fablake

fablake is a production-oriented Python library for Microsoft Fabric OneLake Lakehouse access.

It provides:

• a filesystem-first Lakehouse API
• a pathlib-like LakehousePath for Files paths
• table root locators and discovery helpers for Tables interop
• filesystem interop for pandas, PyArrow, Polars, and DuckDB
• runtime support for both Microsoft Fabric notebooks and local development

Install

Install with uv:

uv add fablake

For local authentication via Azure credentials, also install:

uv add azure-identity

pip install fablake azure-identity also works.

You can also install the auth extra with pip install "fablake[auth]".

Quick start

from fablake import Lakehouse

lh = Lakehouse(workspace="Finance", lakehouse="Ops", schema_enabled=True)

path = lh.files / "raw" / "orders" / "2026-01-01.json"

if path.exists():
    print(path.read_text())

table_root = lh.table("orders", schema="brz")
print(table_root)

for table in lh.tables.list("db*.*der"):
    print(table)

Core concepts

• lh.files targets the Lakehouse Files root.
• lh.table(name, schema=...) returns a Tables locator, not a dataframe.
• lh.tables.list(pattern) returns matching LakehouseTable locators.
• lh.fs exposes the underlying filesystem handle for library interop.

Interop and runtime

• Use lh.fs as the filesystem layer for pandas, PyArrow, Polars, and DuckDB workflows.
• Inside Microsoft Fabric notebooks, omitted credentials use notebook/runtime auth resolution.
• In local development, install azure-identity and use your local Azure credential flow.

Filesystem interop example:

from fablake import Lakehouse
import duckdb
import pandas as pd
import polars as pl
import pyarrow.parquet as pq

lh = Lakehouse(workspace="Finance", lakehouse="Ops", schema_enabled=True)
path = lh.files / "raw" / "orders.parquet"

# pandas + PyArrow engine
df_pd = pd.read_parquet(path, engine="pyarrow", filesystem=lh.fs)

# PyArrow
table = pq.read_table(path, filesystem=lh.fs)

# Polars (via file-like stream)
with lh.fs.open(path, "rb") as stream:
    df_pl = pl.read_parquet(stream)

# DuckDB
duckdb.register_filesystem(lh.fs)
df_duck = duckdb.sql(f"SELECT * FROM read_parquet('{path}')").df()

Documentation

• User guide: docs-site/docs/index.md
• Python API docs: docs-site/docs/api/index.md
• Usage examples: docs-site/docs/usage-examples.md
• Changelog: CHANGELOG.md
• Version selector source: docs-site/docs/versions.json

Build docs locally:

uv sync --group docs
uv run python scripts/docs.py serve

Other helper commands:

uv run python scripts/docs.py build
uv run python scripts/docs.py api
uv run python scripts/docs.py sync