Motivation

• Organize node-type specific constants and helpers out of the large structure.py to improve modularity and readability.
• Centralize schema building and sanitization logic per node type so maintenance and future extensions are easier.

Description

• Added a new velvetflow/planner/node_types/ package with modules for each node type: action.py, condition.py, data.py, loop.py, reasoning.py, and switch.py and a central __init__.py that re-exports utilities.
• Moved data-node schema construction into node_types/data.py as build_data_node_output_schema and loop-related helpers into node_types/loop.py (extract_loop_body_context, validate_loop_exports, fallback_loop_exports, ensure_loop_items_fields).
• Implemented shared param/field sanitization in node_types/sanitizers.py (previously _filter_supported_params, _sanitize_builder_node_params, _sanitize_builder_node_fields) and updated structure.py to import and use filter_supported_params, sanitize_builder_node_params, sanitize_builder_node_fields, and build_data_node_output_schema from the new package.
• Removed the old in-file node-type constants and duplicated helper implementations from structure.py, updated call sites to the new function names, and kept existing behavior/validation logic intact.

Testing

• Ran bytecode compilation for changed modules with python -m compileall velvetflow/planner/structure.py velvetflow/planner/node_types, which completed successfully.
• Performed a lightweight runtime import check with python - <<'PY'\nfrom velvetflow.planner.structure import plan_workflow_structure_with_llm\nprint('ok')\nPY, which printed ok indicating the planner entrypoint imports and basic wiring are functional.

Codex Task