The notebook.link team is building "serverless scientific computing" on top of the modern Conda ecosystem & the prefix.dev package server. Read the guest blog on how they shipped interactive WASM for Python, R, C++, Fortran and more: prefix.dev/blog/server...
Find the documentation here:
pixi.prefix.dev/latest/buil...
No more duplicate dependencies in pyproject.toml!
Pixi Python packages can now set `ignore-pypi-mapping = false` to automatically convert `pyproject.toml` dependencies into conda package requirements.
The goal is no configuration, this is a step in that direction!
Do you need more control over the packages that will be installed in your Pixi workspaces?
You can now define [constraints], which allow you to specify which versions are allowed when a package is installed.
Yes, you can do `pixi global install --git github.com. --branch foobar` or `pixi add --git ...`
Just run `pixi build` and the package will come rolling out.
Don’t like writing recipes? With Pixi packages, it’s a lot less work to package up remote projects!
This is all the configuration needed to build the ruff project into a conda package with Pixi. 🤯
Of course! Do you want a demo sometime soon?
Our documentation has been enhanced to cover all the debugging scenarios: rattler-build.prefix.dev/latest/debu...
Let us know how it goes if you try it out!
We want to make it easier to build packages!
For that, we have a unified debug experience in `rattler-build` now. You can run `debug setup`, enter a `debug shell`, and even add some dependencies into the host environment with `debug host-add`.
This is very powerful for the R ecosystem as often, system level dependencies are underspecified. CRAN packages generally need to be built from source. pixi-build-r makes that super smooth - you don't even have to release to CRAN! You can just depend on another project from git.
Pixi can now natively build R packages - for your own packages or thirdparty packages from CRAN or Github. The new `pixi-build-r` build backend makes it possible. pixi-build-r will install compilers, system-level dependencies and R / r-packages from @condaforge.
Exciting news!
We're thrilled to welcome Pavel as our first Pixi core maintainer from outside Prefix! With his extensive contributions, we have full confidence in his stewardship of the project. Thank you for your past work, and here’s to an even stronger collaboration ahead!
We're looking forward to more converted recipes on tdejager.github.io/are-we-reci... 🚀
Low level packages such as `zlib`, `zstd` or large packages such as Pytorch are often "split" into headers, libraries and man-pages. The staging output makes it possible to build once, then split. Slice and dice the packages.
Discuss it here: github.com/conda/ceps/...
The RFC period has started on the "staging cache" CEP for the v1 recipe format. This might sound like gibberish but is a powerful feature that will allow us to convert more (complicated!) packages on @condaforge to the V1 recipe format!
The command to try this yourself:
pixi exec -s newton-all -s pytorch -s usd-exchange python -m newton.examples robot_policy
You can now install the Open-Source physics engine, Newton, with Pixi! 🤖
Getting a full humanoid robot simulator has never been easier, just run one command!
Check out the original announcement: github.com/newton-phys...
#robotics #NVIDIA #warp #newton
We couldn't have done it without the support of many great people including Dan Yeaw and Jakov Smolić from @anacondainc who is also sponsoring this work. We are grateful to Axel Obermeier
for his input in design discussions & the many people who convert feedstocks to recipe v1
A code snippet shows details of a Flask package source with its attestation that will be validated by Rattler-Build
We also worked on making the conda ecosystem supply-chain security aware. This is why we added experimental support for validating attestations of sources.
Recipe v1 YAML for a software package named "foo," detailing builds and requirements for staging, library, and development packages.
One of the most requested features is to have a common cache for package builds similar to what conda-build provides. Our goal was to have a solid design for that and after many iterations we are now confident with our approach.
Steps for using `rattler-build debug` to troubleshoot and fix failed builds by entering environments, adding dependencies, and generating patches.
Iterating on recipes can be painful if you always have to iterate from scratch. This is why introduced `rattler-build debug` that lets you directly jump into your fail build without re-building environments or artifacts.
A coding interface showcasing a recipe YAML configuration for the numpy package, including version and environment details.
For a long time, we wished for an online playground so we can quickly check how a recipe renders and what packages produces. Thanks to web assembly, we are now able to provide exactly that: playground.rattler.build/
Code snippet for where you assemble a package from scratch without any recipe or build step.
For interactive development, Python is still king, which is why made sure to publish a convenient and powerful Python wrapper. We even expose functionality not found on the CLI, like the package assembler that lets you assemble packages directly without a recipe.
We also extracted library crates so you can create your own application based on Rattler-Build code. With this crate, for example, you can parse YAML files with the same jinja functionality built-into recipe V1 files: crates.io/crates/ratt...
Error message from a script indicating an undefined variable issue in a Jinja template within a YAML file.
Our parser is now stricter. This means great error messages for undefined variables and ambiguous match specs. Unfortunately that also means that some recipes that used to work, have to be adapted to work with the newest Rattler-Build version.
Paxton holds a tool and poses against a bright yellow background, promoting Rattler-Build updates.
We improved our package builder Rattler-Build a lot over the past months, so we wrote a blog post about it!
prefix.dev/blog/whats-...
Your Github Actions will be simple, cached and fully reproducible, too.
You can install k9s, helm, terraform, linters ... every version transparently locked in a pixi.lock file - on macOS, Windows & Linux.
Pixi tasks give your team consistent entry points with almost no setup time.
You thought Pixi was only for data science and Python?! Nope! Pixi is the perfect tool to manage your Kubernetes deployments, turbocharged by the github-releases channel we can offer many of the most used packages!
No more curl-pipe-to-bash.
Read more: prefix.dev/blog/kubern...
