Create a multi-platform lockfile#

This tutorial starts with a small environment.yml, resolves it for Linux and macOS, and writes a pixi.lock. Along the way, it shows the difference between conda-presto’s native result and an exporter result.

Before you start#

Install conda-presto with Quick start, then verify the subcommand:

conda presto --help

Create a working directory containing this environment.yml:

name: science
channels:
  - conda-forge
dependencies:
  - python=3.13
  - numpy
  - pandas

conda-presto reads this file through conda’s environment specifier registry. It does not create the science environment.

Inspect the native solve result#

First resolve only Linux and keep the native JSON:

conda presto -f environment.yml -p linux-64 > linux-result.json

Confirm that the solve succeeded:

jq '.[0] | {platform, package_count: (.packages | length), error}' linux-result.json

The result has one platform, a nonzero package count, and an error value of null. Native JSON can represent a failure for one platform while preserving successful results for the others.

Add another platform#

Request Linux and Apple silicon macOS:

conda presto -f environment.yml \
  -p linux-64 \
  -p osx-arm64 > multi-platform-result.json

Inspect the result order and package counts:

jq -r '.[] | "\(.platform): \(.packages | length) packages"' multi-platform-result.json

The output order follows the -p arguments even though the solves may run in parallel.

Write a pixi.lock#

Run the same solve through the pixi-lock-v6 exporter:

conda presto -f environment.yml \
  -p linux-64 \
  -p osx-arm64 \
  --format pixi-lock-v6 > pixi.lock

Exporter output is a single document, so every requested platform must solve successfully. The native JSON path is the better diagnostic form when you need per-platform errors.

Confirm that the lockfile contains both platforms:

rg 'linux-64|osx-arm64' pixi.lock

Reuse the lockfile#

conda-presto can read the lockfile through the same conda plugin system:

conda presto -f pixi.lock \
  -p linux-64 \
  --format conda-lock-v1 > conda-lock.yml

Because the requested platform is already present and both formats are lockfiles, this conversion reuses package records without solving again.

What you learned#

conda presto resolves without installing packages.
Native JSON keeps one result per platform and can represent partial failure.
--format uses conda exporter plugins and requires successful environments.
Lockfile-to-lockfile conversion can skip solving when the input covers every requested platform.

For individual CLI tasks, see Resolve from the CLI and Transcode lockfiles without solving. Exact flags are in CLI reference.