pyfabric

Python library for creating, validating, and locally testing Microsoft Fabric items that are compatible with Fabric git sync.

Installation

pip install pyfabric

Optional dependencies

pip install pyfabric[azure]         # Azure authentication and REST client
pip install pyfabric[data]          # OneLake and SQL data access
pip install pyfabric[lakehouse-io]  # Delta table writes (deltalake + pyarrow)
pip install pyfabric[testing]       # DuckDB Spark mock and pytest fixtures
pip install pyfabric[dataagent]     # Data agent MCP query client
pip install pyfabric[deploy]        # parameter.yml substitution (PyYAML)
pip install pyfabric[all]           # All optional dependencies

AI assistant context (optional)

pyfabric ships with reference memories so any AI coding assistant can learn the library without you re-explaining it every session. Two ways to install, depending on your assistant:

Claude Code — one command, installs into your active Claude profile:

pyfabric install-claude-memory            # $CLAUDE_CONFIG_DIR/memory or ~/.claude/memory
pyfabric install-claude-memory --dry-run  # show what would change
pyfabric install-claude-memory --force    # overwrite locally-edited copies

Safe to re-run after upgrades — MEMORY.md is merged (never overwritten), and individual .md files are skipped unless --force is passed.

Any other assistant — emit the memories as portable markdown and redirect to whatever file your tool reads:

pyfabric emit-context > .github/copilot-instructions.md   # GitHub Copilot
pyfabric emit-context > .cursorrules                      # Cursor
pyfabric emit-context > .continuerules                    # Continue
pyfabric emit-context > CONVENTIONS.md                    # Aider

emit-context strips Claude-specific frontmatter and skips the index file, so the output is plain markdown any LLM can consume. Commit the generated file so teammates' assistants pick it up too, and re-run after upgrading pyfabric.

Quick start

First-run smoke test

After pip install --pre "pyfabric[azure]", verify install + auth + the core read APIs in one command:

pyfabric demo "<your_workspace_display_name>"
pyfabric demo "<your_workspace_display_name>" --show-definitions

It lists accessible workspaces, resolves the one you named (with case-insensitive and "did you mean" fallbacks), and summarises its items. Read-only — no writes, no pushes. Equivalent script form: examples/workspace_demo.py.

Validate a Fabric workspace

from pathlib import Path
from pyfabric.items.validate import validate_workspace

results = validate_workspace(Path("my_workspace/"))
for r in results:
    status = "OK" if r.valid else "FAIL"
    print(f"{status}: {r.item_path.name}")
    for e in r.errors:
        print(f"  ERROR: {e.message}")

List workspaces with the REST client

from pyfabric.client.auth import FabricCredential
from pyfabric.client.http import FabricClient
from pyfabric.workspace.workspaces import list_workspaces

cred = FabricCredential(tenant="contoso")
client = FabricClient(cred)

for ws in list_workspaces(client):
    print(f"{ws['displayName']}  {ws['id']}")

Test notebook logic locally

# In your test file (pytest)
def test_my_notebook(fabric_spark, mock_notebookutils):
    # fabric_spark is a DuckDB-backed SparkSession replacement
    df = fabric_spark.sql("SELECT 1 AS value, 'hello' AS msg")
    rows = df.collect()
    assert rows[0]["value"] == 1

    # mock_notebookutils replaces Fabric notebookutils
    mock_notebookutils.fs.mkdirs("/output")
    mock_notebookutils.fs.put("/output/result.txt", "done")

Documentation

Sub-packages

Supply Chain Security

Every release of pyfabric includes supply chain security attestations:

• SLSA Build Provenance — each release is built in GitHub Actions and attested with SLSA provenance, verifiable with gh attestation verify
• SBOM (SPDX) — a Software Bill of Materials in SPDX JSON format is generated for every release and attached as a release asset
• PyPI Trusted Publisher — packages are published to PyPI via OpenID Connect with no long-lived credentials
• Dependency Review — every PR is scanned for known vulnerabilities via pip-audit and GitHub's dependency review action

Verifying a release

# Verify SLSA provenance of a downloaded package
gh attestation verify pyfabric-*.whl --repo Creative-Planning/pyfabric

# Download the SBOM for a specific release
gh release download v0.1.0a2 --repo Creative-Planning/pyfabric --pattern "*.spdx.json"

The SBOM for each release is available at: https://github.com/Creative-Planning/pyfabric/releases/download/<tag>/pyfabric-build.spdx.json

Requirements

• Python 3.12 or later

Developer setup

After cloning, install dev dependencies and git hooks so that lint, format, type check, and tests run automatically before commit/push:

pip install -e ".[dev]"
pre-commit install
pre-commit install --hook-type pre-push

See CONTRIBUTING.md for full details.

Contributing

This project is maintained by Creative Planning. We do not accept external contributions at this time. You are welcome to fork and modify under the MIT license. See CONTRIBUTING.md.

License

MIT