Summary

• Add optional model and provider fields to todo items
• Allow per-task model override when delegating to subagents
• Resolve credentials per-task instead of globally
• Add tests for new model/provider fields

Changes

tools/todo_tool.py

• Extended TodoStore to support optional model and provider fields
• Updated schema to include new fields with descriptions
• Validation handles the new optional fields

tools/delegate_tool.py

• Added _resolve_task_model() function to resolve per-task model credentials
• Modified _resolve_delegation_credentials() to accept task-level overrides
• Each task now resolves its own credentials independently
• Updated DELEGATE_TASK_SCHEMA with model/provider properties

tests/tools/test_todo_tool.py

• Added tests for model and provider field handling
• Tests cover write, merge, and validation scenarios

Use Case

When delegating tasks to subagents, you can now specify a different model/provider per task:

{
  "id": "1",
  "content": "Use GPT-4 for this task",
  "status": "pending",
  "model": "gpt-4o",
  "provider": "openai"
}

This enables mixed-model workflows where different tasks use different models based on their requirements.