• Python Environments
• conda 26.3 or newer
• A trusted VS Code window
• A local file project registered with Python Environments for workspace and project-input features
• conda-workspaces 0.7 or newer for workspace features
• conda 26.5 or newer for workspace PyPI dependencies
• conda-lockfiles 0.2 or newer for conda-lock.yml and conda-lock.yaml creation inputs

Set conda-code.condaExecutable when the desired conda is not available through CONDA_EXE, the Python extension's python.condaPath setting, or PATH.

Usage

Install Conda Code and open the Python Environments view. The Conda Code provider lists regular conda environments immediately. Opening a registered Python project with a supported workspace manifest adds its installed workspace environments to the same provider. Tasks declared by that manifest appear under Tasks: Run Task with the conda-workspaces source. Select one of that workspace's Conda Code environments first to run the task there.

Run Conda Code: Refresh Environments after an external change when automatic manifest refresh is not sufficient. The command also refreshes workspace task discovery.

Development

Use the Node.js release pinned in .nvmrc.

npm ci
npm run typecheck
npm test
npm run docs

Run the Extension launch configuration in an Extension Development Host. Build an installable package with:

npm run vsix