Python HITL docs/samples: use AgentResponse.value instead of model_validate_json for structured outputs

[pamelafox]

Summary

The HITL documentation page and the guessing_game_with_human_input.py sample use model_validate_json() to parse structured agent output inside workflow executors:

text = result.agent_response.text
last_guess = GuessOutput.model_validate_json(text).guess

Since the agent is configured with response_format=GuessOutput, the AgentResponse object already has a .value property that returns the parsed Pydantic model directly:

last_guess = result.agent_response.value.guess

Using .value is more idiomatic and avoids redundant parsing. It also better communicates that structured outputs are a first-class feature — the framework already handles the JSON parsing when response_format is set.

Affected files

• Docs: https://learn.microsoft.com/en-us/agent-framework/workflows/human-in-the-loop?pivots=programming-language-python
• Sample: python/samples/03-workflows/human-in-the-loop/guessing_game_with_human_input.py (link)

Context

It's possible model_validate_json was used because .value wasn't available on AgentResponse when the sample was originally written, or because .value support inside AgentExecutorResponse.agent_response was added later. Either way, .value is available today and would be the cleaner approach.

This also applies to other workflow samples that use model_validate_json with structured response_format agents (e.g. agents_with_approval_requests.py, checkpoint samples, etc.).

[LEDazzio01]

I've submitted PR #4405 to address this. The change replaces the manual model_validate_json parsing in guessing_game_with_human_input.py with the cleaner .value access:

-        text = result.agent_response.text
-        last_guess = GuessOutput.model_validate_json(text).guess
+        last_guess = result.agent_response.value.guess

I've also updated the docstring to explain the .value usage.

Note: other sample files (edge_condition.py, multi_selection_edge_group.py) also use model_validate_json with structured outputs, but their usage is in edge conditions and ChatClient.get_response() contexts (not AgentExecutorResponse), which are more nuanced. Happy to submit a follow-up PR for those if desired.