Weaverstack

A data-engineering runtime for Microsoft Fabric, built around a central control plane.

One mandatory Weaver Lakehouse holds your installed source repositories and the authoritative catalogue. Destination Lakehouses and Warehouses hold only materialised output — no copied runtime, no per-target catalogue, no attachment requirements.

Folder, Delta and SQL Warehouse are materialisation targets. You describe objects in one repository; Weaver routes them to the physical targets you name, builds one global dependency graph across all three forms, and certifies each object in the central catalogue only once it has built.

Status: pre-alpha. The current CLI, Workspace and catalogue work follows the Weaver master plan.

Installation

pip install weaverstack        # core, for a Fabric Environment or notebook
pip install 'weaverstack[cli]' # plus the optional desktop CLI

Requires Python 3.11 or later. Tested on macOS, Linux and Windows across Python 3.11 and 3.12.

Local Spark development additionally needs a JDK, and on Windows runs under WSL — Spark's local writes need a winutils.exe that Windows does not carry. Everything else, including the CLI and the whole Fabric path, runs natively on all three.

Local development

Weaver runs against a local filesystem standing in for Lakehouses, so build and load can be developed without touching a workspace. It needs a JDK and a matched Spark/Delta pair — all optional, none of it required to use Weaver on Fabric.

weaver doctor

reports what is present and what to install. See design/local-setup.md.

CLI lifecycle

One Workspace configuration can abbreviate the full desktop lifecycle:

weaver install     --workspace-config workspace.yml
weaver build ./estate --workspace-config workspace.yml --bind Lakehouse/Sales_Dev

The same commands work against a local folder with --workspace-type local, except Warehouse work, which remains Fabric-only. See CLI usage for repository sources, physical-first bindings, wipe and unbind.

Documentation

• Authoritative master plan
• How Weaver build works — incremental selection, bundle order and certification
• Where your Weaver document repository lives — a folder of files, and how it reaches Fabric
• CLI usage — signing in, workspaces, capacity, wipe
• Local development setup
• Fabric integration tests
• Agent guide

Licence

Apache 2.0. See LICENSE.