Video summary

Rayfin and Fabric Apps Overview

Main summary

Key takeaways

Technology

Summary of Technological Concepts / Product Features

Problem Scenario (Why Rayfin + Fabric Apps)

You may already have data in Microsoft Fabric OneLake, including:

  • Structured data with a star schema
  • A semantic model

Or you may be ingesting/mirroring:

  • Structured, semi-structured, or Parquet data
  • Via data virtualization

Teams want to build applications that:

  • Surface this data to users/clients
  • Provide durable state (so you don’t have to be a database expert)
  • Use governance, compliance, security, and authentication aligned with enterprise requirements
  • Support AI/agentic-era prototyping without breaking compliance
  • Are easy to host and scale

What Rayfin Is

Rayfin is positioned as an SDK (tooling layer) for building:

  • Data models
  • Applications

…in a way that integrates with Microsoft Fabric.

It targets the agentic era: you define intent, and AI can help generate the code/model artifacts.

Rayfin + Fabric Apps Platform Concept

Rayfin is described as enabling Fabric apps with a managed backend-as-a-service approach.

Key UX/portal idea:

  • After scaffolding and customization, you mostly just publish.

The emphasis in the video:

  • Rayfin deploys on top of Fabric, so Fabric handles underlying infrastructure concerns.

Key Implementation Details (Data / Durable Storage)

Data Modeling Using TypeScript

Rayfin lets you define entities in TypeScript, such as:

  • Villain
  • Hero
  • Event (example domain)

Entities use decorators for:

  • Fields
  • Roles / permissions
  • Relationships

AI can generate these TypeScript definitions from natural language intent.

Deployment “Materialization” into Fabric Storage

When you deploy a Rayfin-scaffolded project:

  • The TypeScript models are decompiled/materialized into Fabric primitives
  • The demo notes that, today, this becomes SQL tables / SQL schema
    • (with the expectation that storage mechanisms could evolve)
  • Rayfin exposes the data via a GraphQL endpoint for application interaction

Why This Helps

The resulting data becomes part of OneLake, making it available for:

  • Organization-wide analytics
  • AI integrations
  • Secure access governed by Fabric permissions

Key Implementation Details (Authentication / Governance)

Authentication Handled by Fabric

Authentication for the app is handled by Fabric, using Single Sign-On (SSO).

Specifically mentioned:

  • Microsoft Entra ID (with the note that this could expand later).

Governance and Compliance

Because the backend/hosting are inside Fabric, the app inherits:

  • Fabric governance/compliance/security/authorization

The video frames this as reducing the risk that “AI vibe code” will produce:

  • insecure infrastructure
  • misconfigured deployments

Key Implementation Details (Hosting / Frontend)

Hosted Applications in Fabric

Rayfin provides hosting for pre-rendered frontend content, such as:

  • HTML
  • JavaScript
  • CSS

The browser loads this hosted content, while Rayfin/Fabric wiring connects it to:

  • the backend
  • the data layer

Flexible Usage Modes

Rayfin can be used to support multiple patterns:

  • Create new durable data models (Rayfin-defined) and host the full app in Fabric
  • Build a frontend that connects to existing OneLake data
    • Example mentioned: querying via DAX
  • “Either/or” combinations:
    • durable storage +/or frontend +/or existing data connectivity

Scaling

The Fabric app is described as able to auto-scale as needed.

Tutorial / Demo Flow Highlighted

Project Scaffolding (Two-Step Idea)

Using the SDK:

  1. Scaffold the project
  2. Customize (optionally with AI)
  3. Publish to Fabric

CLI / Template Setup

A terminal command shown:

npm create Microsoft Raven latest

Options mentioned include:

  • Blank app
  • Data app connected to existing OneLake/semantic model data
  • Starter templates (e.g., a to-do app)

Hands-on Example Scenario (US Map Dashboard)

The speaker creates a Fabric app and selects a data app option.

They use:

  • An existing semantic model built over three tables:
    • attacks / villains / heroes

AI-assisted customization example:

  • “create a map of the US based on the semantic model”

They add interactive features such as:

  • a time slider

Upload/publish process:

  • Test locally
  • Publish using an npx Rayfin command (as described)

Result:

  • A customized dashboard visualizing villain activity across the US
  • Interactive hover and time filtering

Positioning vs. Power BI

Power BI remains important for:

  • GUI-driven semantic model creation
  • zero-code dashboards

Rayfin is positioned as complementary by enabling:

  • vibe coding for highly customized experiences
  • publishing experiences via a URL
  • deeper app integration with OneLake data
  • easier creation of new compliant apps with durable storage

Key Takeaways / Analysis Points

Rayfin aims to reduce friction for:

  • developers and non-developers

It does this by:

  • Letting teams define data models in TypeScript
  • Using AI to generate those definitions
  • Automatically deploying them into Fabric-managed storage and exposing via GraphQL
  • Ensuring Fabric-native security + SSO (Entra ID mentioned)
  • Handling hosting/scaling inside Fabric for both backend and frontend

Main Speakers / Sources

  • Speaker/source: Single presenter narrating the demo (name not provided in subtitles).
  • Product sources referenced: Rayfin SDK and Microsoft Fabric (including OneLake, Fabric Apps, Entra ID/SSO, GraphQL, Fabric primitives, and semantic models).

Original video