Published

Introducing the First Stable FAB Format for Flower Hub

Share blogpost

We’re excited to introduce the first stable version of the Flower App Bundle (FAB) format for Flower Hub.

This is a foundational step for the Flower ecosystem.

Flower Hub is not just a place to upload apps, it’s the center of a growing ecosystem of collaborative AI applications, and for that ecosystem to thrive, apps must remain compatible, reusable, and future-proof.

The stable FAB format makes that possible.

A Foundation for a Growing Ecosystem

As Flower evolves, so do the requirements for apps:

  • Richer metadata
  • Improved packaging standards
  • Stricter validation rules

But evolution creates a challenge: How do you introduce new capabilities without breaking existing apps?

An ecosystem only works if:

  • apps published today still work tomorrow
  • new features don’t force immediate migration
  • old and new versions can coexist seamlessly

The stable FAB format creates a clear compatibility boundary, giving developers confidence today while allowing Flower to innovate tomorrow.

Why this matters

As a user or publisher of collaborative AI apps, you want:

  • Apps that continue to work across platform updates
  • A consistent and predictable packaging standard
  • The ability to adopt new features on your own timeline

Without a defined app format, even small changes can introduce risk:

  • New validation rules might unintentionally break older apps
  • Newer apps have no clean way to signal advanced requirements

The stable FAB format solves this.

It provides a structured way to preserve backward compatibility while enabling forward-looking innovation.

Powering Flower Hub

Flower Hub depends on a shared understanding of how apps are packaged and validated.

The FAB format is that shared contract.

It ensures that apps from different publishers can:

  • Coexist in a single ecosystem
  • Follow consistent validation rules
  • Be reliably discovered, shared, and executed

This is not just a packaging detail; it’s core infrastructure for the Flower ecosystem.

With a stable FAB format, Flower Hub can evolve safely, introducing new requirements without disrupting what already works.

How It Works

At the center of the FAB format is a simple but powerful field: fab-format-version.

Defined under [tool.flwr.app] in pyproject.toml, this field tells Flower which app format and validation rules to apply during the build process.

For example:

[project]
name = "my-federated-app"
version = "0.1.0"
dependencies = [
  "flwr>=1.27.0,<=2.0.0",
]

[tool.flwr.app]
publisher = "your-username"
fab-format-version = 1
flwr-version-target = "1.28.0"

Key points:

  • fab-format-version is not the app version
  • it is not the Flower runtime version
  • it defines the app format and validation rules
  • it governs both metadata and FAB content validation

If omitted, the app defaults to legacy format (0).

Built for Evolution

The stable FAB format is designed with long-term evolution in mind:

  • Older apps continue to work under legacy rules
  • Newer apps can adopt stricter validation when ready
  • Platform updates no longer risk breaking the ecosystem

This means developers can move forward at their own pace, without sacrificing compatibility.

For Flower Hub Publishers

If you publish apps on Flower Hub, now is the perfect time to upgrade.

By adopting the latest fab-format-version, you:

  • Make your app’s format explicit
  • Align with current validation standards
  • Prepare your app for future platform capabilities

Review your app metadata, validate your FAB, and republish using the latest format.

Follow the guide: How to Publish an App on Flower Hub

Learn More

To explore the full details:

In Summary

The first stable FAB format establishes a reliable, forward-compatible foundation for Flower Hub.

It ensures that:

  • Today’s apps remain usable tomorrow
  • New capabilities can be introduced safely
  • The ecosystem can grow without fragmentation

FAB is more than a file format, it’s the backbone of a scalable collaborative AI ecosystem.

Share Blogpost