42 lines (27 loc) · 1.95 KB

Launch DIAL Chat with Azure Model

Introduction

In this tutorial, you will learn how to quickly launch DIAL Chat with a custom model created in Azure OpenAI Studio.

Prerequisites

Docker engine installed on your machine (Docker Compose Version 2.20.0 +).

Refer to Docker documentation.
Account in MS Azure OpenAI Studio.

Refer to Create and Deploy OpenAI Model in Azure to learn how to create and deploy a model in your MS Azure.

Step 1: Get DIAL

Download a folder with the necessary files to launch DIAL.

Step 2: Configuration

In the dial-docker-compose/model/core folder, you can find a config.json configuration file.

In config.json, you can add your Azure model credentials to the DIAL Core configuration:

Supply your Azure API Keys for your deployments for the key parameter.
Replace http://azure_deployment_host with your GPT endpoint for the endpoint parameter. Note: in the endpoint, replace gpt-4 with your Azure deployment name, in case it is different.
```
      "upstreams": [
      {
        "endpoint": "http://azure_deployment_host/openai/deployments/gpt-4/chat/completions",
        "key": "AZURE_MODEL_API_KEY"
      }
]
```
Refer to Create and Deploy OpenAI Model in Azure to learn how to create and deploy a model in MS Azure.

Step 3: Launch DIAL Chat

Run docker compose up from the folder with the docker-compose file.
Open http://localhost:3000/ in your browser to launch the DIAL Chat application.