Workspace discovery

Candidate files

Conda Code searches local workspaces for:

1. conda.toml

2. pixi.toml

3. pyproject.toml

The order above is the priority when more than one candidate exists in a directory. At most one workspace is published for that directory, and only that selected manifest contributes tasks. See Workspace tasks.

When another provider owns a higher-priority manifest in an unregistered directory, Find workspaces does not offer a lower-priority manifest from the same directory. A manifest that fails Conda Code’s own validation does not prevent consideration of the next candidate.

The search excludes .git, .conda, .pixi, and node_modules.

Python project registration

The candidate directory must be an exact Python project root registered with Python Environments. A manifest elsewhere in the VS Code window is ignored.

Conda Code first asks for a complete workspace snapshot:

conda workspace --file /path/to/manifest info --json --packages

When supported, that one response supplies the manifest identity, environment composition, dependency provenance, prefixes, and installed package records. Conda Code otherwise uses the conda-workspaces 0.7 commands for the same discovery. Only installed environments are published. A declared but uninstalled environment becomes available through the environment creation flow.

Manifest validation failures reported by conda-workspaces appear in the Problems panel on the affected manifest. Conda Code uses the backend message and location when available. Current conda-workspaces releases report these as file-level diagnostics. Correcting, removing, or changing ownership of the manifest clears the diagnostic on the next refresh. An invalid manifest is not retried through the older discovery commands.

Pixi Code rule

When renan-r-santos.pixi-code is installed, Conda Code skips:

• every pixi.toml

• a pyproject.toml containing a [tool.pixi] table

conda.toml remains eligible.

Skipped manifests contribute neither environments nor tasks.

Prefix routing

Each published workspace environment records its manifest, Python project, environment name, features, direct dependencies, installed packages, prefix, and Python path. A complete snapshot can also report the exact environment, feature, platform, or top-level default declaration for a direct dependency. The environment description and tooltip show the workspace-wide lockfile status and its reported reason when conda-workspaces provides them.

When exactly one manifest reports a prefix, environment and package changes use workspace commands. When several manifests report the same prefix, Conda Code hides it and refuses changes.

Refresh

Conda Code watches:

• conda.toml

• pixi.toml

• pyproject.toml

• conda.lock

File changes and refreshes scoped to a registered Python project inspect only the affected workspaces. Conda Code still reconciles the full published environment set after that inspection.

The manual Conda Code: Refresh Environments command clears package data and refreshes all environment and task sources. It also starts forced conda-information enrichment in the background.