Require full reasoning params when adding nodes

[zhongkaifu]

Motivation

• Ensure reasoning nodes always include the full set of parameters the executor and planner expect so LLM-driven nodes are well-specified and predictable.
• Surface clear validation errors when a reasoning node is created with missing fields to guide callers to supply complete prompts, context, expected output schema and toolset.

Description

• Updated the reasoning params schema in velvetflow/planner/params_tools.py to mark system_prompt, task_prompt, context, expected_output_format, and toolset as required in _reasoning_params_schema.
• Added runtime validation in add_reasoning_node in velvetflow/planner/structure.py to detect empty or missing reasoning fields and return a structured validation error listing missing_fields, required_fields, and required_out_params_schema (which must match expected_output_format).
• The validation error message was clarified to call out that out_params_schema must align with expected_output_format.

Testing

• No automated tests were run as part of this change.

---

Codex Task

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0d3572a488

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Allow empty context/toolset for reasoning nodes

The new validation treats any empty list/dict/blank string as “missing” for all required reasoning fields, so toolset: [] or context: {} now causes add_reasoning_node to reject nodes that intentionally have no tools or no context. This is a regression because the executor already tolerates toolset being empty (it resolves to no tools) and can build prompts without context; workflows that previously created tool-free or context-free reasoning nodes will now fail validation even though they still execute correctly.

Useful? React with 👍 / 👎.