rag-question-answer

Copyright IBM Corp. 2020, 2024

RAG question & answer application | IBM watsonx.ai Flows Engine

This is an example "question and answer" application built using watsonx.ai Flows Engine. You can recreate the content in this directory by following along this tutorial. The example uses a vector database to make a LLM context aware of a set of documents.

Deploy the flows

To deploy the RAG flows, make sure to install and authenticate the wxflows CLI. Then add your credentials to watsonx.ai in the .env file or use the shared "getting started" instance:

# Shared "getting started"
STEPZEN_WATSONX_HOST=shared

# Your own watsonx.ai credentials (optionally)
# STEPZEN_WATSONX_AI_APIKEY=
# STEPZEN_WATSONX_HOST=
# STEPZEN_WATSONX_PROJECTID=

The use of STEPZEN_WATSONX_AI_TOKEN is deprecated since wxflows 1.0.0rc189 and up. Please update your CLI when you're on an older version.

Extract the file watsonxdocs.zip, it contains a set of HTML documents that we'll prepare for upload to the vector database.

Run the command wxflows init --interactive and follow the steps, use ./watsonxdocs as the data source and watsonxdocs as the collection name.

Uncomment the flow myRag like this:

[wxflows.deployment]
# Example flows.
flows="""
    // myPrompt = ragAnswerInput | topNDocs | promptFromTopN | ragInfo
    myRag = ragAnswerInput | topNDocs | promptFromTopN | completion(parameters:myRag.parameters) | ragInfo
    // myRagWithGuardrails = ragAnswerInput | topNDocs | promptFromTopN | completion(parameters:myRagWithGuardrails.parameters) | ragScoreInfo | hallucinationScore | ragScoreMessage | ragInfo
"""

The data needs to be uploaded to the vector database, which you can do by running wxflows collection deploy. Running this command will take a couple of minutes to finish.

Then you can run the command wxflows deploy to deploy your flows to a live endpoint.

Run the application in the browser

You can run this application in the browser using the free service StackBlitz:

In the .env file add the values generated by the following command:

   echo VITE_WXFLOWS_ENDPOINT=https://$(wxflows whoami --account).$(wxflows whoami --domain)/wxflows-genai/watsonxdocs/graphql 
   echo VITE_WXFLOWS_APIKEY=$(wxflows whoami --apikey)
   echo VITE_WXFLOWS_COLLECTION=watsonxdocs

You need to be logged into the wxflows CLI to generate the above values!

Run the application locally

Prerequisites

• Follow the instructions to set up a RAG flow

Installation

You need to have Node.js (>16) installed on your machine, from the directory app run:

npm install

This will install all the needed dependencies for the project.

Add your environment variables for watsonx.ai Flows Engine to a new .env file, you need to be logged into the wxflows CLI, by running:

   echo VITE_WXFLOWS_ENDPOINT=https://$(wxflows whoami --account).$(wxflows whoami --domain)/wxflows-genai/watsonxdocs/graphql >> .env
   echo VITE_WXFLOWS_APIKEY=$(wxflows whoami --apikey) >> .env
   echo VITE_WXFLOWS_COLLECTION='watsonxdocs' >> .env

Run the application

You can now start the application by running:

npm run dev

The location of the application will be shown in your terminal.

Make sure the flowName variable on line 12 in /src/callWxflows.js match your value in wxflows.toml.

Support

Please reach out to us on Discord if you have any questions or want to share feedback. We'd love to hear from you!