Update sample code for: LangGraph Tutorial: Build Stateful AI Agents in Python

[martin-martin]

Where to put new files:

• New files should go into a top-level subfolder, named after the article slug. For example: my-awesome-article

How to merge your changes:

1. Make sure the CI code style tests all pass (+ run the automatic code formatter if necessary).
2. Find an RP Team member on Slack and ask them to review & approve your PR.
3. Once the PR has one positive ("approved") review, GitHub lets you merge the PR.
4. 🎉

---

Updates the existing python-langgraph/ companion code for the LangGraph Tutorial: Build Stateful AI Agents in Python tutorial refresh.

Changes

• Bump to langgraph 1.2.5, langchain-openai 1.3.2, langchain-core 1.4.7, pydantic 2.13.4 (regenerated requirements.txt; raised pyproject.toml floors)
• Switch the model to gpt-5.4-nano
• Fix check_escalation_status_node to fall back to 0 when no fine is extracted (avoids a TypeError on notices without a max potential fine)
• Store the boolean answer.is_true in follow_ups (was storing the model object, violating the dict[str, bool] type)
• Drop the unused critical_fields_missing key from the agent's initial state
• Point the README API-key link at platform.openai.com/api-keys and align the README title

Verified end-to-end against OpenAI on Python 3.12 and 3.14 with the pinned requirements.txt (structured output, the escalation/no-fine path, the follow-up cycle, and the agent). ruff format/ruff check (v0.14.1) pass.