https://github.com/MicrosoftJames/streamlit-with-promptflow

https://github.com/MicrosoftJames/streamlit-with-promptflow

Last synced: about 2 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/MicrosoftJames/streamlit-with-promptflow
• Owner: MicrosoftJames
• Created: 2023-12-22T11:34:14.000Z (9 months ago)
• Default Branch: main
• Last Pushed: 2024-01-26T11:49:34.000Z (8 months ago)
• Last Synced: 2024-04-27T16:32:14.906Z (5 months ago)
• Language: Python
• Size: 604 KB
• Stars: 15
• Watchers: 1
• Forks: 4
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        


  



# Streamlit with Prompt flow Demo



  



This project demonstrates how to use [Streamlit](https://streamlit.io/) to build a simple chat app that uses [Prompt flow](https://microsoft.github.io/promptflow) to generate responses with a model deployed in Azure OpenAI Service. The app uses two flows:

1. [chat](flows/chat) - for generating responses to user input given the context of the conversation

2. [make_title](flows/make_title/) - for generating a title for a conversation using the initial user question

# Install dependencies

You can install the dependencies using pip:

```bash

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](flows/chat/flow.dag.yaml) and [flows/make_title/flow.dag.yaml](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://.openai.azure.com`

# Run the app locally

You can run the app using the following command:

```bash

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://.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

Your service principal must have the ability to create to app registrations and managed identities.