feat: graph rag with Vertex AI tutorial

[codingphun]

Description

Thank you for opening a Pull Request!
Before submitting your PR, there are a few things you can do to make sure it goes smoothly:

• Follow the CONTRIBUTING Guide.
• You are listed as the author in your notebook or README file.
  • Your account is listed in CODEOWNERS for the file(s).
• Make your Pull Request title in the https://www.conventionalcommits.org/ specification.
• Ensure the tests and linter pass (Run nox -s format from the repository root to format).
• Appropriate docs were updated (if necessary)

Fixes #<issue_number_goes_here> 🦕

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[koverholt]

Thanks for the contribution of this sample notebook! I left some comments throughout, but mainly this notebook needs to be formatted more closely to the notebook template in this repo.

[koverholt]

Update all of these links to point to the tutorial_graph_rag.ipynb file.

[koverholt]

Use a more descriptive title, such as Building a GraphRAG Question-Answering System with LangChain on Vertex AI or similar.

[koverholt]

Is there a canonical link you can use in the Google Cloud or LangChain docs for users looking for more information on the GraphRAG approach?

[koverholt]

Please include all of the standard setup and config steps from the notebook template in this repo.

[koverholt]

I've uploaded this image to our public GCS bucket associated with this repo, so you can reference it as a Markdown image or HTML image reference instead of base64 encoded:

https://storage.googleapis.com/github-repo/generative-ai/gemini/reasoning-engine/images/graphrag.png

[koverholt]

Use in-line code formatting for syntax such as LLMGraphTransformer.

[koverholt]

Be sure to stylize Langchain correctly as LangChain here and throughout.

[koverholt]

Use gemini-1.5-pro instead of a specific checkpoint to make this more future-proof for the Gemini 1.5 series.

[koverholt]

Have you tested (or can you test) this with all package versions set to latest available versions? For example, the current version of google-cloud-aiplatform is 1.72.0.

[koverholt]

We should have users define all constants in a cell near the top (like you have), then reuse that throughout to avoid having to re-define these in multiple places throughout the notebook.

[holtskinner]

Suggested change

	"In this tutorial, you will see a complete walkthrough of building a question-answering system using the GraphRAG method. You'll learn how to create a knowledge graph from scratch, store it efficiently in Spanner Graph, enhance search accuracy with embedding vectors in Spanner Vector Database, and finally, deploy a functional FAQ system with Langchain and Vertex AI Reasoning Engine."
	"In this tutorial, you will see a complete walkthrough of building a question-answering system using the GraphRAG method. You'll learn how to create a knowledge graph from scratch, store it efficiently in Spanner Graph, enhance search accuracy with embedding vectors in Spanner Vector Database, and finally, deploy a functional FAQ system with LangChain and Vertex AI Reasoning Engine."

[holtskinner]

Can you add links to the docs for each of the features/products mentioned?

[holtskinner]

Put all imports here.

[holtskinner]

Remove these image outputs, they make the notebook very large and difficult to load.

[holtskinner]

move these imports to the top of the notebook

[holtskinner]

This database insert seems very complicated. Is there a simpler way to notate all of this?

[holtskinner]

Suggested change

	"Now the graph nodes, relationship and embedding vectors are stored in the Spanner database. Let's try to query the database. We will ask the original question \"Who influced Larry Page?\" in the Vector database to get a semantic search result. Then we will get the node \"Lawrence Edward Page\" from the result."
	"Now the graph nodes, relationship and embedding vectors are stored in the Spanner database. Let's try to query the database. We will ask the original question \"Who influenced Larry Page?\" in the Vector database to get a semantic search result. Then we will get the node \"Lawrence Edward Page\" from the result."

[holtskinner]

A lot of this seems to be repeated in multiple cells. Can this be condensed?

[holtskinner]

Do you need to build a custom retriever for this? There should already be a Spanner Retriever in LangChain.

Also, would it be possible to just define these methods once in the Chain instead of running them multiple times in the notebook? It seems like there's lots of duplication.

[holtskinner]

Suggested change

	"Give service-PROJECT_NUMBER@gcp-sa-aiplatform-re.iam.gserviceaccount.com Cloud Spanner Database User role or Vertex AI Reasoning Engine will not have access to the Spanner database"
	"Give `service-PROJECT_NUMBER@gcp-sa-aiplatform-re.iam.gserviceaccount.com` Cloud Spanner Database User role or Vertex AI Reasoning Engine will not have access to the Spanner database"