diff --git a/data-load.ipynb b/data-load.ipynb
index c6d8eca..65bf02b 100644
--- a/data-load.ipynb
+++ b/data-load.ipynb
@@ -8,29 +8,23 @@
"# Neo4j Generative AI - Data Loading\n",
"[](https://colab.research.google.com/github/neo4j-product-examples/genai-workshop/blob/main/data-load.ipynb)\n",
"\n",
- "This workshop will teach you to how to use Neo4j for Graph-Powered Retrieval-Augmented Generation (GraphRAG) to enhance GenAI and improve response quality for real-world applications.\n",
+ "This workshop will teach you how to use Neo4j for Graph-Powered Retrieval-Augmented Generation (GraphRAG) to enhance GenAI and improve response quality for real-world applications.\n",
"\n",
"GenAI, despite its potential, faces challenges like hallucination and lack of domain knowledge. GraphRAG addresses these issues by combining vector search with knowledge graphs and data science techniques. This integration helps improve context, semantic understanding, and personalization, making Large Language Models (LLMs) more effective for critical applications.\n",
"\n",
"We walk through an example that uses real-world customer and product data from a fashion, style, and beauty retailer. We show how you can use a knowledge graph to ground an LLM, enabling it to build tailored marketing content personalized to each customer based on their interests and shared purchase histories. We use Retrieval-Augmented Generation (RAG) to accomplish this, specifically leveraging not just vector search but also graph pattern matching and graph machine learning to provide more relevant personalized results to customers. We call this graph-powered RAG approach “GraphRAG” for short.\n",
"\n",
"This notebook walks through the first steps of the process, including:\n",
- "- Building the knowledge graph and generating text embeddings from scratch \n",
+ "- Building the knowledge graph and\n",
+ "- generating text embeddings from scratch\n",
"\n",
- "*If you would rather start from a database dump and skip data loading, please skip to [genai-workshop.ipynb](https://github.com/neo4j-product-examples/genai-workshop/blob/main/genai-workshop.ipynb)* \n",
+ "[genai-workshop.ipynb](https://github.com/neo4j-product-examples/genai-workshop/blob/main/genai-workshop.ipynb) contains the rest of the workshop including\n",
+ " - Vector search\n",
+ " - Graph patterns to improve semantic search\n",
+ " - Augmenting semantic search with graph data science\n",
+ " - Building an example LLM chain and demo app\n",
"\n",
- "\n",
- "\n",
- "Please also see the following companion notebooks to complete the workshop: \n",
- "\n",
- "- [genai-workshop.ipynb](https://github.com/neo4j-product-examples/genai-workshop/blob/main/genai-workshop.ipynb)\n",
- " - Vector search \n",
- " - Using graph patterns in Cypher to improve semantic search with context\n",
- " - Further augmenting semantic search with knowledge graph inference & graph data science \n",
- " - Building the LLM chain and demo app for generating content \n",
- " \n",
- "- [genai-example-app-only](https://github.com/neo4j-product-examples/genai-workshop/blob/main/genai-workshop-app-only.ipynb)\n",
- " - Building the LLM chain and demo app for generating content \n",
+ "If you would rather start from a database dump and skip this data loading, you can do so using [this dump file](https://storage.googleapis.com/gds-training-materials/Version8_Jan2024/neo4j_genai_hnm.dump).\n",
" "
]
},
@@ -39,7 +33,7 @@
"id": "a527357f-a5e8-42c9-8966-f37846097c1d",
"metadata": {},
"source": [
- "### Some Logics\n",
+ "### Some Logistics\n",
"1. Run the pip install below to get the necessary dependencies. this can take a while. Then run the following cell to import relevant libraries\n",
"2. You will need a Neo4j database environment with the [graph data science library](https://neo4j.com/docs/graph-data-science/current/installation) installed e.g. \n",
" - [AuraDS](https://neo4j.com/docs/aura/aurads/) \n",
@@ -54,12 +48,12 @@
"metadata": {
"tags": []
},
- "outputs": [],
"source": [
"%%capture\n",
- "%pip install sentence_transformers langchain langchain-openai langchain_community openai tiktoken python-dotenv gradio graphdatascience altair neo4j_tools\n",
+ "%pip install sentence_transformers langchain langchain-openai langchain_community openai tiktoken python-dotenv gradio graphdatascience\n",
"%pip install \"vegafusion[embed]\""
- ]
+ ],
+ "outputs": []
},
{
"cell_type": "code",
@@ -68,14 +62,12 @@
"metadata": {
"tags": []
},
- "outputs": [],
"source": [
"import pandas as pd\n",
"import numpy as np\n",
"from dotenv import load_dotenv\n",
"import os\n",
"from graphdatascience import GraphDataScience\n",
- "from neo4j_tools import gds_db_load, gds_utils\n",
"from langchain_openai import OpenAIEmbeddings, ChatOpenAI\n",
"from langchain.vectorstores.neo4j_vector import Neo4jVector\n",
"from langchain.graphs import Neo4jGraph\n",
@@ -83,7 +75,8 @@
"from langchain.schema import StrOutputParser\n",
"from langchain.schema.runnable import RunnableLambda\n",
"import gradio as gr"
- ]
+ ],
+ "outputs": []
},
{
"cell_type": "code",
@@ -92,12 +85,12 @@
"metadata": {
"tags": []
},
- "outputs": [],
"source": [
"pd.set_option('display.max_rows', 10)\n",
"pd.set_option('display.max_colwidth', 500)\n",
"pd.set_option('display.width', 0)"
- ]
+ ],
+ "outputs": []
},
{
"cell_type": "markdown",
@@ -114,7 +107,7 @@
"\n",
"To make this easy, you can write the credentials and env variables directly into the below cell.\n",
"\n",
- "Alternatively, if you like, you can use an environments file. This is a best practice for the future, but fine to skip if you're just exploring."
+ "Alternatively, if you like, you can use an environment file. This is a best practice for the future, but fine to skip for now."
]
},
{
@@ -124,7 +117,6 @@
"metadata": {
"tags": []
},
- "outputs": [],
"source": [
"# Neo4j\n",
"NEO4J_URI = 'copy_paste_your_db_uri_here' #change this\n",
@@ -136,7 +128,8 @@
"LLM = 'gpt-4o'\n",
"os.environ['OPENAI_API_KEY'] = 'sk-...' #change this\n",
"OPENAI_API_KEY = os.getenv('OPENAI_API_KEY')"
- ]
+ ],
+ "outputs": []
},
{
"cell_type": "code",
@@ -145,7 +138,6 @@
"metadata": {
"tags": []
},
- "outputs": [],
"source": [
"# You can skip this cell if not using a ws.env file - alternative to above\n",
"from dotenv import load_dotenv\n",
@@ -163,7 +155,8 @@
" # AI\n",
" LLM = 'gpt-4o'\n",
" OPENAI_API_KEY = os.getenv('OPENAI_API_KEY')"
- ]
+ ],
+ "outputs": []
},
{
"cell_type": "markdown",
@@ -202,7 +195,6 @@
"metadata": {
"tags": []
},
- "outputs": [],
"source": [
"# Use Neo4j URI and credentials according to our setup\n",
"gds = GraphDataScience(\n",
@@ -212,7 +204,8 @@
"\n",
"# Necessary if you enabled Arrow on the db - this is true for AuraDS\n",
"gds.set_database(\"neo4j\")"
- ]
+ ],
+ "outputs": []
},
{
"cell_type": "markdown",
@@ -229,21 +222,10 @@
"metadata": {
"tags": []
},
- "outputs": [
- {
- "data": {
- "text/plain": [
- "'2.9.0+73'"
- ]
- },
- "execution_count": 12,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
"source": [
"gds.version()"
- ]
+ ],
+ "outputs": []
},
{
"cell_type": "markdown",
@@ -254,7 +236,7 @@
"This workshop will leverage the [H&M Personalized Fashion Recommendations Dataset](https://www.kaggle.com/competitions/h-and-m-personalized-fashion-recommendations/data), a sample of real customer purchase data that includes rich information around products including names, types, descriptions, department sections, etc.\n",
"\n",
"*Bonus!*\n",
- "The data we use is a slightly sampled and preformatted version of the kaggle data. If you are interested in what we did, you can find the details [here](https://github.com/neo4j-product-examples/genai-workshop/blob/main/data-prep.ipynb)"
+ "The data we use is a sampled and preformatted version of the Kaggle data. If you are interested in what we did, you can find the details [here](https://github.com/neo4j-product-examples/genai-workshop/blob/main/data-prep.ipynb)"
]
},
{
@@ -264,7 +246,6 @@
"metadata": {
"tags": []
},
- "outputs": [],
"source": [
"import pandas as pd\n",
"\n",
@@ -274,7 +255,8 @@
"article_df = pd.read_csv('https://storage.googleapis.com/neo4j-workshop-data/genai-hm/article.csv')\n",
"customer_df = pd.read_csv('https://storage.googleapis.com/neo4j-workshop-data/genai-hm/customer.csv')\n",
"transaction_df = pd.read_csv('https://storage.googleapis.com/neo4j-workshop-data/genai-hm/transaction.csv')"
- ]
+ ],
+ "outputs": []
},
{
"cell_type": "markdown",
@@ -293,53 +275,14 @@
"metadata": {
"tags": []
},
- "outputs": [
- {
- "data": {
- "text/html": [
- "\n",
- "\n",
- "
\n",
- " \n",
- " \n",
- " | \n",
- "
\n",
- " \n",
- " \n",
- " \n",
- "
\n",
- "
\n",
- "\n",
- "
\n",
- " \n",
- " \n",
- " | \n",
- "
\n",
- " \n",
- " \n",
- " \n",
- "
\n",
- "
\n",
- "\n",
- "
\n",
- " \n",
- " \n",
- " | \n",
- "
\n",
- " \n",
- " \n",
- " \n",
- "
\n",
- "
\n",
- "\n",
- "
\n",
- " \n",
- " \n",
- " | \n",
- "
\n",
- " \n",
- " \n",
- " \n",
- "
\n",
- "