Skip to content

graph-rag

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

graph-rag

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
README.md
README.md
 
 
graph-rag-langchain-26ai.ipynb
graph-rag-langchain-26ai.ipynb
 
 
requirements.txt
requirements.txt
 
 

README.md

GraphRAG Demo

This demo shows a graph + LLM workflow using Oracle Graph, Oracle AI Database 26ai, and LangChain. It creates a small movie recommendation graph, queries graph relationships, and sends graph query results to an LLM for natural-language responses.

What's Included

Path Purpose
graph-rag-langchain-26ai.ipynb Jupyter notebook for the GraphRAG-style LangChain demo.

Prerequisites

  • Oracle AI Database 26ai or Autonomous AI Database 26ai.
  • Python and Jupyter.
  • An LLM API key or compatible provider configuration.

Install the Python dependencies from this folder:

python -m pip install -r requirements.txt

Run The Demo

Open the notebook from this folder:

jupyter lab graph-rag-langchain-26ai.ipynb

Before running the LLM cells, update the database connection values and LLM credential placeholders in the notebook.

The notebook reads these optional environment variables:

Variable Purpose
OPENAI_API_KEY API key used by langchain-openai.
OPENAI_MODEL Chat model name to use for the demo.

Objects Created

The notebook creates these database objects:

Object Type Purpose
MOVIES Table Movie titles, genres, and summaries.
MOVIES_CUSTOMER Table Sample customers.
WATCHED_MOVIE Table Which customer watched which movie and when.
CUSTOMER_WATCHED_MOVIES Property graph Graph connecting customers to watched movies.

Expected Output

Before the LLM cells, the graph query should print customer and movie rows such as John Doe watching Inception, The Matrix, and The Godfather.

After the LLM cells are configured, the notebook should produce natural-language responses about movie genres and customer viewing preferences.

Reset The Demo

Run these statements if you want to remove the GraphRAG demo objects and rerun the notebook from the beginning:

DROP PROPERTY GRAPH CUSTOMER_WATCHED_MOVIES;
DROP TABLE WATCHED_MOVIE PURGE;
DROP TABLE MOVIES_CUSTOMER PURGE;
DROP TABLE MOVIES PURGE;

If an object does not exist yet, your SQL client may show an expected drop error. Continue with the remaining cleanup statements.

Notes

This demo is separate from the Bank Graph demo. It creates its own movie/customer graph inside the notebook and does not use the Bank Graph CSV files.

Related Links