Here I explain to you how I created my own voice assistant on my home server. My goal is to keep my data my sat. For me it was also important to have the issue on my Sonos boxes. Involving artificial intelligence seemed too demanding for me in the beginning and therefore here in Part 1 only the choice of the system.
The right software: Snips.ai is the first result of a Google search but was unfortunately discontinued this January (2020). What was out of the question for me was Dialogflow (formerly api.ai) from Google, which offers minimal processing and processes everything externally. It quickly became difficult to find a platform that does not process speech analysis externally.
The data processing concept of voice assistants is: 1. Convert language to text (STT) 2. process this text and 3. output the result as a language (TTS).
According to Mycroft, which has a very flexible framework, for me mainly Home-assistant.io with Hass.io was an option. With its many interfaces such as Sonos or the German Weather Service, but also Ifttt around only a few of the approx. 1500 application. Home Assistant lets you control a lot and also offers a good graphical interface.
For the voice assistant, Rhasspy stands out for complete offline availability as STT and MaryTTS or Pico TTS. More on this in Part 2.
In the next article (Part 2) the installation of Homeassistant.