fabric-platform

Python SDK for the Fabric AI workflow platform.

Installation

pip install fabric-platform

Quick Start

from fabric_platform import FabricClient

# API key auth
fabric = FabricClient(api_key="fab_your_key")

# Or: native email/password auth
fabric = FabricClient(base_url="https://fabric.example.com")
auth = fabric.login("alice@example.com", "SecurePass1!")

# Or: consumer-driven social login (your app handles OAuth)
auth = fabric.social_login(
    provider="google",
    provider_user_id=google_user["sub"],
    email=google_user["email"],
    display_name=google_user["name"],
)

# Use the authenticated client
me = fabric.get_me()
orgs = fabric.get_my_organizations()

Features

• Auth: signup, login, social login, change password, logout-all
• Organizations & Teams: CRUD, memberships, invitations
• Workflows: submit, monitor, stream events
• Assets: upload, download, galleries
• Webhooks & Schedules: create, manage, delivery history
• Admin: user management (list, unlock, verify, delete)

Typed workflows

Fabric derives JSON Schemas from each workflow's Pydantic task types at server boot and stores them in the registry. Discover a workflow's contract without reading the source:

schemas = fabric.get_workflow_schemas("research/trends")
print(schemas["input_schema"])    # JSON Schema for workflow input
print(schemas["output_schema"])   # JSON Schema for workflow output
print(schemas["task_schemas"])    # Per-task breakdown

Opt in to server-side validation on submission:

from fabric_platform.errors import FabricValidationError

try:
    run = fabric.submit_run(
        "research/trends",
        input={"topic": 42},  # wrong type
        validate=True,
    )
except FabricValidationError as e:
    print(e.response.json()["error"]["details"])

See the Typed Workflows guide for the full pattern, CI integration, and migration tips.

Documentation

See the Fabric documentation for the full API reference.