azure-open-ai-batch-demo

A demo on how to run Azure OpenAI generation batch jobs on ACA.

Prerequisites

• Python 3.6 or later

• Azure CLI

• Azure Service Bus namespace (Will be created by the setup.sh script if it doesn't already exist)

Setup

1. Clone the repository:

       git clone https://github.com/Azure-Samples/azure-open-ai-batch-demo.git    

2. Create a Service Bus and deploy Azure Container App:

   ./setup.sh

This script creates a new resource group, Service Bus namespace, and two queues (pendingPrompts and generatedPrompts) if they don't already exist. It also outputs the connection string for the Service Bus namespace.

3. Set the SERVICE_BUS_CONN_STR environment variable:

   export SERVICE_BUS_CONN_STR=<your-connection-string>

Replace <your-connection-string> with the connection string output by the setup.sh script.

4. Install the Python dependencies:

   pip install -r requirements.txt

Usage

1. Send prompts to the pendingPrompts queue:

   python send_prompts.py

   This script reads a list of services from azure-services.txt and sends a prompt for each service to the pendingPrompts queue.

2. Generate responses using Azure OpenAI:

   python generate_responses.py

   This script reads prompts from the pendingPrompts queue, generates responses using Azure OpenAI, and writes the responses to generated-responses.txt.