cookbook(langgraph): use gpt-4.1-mini + sharpen Pattern-1 prompts

[DK09876]

Summary

Two related changes to notebooks/06-langgraph-react-agent.ipynb to make the memory demos work reliably end-to-end. Pairs with vectorize-io/hindsight#1673.

Same template as the OpenAI Agents polish PR #40 — both notebooks were showing the same gpt-4o-mini tool-use variance under the V1/V2 audit's cookbook checks.

What & why

1. Model upgrade: gpt-4o-mini → gpt-4.1-mini (cells 3 and 9)

Under gpt-4o-mini, cell 7's cross-session recall returned:

"It looks like I don't have any information stored about your IDE or your job. Could you let me know what you use and what your job is?"

Even though cell 5 had been intended to store Alice's facts via hindsight_retain. With gpt-4.1-mini the agent reliably calls the tool both ways. Cell 7 now returns:

"You use the IDE Visual Studio Code (VS Code), and your job is a data scientist working with Python and SQL."

2. Pattern-1 prompt sharpening (cells 7 and 8)

Same template the Claude SDK cookbook adopted (commit d92ccab373 on main): nudge the agent to call recall_memory explicitly before answering, instead of leaving the tool-use decision to the model.

• Cell 7: "What IDE do I use? And what's my job?" → "Recall what you know about me first, then answer: what IDE do I use and what's my job?"
• Cell 8: same template ("Use the recall tool to refresh what you know about me first, then recommend...").

Cell 8 now returns memory-grounded recommendations:

"Alice, based on what I know about you as a data scientist who specializes in Python and SQL and prefers dark mode in VS Code, here is a recommended setup..."

Relationship to other PRs

• Pairs with LangGraph: add memory_instructions, fix nodes, remove BaseStore hindsight#1673 (integration code unchanged here).
• Pairs with cookbook(openai-agents): use gpt-4.1-mini + sharpen Pattern-1 prompts #40 (OpenAI Agents) — same model+prompt pattern fix for the same gpt-4o-mini variance issue.
• The unmerged PR LangGraph notebook: replace BaseStore with memory_instructions #36 (dk/langgraph-basestore-cookbook) adds a Pattern 3 / memory_instructions cell that also benefits from the gpt-4.1-mini bump. Once LangGraph notebook: replace BaseStore with memory_instructions #36 lands, that cell can be polished in a follow-up.

Test plan

• Cookbook 16/16 cells execute against canonical PR #1673 build, 0 errors.
• Cell 7 (Pattern 1 cross-session) reliably surfaces Alice's facts (VS Code / data scientist / Python+SQL).
• Cell 8 (recommendations) grounds recommendations in recalled context.
• Pattern 2 (Graph nodes, cells 11+13) unchanged behavior.
• Reviewer to verify end-to-end against the merged hindsight-langgraph.

🤖 Generated with Claude Code