grace (n) -

1. elegance and beauty of movement, form, or expression
2. a pleasing or charming quality
3. goodwill or favor
4. a sense of propriety and consideration for others ¹

Grace is a new, modern, cloud-native version control system.

Grace is easy to use, easy to understand, and consistently fast. And it's powerful, ready to handle large repositories and large file sizes.

Grace Server scales up by running on Kubernetes and massive PaaS services from large cloud providers.

Grace Client runs in the background, making it ambient, faster, and more valuable to your everyday work as a developer.

Grace connects you with others working in your repository, across the globe, in real-time, enabling new experiences and new ways of sharing.

to be clear:

Grace is new 🧑🏼‍🔬 and alpha-level right now. 🔥🧯 The parts that are implemented work, but there's much more to do. Grace is not ready to be relied on as a production version control system yet. It should not be used for anything other than its own design, development and testing, and for feedback.²

I gave my first conference talk about Grace at NDC Oslo 2023. You can watch it here:

For a list of (mostly imagined) frequently asked questions, please see the Frequently Asked Questions page.

If you'd like to read about some of the design thinking and motivations behind Grace - topics like UX, performance, scalability, monorepos, Git, why F#, and more - please read Design and Motivations.

Stop being afraid of your version control system, and start enjoying it instead.

Grace is easy-to-use, easy-to-understand, and fast, with cool new features that let version control fade into the background while you're working, and help you remember where you were when you get interrupted.

There are fewer concepts to understand in Grace, so learning it is easy, and understanding what it's doing is simple. There's a simple-to-understand grammar, built-in aliases for common gestures, and common-sense defaults.

Grace is powerful source control, minus the fear.

Great UX requires great speed. And Grace is fast.

Grace Server is designed to run on fast, cloud-based PaaS services for incredible scale and performance. Grace uses virtual actors as networked, in-memory data caches to maximize performance.

Grace CLI adds just milliseconds to the server response time for each command.

Grace's GUI apps will be platform-native, with all of the performance and stick-to-your-finger-ness that native apps have always had.

Grace Server can even precompute views and projections that you're likely to want, like diffs and directory contents, and garbage-collect them when they're no longer needed.

Grace CLI includes a command - grace watch - that watches your working directory for changes, and keeps a live connection to Grace Server, connecting you in real-time to your team around the world.

grace watch takes action for you, whether that's automatically uploading and downloading new file versions, processing notifications from other repo users, or running custom actions that you create.

And don't worry about having yet-another application running in the background. grace watch is small and very quiet when nothing is going on.

In Git, commit is an overloaded concept. It can mean:

• "I'm partially done" - local commit after each unit of work is done
• "I'm really done" - ready for the pull request
• "Merge" - merges are called commits, even if they're not.

And then you get the "squash" vs. "don't squash" debate. Sigh.

Grace simplifies this by breaking these usages out into their own gestures and events:

• grace checkpoint - this means "I'm partially done", for you to keep track of your own progress
• grace commit - this is "I'm really done" or "This version is a candidate for promotion"; you'd use a commit for a PR
• grace promote - in Grace, promotions replace merges; a promotion is how Grace moves code from a child branch to a parent branch

and, introducing:

• grace save - this is used by grace watch for saving versions of files between checkpoints, because...

By default, grace watch uploads new file versions after every save-on-disk, along with a new snapshot of the entire directory structure of the repo, including new SHA-256 hashes.

It happens so quickly you don't even notice it.

And it gives you some very cool things, like:

• file-level undo for as far back as your repository allows,
• very fast grace checkpoint, grace commit, and grace promote commands, and,
• a Version History view that will let you to flip through your versions, helping you get remember where you were when you get interrupted, and enabling easy, instant restoration of any of your past changes.

Grace is event-sourced. That means that everything that changes the state of the repository - every save and commit, every branch name change, every everything, is stored as a separate event.

As they're handled, they're sent to an event processor, which can log them in your choice of format and system. You can see the who / what / where / when of everything that happens in your repository, using your favorite event analytics and stream analytics tools, and even set up custom auditing and automation based on those events.³

When running grace watch, Grace uses SignalR to create a live, two-way communication channel between client and server.

This connection allows Grace to do All The Cool Things. Things like connecting you in real-time to everyone else working in your repository. Things like auto-rebasing. Things like watching for events in your repository, notifying you when you want to be notified, and running custom local actions if you want.

Imagine: there's a promotion to main, your branch gets auto-rebased on those latest changes, and then your local unit test suite gets run automatically so you immediately know if there's a problem.

Grace lets you share your code with team members effortlessly, around the world, for those times when you need another set of eyes on it, or just want to show them something cool.

Auto-rebasing keeps you up-to-date and ready to go, because...

Merge conflicts suck. Finding out that you have one, when you thought you were already done with your work, is anxiety-inducing, and one of the most confusing parts of the user interface in Git. Grace helps you eliminate conflicts by keeping your branch up-to-date, automatically.

When your parent branch gets updated, within seconds, grace watch will auto-rebase your branch on those changes, so you're always coding against the latest version that you'll have to promote to.

Most-to-almost-all of the time, you don't even notice it. No errors, no conflicts.

When there is a problem, auto-rebase lets you find out right away.

You can fix it while you're in flow, and skip the conflict later.

And Grace doesn't just tell you there's a conflict. It uses AI to propose a resolution, and lets you accept it with a single click, or request that it tries again with a different proposal. You can even tell Grace to run a CI/CD pipeline on the resolution before proposing it to you.

C'mon, it's 2024. There's got to be some AI in here, right?

Let's shift left on promotion conflicts. Grace can't eliminate all of them, but it will reduce how often they happen, and it will proactively offer AI help to resolve them, so you don't have to worry about them ever again.

With Grace, there's no need for forking entire repositories just to make contributions. In open-source repos, you'll just create a personal branch against the repo.

You'll own your personal branch, and you can make it public or private. Your branch gets auto-rebased like any other branch in the repository, to keep your version up-to-date. When your change is ready you can submit PR's to get your personal branch's version promoted to a parent branch in the repo.

This is how I expect a large, open-source project in Grace to be: dozens of contributors, each with personal branches, working on a public project that remains securely controlled with ACL's. Everyone gets auto-rebased with every update to main. No networks of forks to manage, no entire copies of the repo. Just individuals working on the same repo, securely, together.

Grace's branching strategy is called single-step and is designed to help reduce conflicts, and to make it easier to work on and promote code to shipping branches (like main).

Single-step branching is both easy-to-use and powerful enough to be all that you need to run your projects.

There's a separate page that describes it in more detail.

At GitHub, I've been fortunate to work with customers who run some of the largest monorepos in the world. I've seen the challenges they face, and I've seen what they need to build and ship code to production.

Grace is ready for it.

Grace is designed to handle monorepos with ease. It's built to scale, and treats every repository, branch, directory version, and file version as a separate entity in a networked cluster of virtual actors. This design allows Grace to scale to handle arbitrarily large repositories, just as easily as it handles small- and medium-sized repositories.

So far, Grace has been tested on repositories as large as 100,000 files with 15,000 directories, with excellent performance, and no limits in sight.

Grace has no problem storing large files. Really large files. It's been tested with 10GB files - not that I think files that large belong in version control - and it should handle even larger files well.

Grace will let you specify how to handle those files, like only downloading them for the Design department, but not for Engineering. It's up to you.

Yeah, I said it.

Grace will have a native GUI app for Windows, Mac, Android, and iOS. (And probably Linux.)

Take Grace with you wherever you go. Merge conflict UI, Version History view, repository browsing, current status and more... all running at full native speed on your devices.

Shipping a native app doesn't mean that we don't also need a great web UI.

Sometimes the best way to share information is using a URL.

CLI + Web UI + GUI... use Grace the way you want to.

Grace goes way beyond just repository-level permissions. Grace uses OpenID and OAuth2 to integrate with your AuthN and AuthZ providers. Want to lock down specific paths in your repository to specific users and groups? With Grace, you'll be able to.

Because grace watch automatically uploads new file versions after every save, and commands like grace switch ensure that everything in your working directory is preserved before switching to another branch or reference, there's no need to stash anything. Just switch and go.

Plan your branching strategy in greater detail than ever before. Grace lets you decide which kinds of references are enabled in each branch.

For instance, main might enable promotions and tags, and disable commits, checkpoints, and saves.

Ordinary user branches might disable promotions, but enable everything else.

It's up to you.

Checkpoints and saves are features to help you be more productive, but we don't need to keep them forever. Grace allows you to control how long they're kept are kept in each repository. 72 hours? One week? A month? Or maybe just "keep the last 1,000 saves". It's up to you.

Sometimes you need to delete a version, whether it's because a secret or password was accidentally checked in, or because of some other security issue. With the right permissions, you'll be able to remove bad versions of your code - with a permanent reference that says that you did.

Every time you run grace promote, grace commit, or grace checkpoint, and every time grace watch uploads a new version of a file, Grace Server can optionally pre-render diffs for you to see, either in the CLI or in the Version History view. This makes seeing your ongoing changes in the Version History view instantaneous.

They're automatically aged out and deleted after a configurable length of time, so they don't just sit there wasting resources forever.

If your repository includes large binary files, like edited video, graphics, or game artifacts, you need to make sure that only one person at a time edits each file. Grace will support optional file locking at the directory level to ensure that you don't have to re-do work.

(Don't worry... if you don't need file locking, you don't have to turn it on.)

Grace Server is a modern Web API application. Like Grace's CLI, the Web API is easy to use, and easy to understand.

Grace ships with a .NET SDK, which is simply a projection of the Web API into .NET (and which Grace itself uses). SDK's for other platforms are welcomed as community contributions.

Grace uses SHA-256 hashes to verify that the files you uploaded, and the directory versions that went with them, are exactly the ones that get retrieved by clients. Grace will include a command to verify the SHA-256 hashes of all downloaded files and directory versions.

Grace repositories have a local cache of file versions that have been uploaded and downloaded. When running grace watch, Grace can download new file versions from multiple branches in the background so your grace switch commands run nearly instantly.

The local file cache is pruned regularly.

When you're running grace promote/commit/checkpoint/save/tag -m <some message>, the <some message> part can be up to 2048 characters.

50 characters... I don't think so. Feel free to share details. The person you help might be future you.

Yes, we know... it's hard to let go. Grace will perform an initial import from a Git repo, and will export to a Git repo.⁴

Grace Server is a modern, cloud-native Web API application. It will ship in a container on Docker Hub. (Of course.)

Grace Server is designed to be easy to deploy and operate. It runs on your choice of dozens of cloud-native databases, components and services, using Dapr, making it flexible and inherently scalable. Grace Server is stateless and scales up and down well using basic KEDA counters.

Grace is written primarily in F#, and is organized into nine .NET projects.

• Grace.CLI is the command-line interface for Grace.
• Grace.Server is an ASP.NET Core project that defines the Web API for Grace.
• Grace.SDK is a .NET class library that is a platform-specific projection of the Web API; it is used by Grace.CLI and wraps the HTTPS calls to Grace Server.
• Grace.Actors holds the code for the Actors in the system; it is used exclusively by Grace.Server.
• Grace.Shared is where all common code goes; it is used by all of the other Grace.* projects.
• (future) Grace.Avalonia will contain a native GUI for Grace for Windows, Android, MacOS, and iOS (and hopefully Linux).
• (future) Grace.Blazor will be an ASP.NET Core project containing a web UI for Grace.
• CosmosJsonSerializer is a custom JSON serializer class, used when deploying Grace with Azure Cosmos DB.

An additional project, Grace.Load, is an experiment to create a load test for Grace.Server.

Solid lines indicate a .NET project reference. Dotted lines indicate network requests.

flowchart LR
    subgraph Server
    Grace.Server-->Grace.Actors
    Grace.Server-->CosmosJsonSerializer
    end

    subgraph Clients
    Grace.CLI-->Grace.SDK
    Grace.SDK-.->|HTTPS|Grace.Server
    Avalonia["(future) Grace.Avalonia"]-->|"(not yet started)"|Grace.SDK
    Blazor["(future) Grace.Blazor"]-->|"(not yet started)"|Grace.SDK
    end

    subgraph Utilities
    Grace.Load-->Grace.SDK
    end

    subgraph Tests
    Grace.Server.Tests-->Grace.Server
    end

    Grace.Actors-->Grace.Shared
    Grace.Server-->Grace.Shared
    Grace.CLI-->Grace.Shared
    Grace.SDK-->Grace.Shared
    Avalonia["(future) Grace.Avalonia"]-->|"(not yet started)"|Grace.Shared
    Blazor["(future) Grace.Blazor"]-->|"(not yet started)"|Grace.Shared

Our Code of Conduct is available here. The tl;dr is:

The name of this project is Grace.

Be graceful in your interactions.

Give grace to everyone participating with us.

Create something together that embodies grace in its design and form.

When in doubt, remember the name of the project.

Grace Server will be shipped as a container, which will be made available on Docker Hub. Dapr's sidecar and actor placement processes are shipped as containers and are available on Docker Hub.

We intend to provide a Docker Compose template, as well as Kubernetes configuration for deployment, allowing for deployment to any major public cloud provider, as well as on-premises hardware.