Skip to content

Latest commit

 

History

History
100 lines (84 loc) · 5.08 KB

README.md

File metadata and controls

100 lines (84 loc) · 5.08 KB

AgenticShop: re-Imagined Shopping Experience for the Era of AI Agents

This repository showcases the power of agentic flow with llama index leveraging the functionality of multi-agent workflow. The solution has been developed to experience the enhanced shopping experience for individuals having keen interest in purchasing electronic gadgets. It is a one-click solution which is easy to test and deploy with the help of Azure developer cli. Refer to the architecture diagram for complete layout:

Key Features

Retail solution accelerator provides the following features:

  • Personalized product details based on user profile
  • Elevated user experience
  • Multi Agent Workflows allows seamless handling of multiple tasks
  • Debug panel using Arize Phoenix tracing for agent triggers and tracking

Architecture Diagram

AZURE ARCHITECTURE

Solution Accelerator Deployment

CloudShell

This solution is designed to be deployable via CloudShell. If you prefer to deploy from your local development machine, please follow the steps below.

Prerequisites

The following serve as prerequisites for deployment of this solution:

  1. Azure Developer Cli
  2. Azure Cli
  3. Azure Cli extension rdbms-connect
  4. An Azure account with an active subscription.
  5. Python 3.8+
  6. Powershell Core (for windows users only)

Deployment Steps

Note: This version of the infrastructure works and successfully deploys the working solution. However, certain modules are still in progress such as Azure Key Vault, Azure App Config etc as well as some best practices. These modules and best practices will be finalized and implemented soon.

Clone Repository

Clone the repository. Once done, navigate to the repository

git clone https://github.com/Azure-Samples/postgres-agentic-shop.git
cd postgres-agentic-shop

Login to your Azure account

To login to azure cli, use the following command:

az login

To login to azure developer cli, use this command:

azd auth login

If the above command fails, use the following flag:

azd auth login --use-device-code

Create new azure developer environment

Initialize and create a new azd environment. Provide a name for your azd environment

azd init

Grant permissions to azd hooks scripts

If you are deploying the solution on Windows OS, grant the following permissions to the current session to execute pwsh scripts

Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass

If you are deploying a unix-like environment on windows OS (for instance cygwin, minGW), grant the following permissions to the current session to execute pwsh scripts

pwsh -NoProfile -Command "Set-ExecutionPolicy -Scope CurrentUser -ExecutionPolicy Bypass"

Solution Deployment

Run the following command to provision the resources.

azd up

Once this command is executed, the prompt asks for subscription for deployment, two locations i.e. one for location of solution accelerator resources and other for location of Azure OpenAI models and the resource group to create. Make sure that you have enough Azure OpenAI model quota in the region of deployment. The Azure OpenAI quota required for this solution is listed below. This configuration can be changed from main.parameters.json file in infra directory using following parameters. The deployment might take some time and will provide progress of deployment in terminal as well as on Azure Portal.

  • GPT-4o: 140K TPM - AZURE_OPENAI_CHAT_DEPLOYMENT_CAPACITY
  • text-embedding-ada-002: 60K TPM - AZURE_OPENAI_EMBED_DEPLOYMENT_CAPACITY

Troubleshooting

  1. Troubleshooting guide for azd cli is here.
  2. Validation error occurs when unsupported characters are used while initializing or creating a new env.
  3. Scope error occurs when user does not have appropriate permissions when running azd cli. Update user permissions to subscription level.
  4. When The resource entity provisioning state is not terminal error occurs, restart the deployment using azd up command.

Tear Down

To destroy all the resources that have been created in the step above as well as remove any accounts deployed by the solution accelerator, use the following command:

azd down --purge

The purge flag deletes all the accounts permanently.

Personalization Workflow

Following is the LlamaIndex workflow generated via it's visualization tool: Screenshot