Write your code. Weaver handles the work around it.

Keep using Python, SQL and Power BI in Fabric. Weaver gives your team a consistent way to load data, test results, deploy changes and connect the work.

Start with what you know

Python and Spark SQL, together

Use Python and Spark SQL together to load files and create Lakehouse tables. Weaver runs the steps in the right order.

Ordinary T-SQL

Keep ordinary SELECTs, without a proprietary modelling language or macros.

Keep using Power BI files

Apply modelling standards, run unit tests and deploy Reports using standard Power BI project files (PBIP and TMDL).

Work in Fabric or your own editor

Write code in VS Code or another editor on Windows, macOS or Linux, or work entirely in Fabric Notebooks. Deploy native Power BI models and Reports without Power BI Desktop.

Built for developers and AI agents

Developers get readable command output; AI agents get structured --json results. Both can build, load, test and inspect the work, with clear results and errors.

First Python pipeline Native document reference

Build reliable data pipelines

Run connected work in the right order.

Lakehouse

FilesPython Folder
Delta TablePython
Delta TableSpark SQL

Warehouse

TableT-SQL

Power BI

Semantic ModelRefresh

Weaver runs selected Lakehouse, Warehouse and Power BI work in the right order, waiting when new data needs to become available.

Connect Lakehouse and Warehouse Add a managed model source

Keep work in the right order

Run Lakehouse work, Warehouse SQL and dependent Power BI model refreshes in one connected workflow.

Turn a SQL query into a working table

Give Weaver a SELECT and a key. It creates the table and loading steps, tracks changes and records the results.

Turn Python and SQL into Lakehouse tables

Use Python and Spark SQL to load files and update Delta Tables, without building a separate pipeline to connect them.

Find what depends on what

Weaver can read dependencies from Python imports and supported T-SQL references. With Spark SQL, you specify them.

Load only what has changed

Apply inserts, updates and explicit deletes using rules you choose, without rebuilding every target.

Pick up where each table left off

Each table or folder tracks its own progress through the source data, even when several share the same source.

Test data and calculations

Compare actual and expected results or find invalid rows using Python, SQL and DAX.

Data-quality safeguards

Keep rejected rows and stop unusually large proposed updates or deletes before they reach your tables.

Write a SELECT. Get the engineering around it.

You write

T-SQL query+ native metadata comment

Weaver supplies

Physical Table
Load procedure & staging
Declared-key reconciliation
Supported dependencies
Change statistics
Health evidence

Declare the key used to reconcile rows. Incremental source logic identifies changes and explicit deletes.

From a query to a managed Table

First Warehouse pipeline Load behaviour

Power BI, engineered for organisations

Keep your Power BI files. Add shared standards, model-specific rules and unit tests.

PowerBI/
├── policy.tmdl          ← organisation-wide policies
├── Sales/
│   ├── Sales.pbip       ← standard Power BI project file
│   ├── Sales.tmdl       ← Sales-specific policies
│   ├── Sales.SemanticModel/
│   ├── Sales.Report/
│   └── tests/           ← unit testing for Semantic Models
│       └── Sales/
│           └── Sales.RevenueMatches.dax
└── Finance/
    ├── Finance.tmdl
    └── Finance.Report/

Add shared rules and Sales-specific changes without rewriting your Power BI files. Test calculations against expected results and deploy Reports connected to the right model.

Organisation-wide modelling standards

Apply shared native TMDL policy across Semantic Models instead of repeating formats and conventions.

Built-in modelling rules

Automatically hide foreign keys and create measure selectors using standard TMDL annotations.

Write your own modelling rules in Python

Add your own Python rules when you need more than Weaver's built-in modelling options.

Reuse models without copying them

Start from a shared model and create versions for different audiences without maintaining complete copies.

Unit testing for Semantic Models

Unit test DAX measures against independent expected results, including SQL calculations over managed Lakehouse or Warehouse sources.

Connect Reports to the right models

Deploy the same Report to different workspaces. Weaver checks that it connects to the intended Power BI model.

Supported native model/Report deployment also works in an appropriately licensed Power BI service workspace, without Lakehouse/Warehouse work or a Weaver Catalogue.

Deploy a native model Power BI reference

Know what's happening across Fabric

One view across your Fabric work — from Lakehouse and Warehouse to Power BI models and tests.

TerminalCLI

$ weaver health \
  --workspace-config workspace-development.yml

Health reportExample output

Health
  Read catalogue                   2.6s
  Check installed items            0.1s
  Read installed objects           3.0s
✓ Health                            5.8s

Weaver Health  Green

Load    Green
  Last load activity   0h 3m ago
  8 succeeded

Tests   Green
  4 succeeded

Build   Green
  Installed estate consistent (29 objects)

Fabric NotebookPython

import weaver

health = weaver.health(
    workspace="Parcel Development",
    catalogue="Warehouse/Catalogue",
)
health.report()

Run the same check from Python inside Fabric. The result can also be used in your own code.

See all your work in one place

Weaver keeps its Catalogue of deployed items, dependencies and run results in an ordinary Fabric Warehouse. There's no separate monitoring service to run.

See what is up to date or needs attention

See what is up to date, waiting, failed or inconsistent across your managed Fabric work.

See run history and results

See what ran, how long it took, row counts, data changes and recorded results.

Recover from failures

Keep results from completed steps, record failures and optionally continue unrelated Load work.

Run only the parts you choose

Select work by name, preview what will run and include related steps when you need them.

Control how much runs at once

Set limits on how much runs together while Weaver keeps dependent work in the right order.

Run and inspect your project Operation boundaries

Built for scale

From 3 objects to 3,000.

3Easy to keep in your head.
30It’s hard to keep track.
300It’s no longer whether you use a framework, but which one.
3,000Scheduling and recovery decide everything.

Weaver has been tested and tuned for thousands of connected objects. It runs independent work together, tracks what data has changed and refreshes SQL access only when needed.

Tuned for each engine

From Delta reads and writes to transactions, T-SQL and Shortcut creation, each operation is tuned for its engine to eliminate waste.

Run independent steps together

Weaver can run separate Spark and Warehouse steps at the same time, within limits you choose.

Process only the changes

Each object tracks its own progress, so it can process inserts, updates and deletes rather than starting over every time.

SQL endpoints refreshed as needed

A refresh runs in the background, and only the consumers that need the new state wait for it.

Stress tested with 3,000 Fabric objects.See how it was tested Get started

Visibility, included

Weaver includes a Power BI dashboard showing the status of your work. One command adds it to your Fabric workspace.

Overview: installed items and Health. — Seven pages: Overview, Health, the Orchestration graph, Performance by time of day, a Gantt of each workflow, an ER diagram with a metadata browser, and a bus matrix for every semantic model.

Orchestration: one table's lineage, readiness and last outcome. — Seven pages: Overview, Health, the Orchestration graph, Performance by time of day, a Gantt of each workflow, an ER diagram with a metadata browser, and a bus matrix for every semantic model.

One command

weaver catalogue-dashboard deploys the semantic model and Report beside the catalogue, and brings them up to date on the next run.

Live over Direct Lake

The dashboard reads the Fabric Catalogue using Direct Lake, so its pages reflect the latest recorded results.

Free to tweak

It is an ordinary Power BI Report. Tweak it to fit your needs, or throw it away and build your own over the catalogue.

When your projects get more complex

Same code. Different workspaces. No rewrites.

Keep one set of Python, SQL and Power BI files. Choose where to deploy by changing the workspace configuration file.

DevelopmentSame project files

weaver build \
  --item Warehouse/Operations \
  --workspace-config workspace-development.yml

Parcel Development

ProductionSame project files

weaver build \
  --item Warehouse/Operations \
  --workspace-config workspace-production.yml

Parcel Production

Both commands deploy the same Warehouse item; the configuration chooses its workspace. See workspace configuration. You can also preview changes first.

Fork your data, not just your code.

Existing source estateSource-owned data
Borrowed development dataWarehouse Views · Lakehouse Shortcuts
Selected changed workMaterialised locally by Build

With Mirror, you can change the parts you're working on and use existing data for the rest, without reloading everything.

Develop using Mirror Scope and reset boundaries

Start with one item.

Begin with a Lakehouse, a Warehouse or a Power BI model, and connect more as you need them.

Get started →Read the documentation →