README.en.md

Projekt OpenWebUI_De

The goal is a simple installation and starting the OpenWebUi in German.

A script for the installation and subsequent registration and settings for German speech output

Documentation

The documentation is still in progress. Documentation

Dependencies

• Linux (tested with Ubuntu 24 LTS Server)

• The RAM should be at least 16 GB Ram for the standard installation of llama3.1 (use via CPU, adaptation for AMD GPU or NVIDIA GPC also possible, see Ollama)

• at least 40 GB storage space hard drive/SSD

• root rights

Installation

Install everything with one command :-)

The password for the root is only required once

  wget https://raw.githubusercontent.com/baefthde/OpenWebUI_De/refs/heads/main/install.sh && chmod +x install.sh && ./install.sh

Execute download and set rights and execute install.sh

The install.sh:

• installed git
• get the current OpenWebUI_De directory via git
• sets the rights and runs OpenWebUI_De.sh

Manually via git

  git clone https://github.com/baefthde/OpenWebUI_De
  cd /OpenWebUI_De
  chmod +x OpenWebUI_De.sh
  ./OpenWebUI_De.sh

OpenWebUI_De.sh

installed:

• Docker via Snap
• ollama as API interface (only internally on 127.0.0.1:11434)
  • llama3.1 as LLM (Large Language Model) language model from Meta AI
• OpenWebUI in Docker (accessible externally via port: 8080)
  • OpenWenUI database at ~/open-webui/webui.db accessible for backup and restore outside of Docker
• watchtower in Docker so that OpenWebUI updates itself automatically every hour
• openedai-speech API interface (externally accessible port: 8000)
  • with German voice output via the voice of Thorsten-Voice

Settings after installation

Open OpenWebUI at http://IP:8080

First registration gets admin rights for configuration (name, email, password), no email will be sent!

• Settings -> General:
  • Set Language to "German Deutsch".
• Settings -> Audio -> TTS Settings
  • Autoplay reply on
• Admin Panel -> Settings -> Connections
  • OpenAI API should be disabled
  • Ollama API is set to http://localhost:11434 via Docker start
• Admin Panel -> Settings -> Audio
  • TTS settings:
    • Text-to-speech-engine: http://IP:8000/v1 OpenAI: sk-111111111
    • TTS voice: thorsten TTS-model: tts-1

The AI ​​chat with voice output can then be tested.

A notice

Other LLMs

Additional LLMs can also be installed both directly via ollama and via the OpenWebUI web interface:

The selection of available LLM models can be found here: https://ollama.com/library

e.g. ollama pull mistral

Other languages ​​or voices

1. The voice files must be of the type language.onnx and language.onnx.json and must be placed in the folder ~/openedai-speech/voices.
2. The configuration file ~/openedai-speech/config/voice_to_speaker.yaml needs to be adjusted, more detailed information can be found at: [https://github.com/matatonic/openedai-speech](https://github.com/matatonic/openedai -speech)

Overview of the projects used

Ollama

https://ollama.com/

https://github.com/ollama/ollama

Open Web UI

openwebui.com

https://github.com/open-webui/open-webui

openedai-speech-integration

openedai-speech-integration

https://github.com/matatonic/openedai-speech

thorsten-voice

https://www.thorsten-voice.de/

https://huggingface.co/rhasspy/piper-voices/tree/main/de/de_DE/thorsten/high

Testing the voices

https://rhasspy.github.io/piper-samples/

Authors

• @baefthde