Skip to main content

Open WebUI with mittwald AI Hosting

Open WebUI can be used as a ChatGPT-like interface with mittwald AI Hosting. It can be automatically installed and configured when an API key is created if your hosting product supports containers. Otherwise, set up Open WebUI in mittwald's container hosting following our deployment guide.

Connecting to mittwald AI Hosting​

When using the managed deployment, your Open WebUI deployment will be automatically configured to use your mittwald AI hosting account. Otherwise, you can follow one of the approaches recommended here.

Using Environment Variables (Recommended)​

The recommended method is to configure the connection during container deployment using environment variables. See the deployment guide for detailed instructions.

Using the Admin Panel​

If not connected automatically, you can set up the connection in the admin panel:

  1. Go to "Settings" and choose "Connections"
  2. In the "OpenAI API" section, add another connection
  3. Insert the base URL: https://llm.aihosting.mittwald.de/v1
  4. Enter your API key

Open WebUI will automatically detect all available models.

Optimizing Model Parameters​

For optimal results, it may be necessary to adjust the default parameters of Open WebUI for each model.

  1. Navigate to the "Models" section in Open WebUI
  2. Select the model you want to configure
  3. Under "Advanced Params", apply the recommended parameters documented in the models section, such as top_p, top_k, and temperature

Using Retrieval-Augmented Generation (RAG)​

Open WebUI offers the ability to store knowledge in the form of documents, which can be accessed as needed. This is known as retrieval-augmented generation (RAG).

Uploading Documents​

  1. In the left menu bar, navigate to "Workspace"
  2. Select the "Knowledge" tab and create a new knowledge base using the "New Knowledge" button
  3. Upload documents that you want to make available
  4. In your chats, you can access these documents by using the "Attach knowledge" option on the chat input field.

Configuring an Embedding Model​

To enable more efficient document processing, you can use an embedding model:

  1. In the "Admin Settings", go to the "Documents" menu item
  2. In the "Embedding" section, select "OpenAI" in the dropdown menu as the embedding model engine
  3. Enter the endpoint: https://llm.aihosting.mittwald.de/v1
  4. Enter your generated API key
  5. Select one of the available embedding models (such as Qwen3-Embedding-8B) under "Embedding Model"
  6. In the "Retrieval" section, optionally adjust the parameters "Top K" and "RAG Template" for optimal results

Configuring Speech-to-Text​

Whisper-Large-V3-Turbo can be configured in Open WebUI for speech-to-text (STT) functionality. This model supports over 99 languages and is optimized for audio transcription via our hosted API.

Admin Panel Configuration​

In the Admin Settings under "Audio", configure the following:

  • Speech-to-Text Engine: Select "OpenAI"
  • API Base URL: Enter https://llm.aihosting.mittwald.de/v1
  • API Key: Enter your API key
  • STT Model: Enter the model name whisper-large-v3-turbo

Hiding Whisper from Chat Models​

Whisper will appear in the model list after connection, but it should be hidden from chat model selection since it's designed for audio transcription, not conversational AI:

  1. Navigate to "Admin settings" > "Models"
  2. Select whisper-large-v3-turbo
  3. Choose "Hide model" to prevent it from appearing as a chat option

User Settings​

You can further specify how Open WebUI interacts with the Whisper model in the user settings (not Administrator panel) under "Audio":

  • Language: Explicitly set the language code (e.g., "de" for German, "en" for English)
  • Instant Auto-Send After Voice Transcription: Enable to send transcriptions directly without confirmation

Recommended Parameters​

For optimal transcription quality, configure these parameters in the admin panel or chat settings:

  • Additional Parameters: Set temperature=1.0, top_p=1.0

Testing Speech-to-Text​

To test the speech-to-text functionality:

  1. Click the microphone icon in a chat interface
  2. Speak in the configured language
  3. The transcription will use our /v1/audio/transcriptions endpoint with support for MP3, OGG, WAV, and FLAC formats (maximum 25 MB file size)

You can now use Whisper in any chat! Chat with your favorite LLM by dictating your question and sending it.