Streamlit with Prompt flow Demo

This project demonstrates how to use Streamlit to build a simple chat app that uses Prompt flow to generate responses with a model deployed in Azure OpenAI Service. The app uses two flows:

chat - for generating responses to user input given the context of the conversation
make_title - for generating a title for a conversation using the initial user question

Install dependencies

You can install the dependencies using pip:

pip install -r requirements.txt

Azure OpenAI Service

The requires access to an existing Azure OpenAI Service deployment with API access.

By default, the app expects a deployment with name gpt-4-turbo.

You can change this in flows/chat/flow.dag.yaml and flows/make_title/flow.dag.yaml by setting the deployment_name.

Environment variables

You must set the following environment variables:

OPENAI_API_KEY - The API key for your Azure OpenAI Service resource
OPENAI_API_BASE - The base URL for your Azure OpenAI Service resource in the form https://<resource-name>.openai.azure.com

Run the app locally

You can run the app using the following command:

streamlit run app.py

Deploy the app

The repository uses Pulumi and GitHub actions to deploy the app to Azure App Service. To run the GitHub actions workflow, you must set the following secrets:

ARM_CLIENT_ID - The client ID of the service principal used to deploy the app
ARM_CLIENT_SECRET - The client secret of the service principal used to deploy the app
ARM_LOCATION_NAME - The Azure region to deploy the app to
ARM_SUBSCRIPTION_ID - The Azure subscription ID to deploy the app to
ARM_TENANT_ID - The Azure tenant ID to deploy the app to
OPENAI_API_BASE - The base URL for your Azure OpenAI Service resource in the form https://<resource-name>.openai.azure.com
OPENAI_API_KEY - The API key for your Azure OpenAI Service resource
PULUMI_ACCESS_TOKEN - The Pulumi access token to use for deployment