Implement Structured Output with Pydantic Models

[ommistry2914]

Summary

Adds structured output support to the ChatGradient client by introducing with_structured_output(). This returns a lightweight wrapper that parses the model's text output as JSON and validates it using a provided Pydantic model, returning typed BaseModel instances (or lists of instances).

What changed

chat_models.py
-Added with_structured_output() to ChatGradient.
-Added StructuredChatGradient helper wrapper which:
-Invokes the underlying LLM,
-Parses string output as JSON,
-Validates using Pydantic and returns model instances (or lists when multiple=True.
-Handles parsing and validation errors with clear ValueError messages.

test_structured_output.py (new)
-Unit tests for single-object success, multiple-object success, invalid JSON, and validation error cases. Tests are isolated and do not hit the network (use a small dummy LLM).

README.md
Example usage showing with_structured_output(Person)

fixes #10

[Copilot]

Pull Request Overview

This PR introduces structured output support to the ChatGradient client, enabling automatic parsing and validation of LLM responses into Pydantic models. The implementation adds a with_structured_output() method that returns a wrapper handling JSON parsing and Pydantic validation.

Key Changes:

• Added with_structured_output() method to ChatGradient that returns a StructuredChatGradient wrapper
• Implemented JSON parsing and Pydantic validation with error handling for both single and multiple object responses
• Added comprehensive unit tests using a mock LLM to validate functionality without network calls

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
langchain_gradient/chat_models.py	Implements with_structured_output() method and StructuredChatGradient wrapper class with JSON parsing and Pydantic validation
tests/unit_tests/test_structured_output.py	Adds unit tests covering success cases, invalid JSON handling, and validation errors using a dummy LLM
README.md	Documents the new structured output feature with usage example

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Creating a BaseMessage directly is incorrect. BaseMessage is likely an abstract base class from langchain_core.messages. Use a concrete message type like SystemMessage instead: from langchain_core.messages import SystemMessage and messages = [SystemMessage(content=str(rf))] + messages.

[Copilot]

Catching broad Exception is too permissive. Catch json.JSONDecodeError specifically to handle JSON parsing errors while allowing other exceptions to propagate normally.

Suggested change

	except Exception as e: # noqa: BLE001 - we want to catch JSON errors
	except json.JSONDecodeError as e: