Update LangGraph notebook dependencies and env-driven setup

[dex-the-ai]

Summary

• Update langgraph/couchbase_persistence_langgraph.ipynb dependencies from current main and keep the explicit langgraph==1.1.10 pin.
• Add langchain-openai, switch the sample agent construction to create_agent, and move the notebook to env-driven local Couchbase/OpenAI configuration.
• Clarify the notebook text around required environment variables and the local Couchbase bucket/scope/collections.

Verification

• uv venv -p 3.12 --seed --clear .venv-fresh-pr-20260429
• uv pip install papermill ipykernel pip-audit
• papermill -k vsc-langgraph-fresh-20260429 langgraph/couchbase_persistence_langgraph.ipynb ... ✅
• pip-audit ✅ (No known vulnerabilities found)

Evidence

• Fresh rerun from main executed the notebook end-to-end with papermill on a clean Python 3.12 environment.
• The executed notebook artifact includes both sync and async checkpoint flows.
• The notebook now relies on CB_* / OPENAI_API_KEY environment variables instead of hard-coded local values.
• Visual evidence is included for the rendered notebook output used in review.

Rendered notebook screenshot

Papermill / audit summary

• papermill completed all 27 cells successfully with kernel vsc-langgraph-fresh-20260429
• pip-audit reported No known vulnerabilities found
• the first rerun attempt only failed because OPENAI_API_KEY was not exported into the kernel environment; rerunning with exported env fixed it

• tutorial-maintenance/runs/couchbase-examples__vector-search-cookbook/2026-04-29-langgraph-fresh-from-main/verification.md
• tutorial-maintenance/runs/couchbase-examples__vector-search-cookbook/2026-04-29-langgraph-fresh-from-main/papermill.log
• tutorial-maintenance/runs/couchbase-examples__vector-search-cookbook/2026-04-29-langgraph-fresh-from-main/dependency-freeze.txt
• tutorial-maintenance/runs/couchbase-examples__vector-search-cookbook/2026-04-29-langgraph-fresh-from-main/screenshots/langgraph-executed-notebook.png

Notes

• This stays artifact-scoped to langgraph/couchbase_persistence_langgraph.ipynb rather than bundling broader cookbook changes.
• The PR title now reflects the actual work: dependency updates plus the env-driven notebook/runtime cleanup needed to keep the artifact executable.

[github-actions]

Caution

Notebooks or Frontmatter Files Have Been Modified

• Please ensure that a frontmatter.md file is accompanying the notebook file, and that the frontmatter is up to date.
• These changes will be published to the developer portal tutorials only if frontmatter.md is included.
• Proofread all changes before merging, as changes to notebook and frontmatter content will update the developer tutorial.

1 Notebook Files Modified:

Notebook File	Frontmatter Included?
langgraph/couchbase_persistence_langgraph.ipynb	✅

0 Frontmatter Files Modified:

Frontmatter File
Note: frontmatter will be checked and tested in the Test Frontmatter workflow.

[gemini-code-assist]

Code Review

This pull request updates the Couchbase persistence notebook for LangGraph to support environment-based configuration and adds the langchain-openai dependency. Several critical issues were identified: the langgraph package version was pinned to a non-existent version (1.1.10), and the code was changed to use create_agent from langchain.agents instead of create_react_agent from langgraph.prebuilt. The latter change is problematic because standard LangChain agents do not support the checkpointer parameter required for the persistence features demonstrated in this notebook.

[gemini-code-assist]

The version langgraph==1.1.10 appears to be incorrect. As of the current release cycle, LangGraph versions are in the 0.x range (e.g., 0.2.x or 0.3.x). Pinning to a non-existent version will cause the %pip install command to fail. Additionally, the PR description mentions keeping the pin, but the diff shows a change from 0.3.22 to 1.1.10.

Suggested change

	"%pip install -U langgraph==1.1.10 langgraph-checkpointer-couchbase langchain-openai\n"
	"%pip install -U langgraph langgraph-checkpointer-couchbase langchain-openai\n"

[gemini-code-assist]

The import create_agent from langchain.agents is likely incorrect for a LangGraph tutorial. LangGraph uses create_react_agent from langgraph.prebuilt to create compiled graphs that support persistence via checkpointers. langchain.agents does not provide a create_agent function that accepts a checkpointer argument, and using the legacy agent stack would break the persistence logic demonstrated in this notebook.

Suggested change

	"from langchain.agents import create_agent\n",
	"from langgraph.prebuilt import create_react_agent\n",

[gemini-code-assist]

Using create_agent here is likely incorrect. Since this notebook demonstrates LangGraph persistence, you should use create_react_agent to ensure the resulting object is a LangGraph CompiledGraph that correctly utilizes the provided checkpointer. Standard LangChain agent factory functions do not support the checkpointer parameter.

Suggested change

	" graph = create_agent(model, tools=tools, checkpointer=checkpointer)\n",
	" graph = create_react_agent(model, tools=tools, checkpointer=checkpointer)\n",

[gemini-code-assist]

As noted in the synchronous example, create_react_agent should be used instead of create_agent to maintain compatibility with LangGraph's checkpointer logic and ensure the graph is properly compiled for persistence.

Suggested change

	" graph = create_agent(model, tools=tools, checkpointer=checkpointer)\n",
	" graph = create_react_agent(model, tools=tools, checkpointer=checkpointer)\n",