feat(datacolumns/datatemplates/propertydata): composite DAC support (Phase 1)

[prasad-albert]

Summary

• DataColumns: add DataColumnType enum, SubDataColumnRef and CompositeDataColumn models; extend DataColumn with type, isSystemDefined, subDataColumns, parentId, validation; new create_composite() method and raw param on get_all()
• DataTemplates: surface type, isSystemDefined, subDataColumns on DataColumnValue from GET responses; new reorder_columns() method for PATCH sequence reordering
• PropertyData: add CompositePropertyValue model; widen PropertyData.value, PropertyValue.value, TaskPropertyCreate.value to accept composite arrays; add type field to TaskDataColumn and PropertyValue
• Parameter groups: add DataType.COLOR; widen ValueValidation.min/max/value to accept float (API returns numbers in DAC validation context)

Closes EX-2118 (Phase 1 — DataColumn / DataTemplate / PropertyData)

[claude]

Code Review

Issues Found

• [src/albert/collections/data_columns.py:178] create_composite is missing the @validate_call decorator. Every other public method in this collection (get_by_name, get_by_id, get_all, delete) has it, and AGENTS.md requires it on all public collection methods.

• [src/albert/resources/data_columns.py:20] SubDataColumnRef has a one-liner docstring but no Numpy-style Parameters section for its fields (id, name, key, required, validation, parent_id). The sibling class CompositeDataColumn (line 43) has a proper Parameters section — SubDataColumnRef should match that standard.

• [src/albert/resources/property_data.py:40] CompositePropertyValue has a one-liner docstring but no Numpy-style Parameters section for its fields (key, id, value). AGENTS.md requires Numpy-style docstrings on all new public classes.

Summary

3 issue(s) found. The missing @validate_call is a runtime correctness gap; the two docstring gaps violate the required Numpy-style standard.

[claude]

Code Review

Issues Found

• src/albert/resources/data_columns.py:66-70 New DataColumn fields (type, is_system_defined, sub_data_columns, parent_id, validation) have no attribute docstrings, breaking the standard this file otherwise follows (every other field is documented). BaseAlbertModel sets use_attribute_docstrings=True, so these fields emit empty JSON Schema descriptions (consumed by the ReAct worker tool factory / expand_schema). Same gap for the new SubDataColumnRef fields and the new type/is_system_defined/sub_data_columns fields on DataColumnValue in data_templates.py.
• src/albert/collections/data_columns.py:173-175 The raw param docstring says "Defaults to True (flat list)" but the actual parameter default is None; that default-value claim describes backend behavior, not the signature. Clarify to avoid confusion.

Summary

2 issue(s) found. No correctness bugs; serialization (by_alias/mode=json), pagination, type hints, and union parsing are sound. Both items are documentation/standards gaps, with the missing attribute docstrings being the more meaningful one since they affect generated schema descriptions.

[claude]

Code Review

Issues Found

• [src/albert/collections/data_templates.py:98-124] New public method reorder_columns is missing from the DataTemplateCollection class docstring Methods section. AGENTS.md requires every collection class docstring to list all public methods. (create_composite was correctly added to DataColumnCollection's Methods list, so this is an inconsistency.)
• [src/albert/collections/data_columns.py:254] create_composite deserializes the response as a single object (CompositeDataColumn(**response.json())), while the sibling create reads response.json()[0] (a list). If the POST /composite endpoint returns a list like the plain create endpoint, this will raise. Please verify the response shape.
• [tests] No integration tests were added for create_composite or reorder_columns.

Summary

2 issues found plus a test gap. The docstring omission is a standards violation; the response-shape assumption in create_composite is a potential runtime error worth confirming against the API. Serialization, pagination mode, type hints, and PR title all look correct.

[lkubie]

Bump: composite DAC Phase 1 has been waiting on review for a while. Re-requested @sbernasek. @prasad-albert if the approach has changed since, say the word and we can re-scope.

[prasad-albert]

Bump: composite DAC Phase 1 has been waiting on review for a while. Re-requested @sbernasek. @prasad-albert if the approach has changed since, say the word and we can re-scope.

I think we need to reverify if this PR is complete with the feature, there might be a few things we're missing here