fix(todo): accept the TodoItem shape todo_write declares

[eastagiletracker]

This PR proposes a fix for ToDoResource.todo_write(), which rejects the TodoItem type its own signature declares. We include this PR work along with a full history of your repo at https://eastagiletracker.com/projects/252. You can sign in with your GitHub ID to claim ownership of the project.

The defect

todo_write() is annotated todos: list[TodoItem], but its validation and rendering index every item as a mapping (if "content" not in todo, todo["status"], todo['activeForm']). Passing the very model the signature declares does not raise — it silently returns a false validation error, because BaseModel.__iter__ yields (name, value) tuples, so "content" not in todo is always true. Passing an attribute-style item, such as the TodoItem dataclass in dana.core.runtime.protocols, raises TypeError instead.

Reproduced on develop at acb72ec:

import asyncio
from dana.core.resource.todo_resource import ToDoResource, TodoItem as Model
from dana.core.runtime import TodoItem as Dataclass

r = ToDoResource(resource_id="todo")
fields = dict(content="Run tests", status="pending", activeForm="Running tests")

print(asyncio.run(r.todo_write(todos=[fields])))                    # renders the list
print(asyncio.run(r.todo_write(todos=[Model(**fields)])))           # declared type
print(asyncio.run(r.todo_write(todos=[Dataclass(content="Run tests", status="pending")])))

Todo List: ...                                        # dict: works
Error: Todo item 0 missing 'content' field            # TodoItem: false error
TypeError: argument of type 'TodoItem' is not iterable # dataclass: crash

To be precise about the blast radius: the tool-call path decodes JSON to dicts, so agents driving todo_write through an LLM are unaffected today. What breaks is calling the method directly on its published signature — ToDoResource is exported from dana.core.resource and attached by DanaCodingAgent and the explore agent — and any caller holding the runtime TodoItem dataclass, which crashes outright.

The change

_as_mapping() normalizes each item to a dict before validation, accepting mappings, pydantic models and attribute objects. This mirrors the handling already in thought_logger._display_todo_list, which comments "Handle both TodoItem objects and dicts" — the same two shapes, reconciled in one more place. Dict callers, including the decoded tool-call path, are byte-for-byte unaffected, and get_todos() now genuinely returns the list[dict] it is annotated to return whatever shape went in. An item that is genuinely missing a field, like the two-field runtime dataclass, now gets the intended missing 'activeForm' field message rather than a TypeError.

Verification, with tests/unit/test_todo_resource.py added — this resource had no direct tests before:

• With the change reverted, 8 of the 12 new tests fail and the 4 dict-path tests still pass, which is what pins the fix to the reported shapes rather than to the suite at large.
• uv run pytest tests/ -m "not live and not deep" goes from 1898 passed / 38 skipped on a clean develop to 1910 passed / 38 skipped here — the 12 new tests, no new failures and no change in what is skipped.
• uv run ruff check dana/ passes, as it does on the baseline.

Covered: each accepted shape, all three rendering identically for the same todo, mixed shapes in one call, stored state, missing fields, invalid status, the error index in a mixed list, and the empty list.

How this was managed

We tracked this work on a board imported from this repository's own issues and pull requests (28 stories), so the fix has a story you can read start to finish: todo_write rejects the TodoItem type its own signature declares, on the Dana Runtime board.

If you'd rather not receive contributions like this, reply no-more-prs on this pull request and we won't open any further ones on your repositories.

---

Lawrence W. Sinclair
CEO / East Agile
linkedin.com/in/lwsinclair/
eastagile.com