OpenAI-UI

The OpenAI UI is designed to be a simple Azure OpenAI frontend. The application is build with Svelte. The App focuses on:

• Easy to setup and integrate
• Easy to customise
• Build on Azure

Setup

There are many ways to use the app. A basic way to get started, is to set up an Azure AppService and pull the current latest image from the ghcr.io repository of the app.

Environment variables

To get the app up and running, several environment variables are required. The bicep scripts will set up the environment.

// AI Search
AiSearch_ContentFields=
AiSearch_Endpoint="your ai search endpoint"
AiSearch_FilePathField=
AiSearch_IndexName="your ai search index name"
AiSearch_Key="your ai search key"
AiSearch_QueryType="'simple', 'semantic', 'vector', 'vectorSimpleHybrid' or 'vectorSemanticHybrid"
AiSearch_SemanticConfiguration=
AiSearch_TitleField=
AiSearch_UrlField=

// Authorization
App_ClaimName=""
App_ClaimValue=""

// Database
Database_AccountEndpoint="your database endpoint"
Database_AccountKey="your database account key"
Database_CollectionName="your database collection name for the search history"
Database_DatabaseName="your database name"

// OpenAI
OpenAi_ApiVersion="2023-10-01-preview"
OpenAi_Embedding="your OpenAI embedding deployment name"
OpenAi_Endpoint="your OpenAI endpoint"
OpenAi_FrequencyPenalty="0.0"
OpenAi_Key="your OpenAI key"
OpenAi_MaxTokens="number of tokens per message"
OpenAi_PastMessagesIncluded="max amount of messages to send to the AI service"
OpenAi_PresencePenalty="0.0"
OpenAi_StopSequences="<|im_end|>"
OpenAi_SystemMessage="your system message to customize the ai responses"
OpenAi_TokenLimit="the token limit for requests"

// Frontend (unsafe)
PUBLIC_App_Autosave="history is stored if true, otherwise new chats are not stored in the db, can be overwritten by the user"
PUBLIC_App_UseHistory="if false, database connection is not required and history is not available to the user"
PUBLIC_App_Version="version of the app"
PUBLIC_App_UseMock="if true, no ai service connection will be established"
PUBLIC_App_UseDocumentSearch="if true, enables search in the configured ai search index"
PUBLIC_OpenAi_Deployments="list of available ai deployments, e.g. gpt-3|gpt-4"

Debug

To debug the application run:

npm install
cp .env.example .env.test
npm run dev -- --mode test

if you have an environment, update the environment variables inside .env.test with the ones matching your environment.

Build

npm run build

To test the build locally, run:

npm run node

and open the browser on http://localhost:3000

Docker

To host the app in a docker container either pull the image from the ghcr.io repository or:

docker build --pull --rm -f "Dockerfile" -t openaiui:latest "."
docker run --rm -d -p 80:80/tcp openaiui:latest

Azure

To develop and deploy the app to Azure, install the Azure cli and login e.g.:

az login
az account set --subscription xxxxxxx-xxxxx-xxxx-xxxxx-xxxxxxxy

Deploy the azure environment:

az deployment sub create --name openai-bicep-deployment --confirm-with-what-if --template-file ./infra/main.bicep --location NorthEurope --parameters ./infra/main.parameters.dev.json

Clean up:

az deployment sub delete --name openai-bicep-deployment