Structured outputs and provider-agnostic LLMs in LangGraph (without LangChain)

[VLazarevic]

Hi all,

I'm working on a project using LangGraph to orchestrate LLM workflows. We're aiming for two core goals:

1. Structured output parsing into Pydantic models
2. A provider-agnostic LLM design (e.g. OpenAI, Gemini, Claude)

Most structured output examples I’ve found rely heavily on LangChain, especially components like PydanticOutputParser, prompt templates, and LLM wrappers like ChatOpenAI. From what I’ve seen, LangGraph doesn’t currently provide built-in equivalents, or maybe I’ve missed something.

For reference:
🔗 https://python.langchain.com/docs/concepts/structured_outputs/

My main question is:

Is there a way to achieve structured outputs in LangGraph (like in the LangChain docs above) without building a full LLM wrapper like ChatOpenAI?

Ideally, we’d like to stay within LangGraph and avoid adding LangChain just for prompt formatting and response parsing.

Any guidance or examples would be really appreciated!