Skip to content

microsoft/fabric-task-flows

Task Flows for Microsoft Fabric

CI Python 3.11+ License: MIT

From problem to production in minutes. Less guessing. More building.


It starts the same way every time — a messy whiteboard and a simple question: "What can we build?"

But what if you didn't need to know Microsoft Fabric? What if you only needed to know what your business needs to solve?

With the @fabric-advisor agent, you describe the problem. After a few prompts, the whole solution is mapped, everything deployed, and everyone can see how the pieces connect — so you can start building, and stop guessing.

Who is this for?

  • Teams with a problem to solve — who don't have time to become platform experts first
  • Architects who want pre-validated patterns instead of starting from scratch
  • Anyone new to Fabric who needs opinionated guidance without the guesswork

How it works

The agent walks you through six phases that chain together automatically:

Phase What happens
Discover You describe your problem — the agent asks questions and recommends an architecture
Design A detailed architecture, with decision records explaining every trade-off
Test A test plan validates against your acceptance criteria
Deploy The agent generates dependency-ordered scripts, CI/CD-ready
Document Everything synthesizes into a human-readable brief

What you get

Each project produces two deliverables — one doc, one deployment:

_projects/your-project/
├── docs/project-brief.md     ← ONE human-readable doc (problem, architecture, decisions, validation)
└── deploy/                   ← CI/CD-ready deployment scripts

No sprawl — one project brief and the scripts to make it real.

Quick start

Prerequisites: Python 3.11+, GitHub Copilot with agent mode

python _shared/scripts/run-pipeline.py start "My Project" --problem "describe your business problem"

The pipeline runner generates agent prompts — paste each into Copilot chat. Use advance and next to progress through phases. See _shared/workflow-guide.md for the full pipeline reference.

What's inside

Resource Description
13 Task Flows Pre-defined architectures for common scenarios — batch, streaming, hybrid, ML, API, governance, and more
7 Decision Guides Opinionated guidance on the choices that matter
Deployment Diagrams Visual maps showing how pieces connect and deploy in order
Item Registry 45 Fabric item types with API paths, CI/CD strategies, and deployment order
Templates Definition files for every deployable Fabric item type

Repository structure

.github/agents/         → The @fabric-advisor orchestrator
.github/skills/         → Composable skills for each phase
decisions/              → Decision guides (the "why" behind each choice)
diagrams/               → Deployment diagrams (how pieces connect)
_shared/registry/       → Canonical registries the agent uses
_shared/templates/      → Starting point templates
_shared/lib/            → Shared Python modules
_shared/scripts/        → Pipeline CLI tools
_shared/tests/          → Test suite
_projects/              → Your project workspaces (gitignored)

Contributing

We welcome contributions — especially to the Item Type Registry. See CONTRIBUTING.md for guidelines on adding task flows, decision guides, skills, and more.

License

MIT

About

Task flows let you describe your problem and have an AI agent generate your Microsoft Fabric solution—workspaces, pipelines, notebooks, data agents, models, reports, and more—delivered as a CI/CD‑ready script, ready to deploy.

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages