hyeonsangjeon / boa303

repo for demos Unlock insights with AWS GenAI services (re:Invent BOA303)

Demos for Unlock insights with AWS GenAI services (re:Invent BOA303)

This repository is complementary to breakout session on "BOA303 - Unlock insights with AWS GenAI services" at re:Invent 2023

🎥 Watch the session here

In these demos, you will explore how RAG ( Retrieval-Augmented Generation ) can enhance AI models by leveraging external data sources to provide context-aware answers and unlock insights.

Learn 2 ways to use your data with GenAI models:

1️⃣ DEMO 1: Deploying Llama2 with Sagemaker, CodeWhisperer + Kendra for data retrieval

2️⃣ DEMO 2: Using Claude 2 with Bedrock + Vector DB for data retrieval

DEMO 1 - RAG approach with Amazon Kendra

Prerequisites:

1. Follow instructions to setup Amazon SageMaker Studio prerequisites

2. Follow instructions to enable Amazon CodeWhisperer extension for SageMaker Studio

Setup for Demo 1

1. Create S3 Bucket and download data

• Open Amazon S3 console. Click Create a bucket. Provide a unique name reinvent-demo-yourname. Keep other options as default. Click Create bucket.
• Go to https://www.macrotrends.net/countries/USA/united-states/inflation-rate-cpi. Click download historical data.
• Download Starbucks annual report PDF.
• Upload downloaded csv and pdf files into S3 bucket.

2. Amazon Kendra Index & data sources

• Open Amazon Kendra console. Click Create an index. Provide index name. Under IAM Select Create a new role (Recommended). Enter role name. Click Next, then Next in the follow-up screens, and Click Create.

• Copy index name and save for future references. Wait for index to become active and click Add data sources.
• Find Amazon S3 Connector. Click Add connector.

• Provide Data Source name. Click Next. Create a new role or select an existing role.
• Click Browse S3 and select the bucket that you created in Step 1. At the bottom select Frequency Run on demand. Click Next, then Next again then Add data resources.

• Once the data source is created, click Sync Now button at the top.
• Let’s add one more data source. Click on Data sources on the left pane then click Add data source. Search for Web and pick Web Crawler V2.0.

• Provide data source name and click Next. Copy URL from step 1 into Source URLs. Select Create a new role (Recommended) at the bottom, type name, and click Next.

• Keep default options and select Frequency Run on demand. Click Next, then Next again then Add data resources.
• Once the data source is created, click Sync Now button at the top.

Demo 1 - Part 1: Notebook updates

1. Open SageMaker Studio and copy Demo1 - RAG with SageMaker and Kendra.ipynb notebook.

2. Install dependencies as needed

   !pip install —upgrade —quiet sagemaker
   !pip install langchain
   
   

3. Update s3_path with path to your bucket created in prerequisites. Update if needed csv file name

Demo 1 - Part 2: Notebook updates

1. Copy your SageMaker Endpoint and replace in the code. In SageMaker Console, go to Sidebar > Home > Deployment dropdown > Endpoints — copy the name of the LLM and paste in the 3rd row below

Demo 1 - Part 3: Notebook updates

1. Update Kendra IndexID in the code and AWS region.

Run your notebook and ask questions! 🚀

DEMO 2 - RAG approach with VectorDB

Prerequisites:

1. Follow instructions to setup Amazon SageMaker Studio prerequisites

2. Follow instructions to enable Amazon CodeWhisperer extension for SageMaker Studio

3. Follow instructions to request access to Antropic Claude model inside Amazon Bedrock

4. Create a free account with pinecone.io and create index “demoindex”.

   1. Click indexes on the left pane then click button Create index.
   2. Give an index the name “demoindex”. Configure dimensions 1536, metric cosine, then click Create index. Save index name and environment name in your notes. We will use it later.
   3. Go to API keys on the left pane and copy API key. Save it. We will use it later.

Demo 2 - Part 1:Bedrock Configuration

1. Open SageMaker Studio and copy Demo2 - RAG with Bedrock and Vector DB.ipynb notebook.

2. Create a keys.env file and add your Pinecone account information. Save this file in the same folder as your notebooks.

   PINECONE_API_KEY=your_pinecone_API_key
   PINECONE_ENVIRONMENT=your_pinecone_env
   

3. Install dependencies as needed

   !pip install -U \
   langchain==0.0.306 \
   boto3 \
   botocore \
   pypdf==3.15.2 \
   pinecone-client==2.2.4 \
   apache-beam==2.50.0 \
   datasets==2.14.5 \
   tiktoken==0.4.0 —force-reinstall —quiet
   
   
   

Execute commands in the notebook! 🚀