Function get_workspace

py-src/data_formulator/workspace_factory.py:129–163

Return the active :class:`Workspace` for *identity_id*. For local/Azure: uses WorkspaceManager with lazy creation. For ephemeral: creates a scratch workspace and materializes table data from ``_workspace_tables`` in the request body. The frontend (IndexedDB) owns all data and

(identity_id: str)

def get_workspace(identity_id: str) -> Workspace:
    """
    Return the active :class:`Workspace` for *identity_id*.

    For local/Azure: uses WorkspaceManager with lazy creation.
    For ephemeral: creates a scratch workspace and materializes table data
    from ``_workspace_tables`` in the request body.  The frontend (IndexedDB)
    owns all data and sends it with every request; the backend writes it to
    temp parquet files so agents/DuckDB can read normally.
    """
    ws_id = get_active_workspace_id()
    if not ws_id:
        raise ValueError("No active workspace. X-Workspace-Id header is required.")

    backend = _get_backend()

    if backend == "ephemeral":
        from data_formulator.datalake.ephemeral_workspace import construct_scratch_workspace

        # Extract workspace tables from the request body
        workspace_tables = []
        from flask import request
        if request.is_json:
            data = request.get_json(silent=True) or {}
            workspace_tables = data.get("_workspace_tables") or []

        return construct_scratch_workspace(identity_id, ws_id, workspace_tables)

    mgr = get_workspace_manager(identity_id)

    # Lazy creation: frontend generates the ID, backend creates on first use
    if not mgr.workspace_exists(ws_id):
        mgr.create_workspace(ws_id)

    return mgr.open_workspace(ws_id, identity_id)