JaviVoiceAssistant

OpenSource VoiceAssitant Javi

https://javi.devellight.studio/

A multifunction and fully offline usable open-source voice assistant written in Java!

⚠️ NOTICE: The LICENSE file only applies to the source-code (java files) in this repo, not to any dependencies! ⚠️
License Infomation of used libraries can be found here: https://theredstonedev-de.github.io/JaviVoiceAssistant/target/site/
Please inform me about bugs you have found via the issues!
⚠️ This is still Work-In-Progress Software! ⚠️

Used libraries in this Project:

MaryTTS (Text-To-Speech)
Vosk (speech recognizer)
QT / QT Jambi (just the simple Text-Overlay)
java-discord-rpc (Discord RPC)
dbus java for communicating with MPRIS

Building and Running

Requirements to run fully:

a PC based on x86_64 (amd64) with a Linux installation
Java (version 18 recommended)
QT 6
Maven (mvn on command line)
Git (optional, only for downloading)

Requirements to run (only basic funtionality):

any x86_64 PC OS doesn't matter
Java
Git (optional, only for downloading)

meida control, volume control and some other features WILL NOT WORK on Windows!

</div>

Running standalone:

Clone the repository:

git clone https://github.com/TheRedstoneDEV-DE/VoiceAssistant

and get the voice model for vosk from (the small one is recommended) VOSK Models unzip it, rename the folder inside it to 'model' and put it into the project root
Build

mvn install
Run the Jar file, which was generated in the 'target' folder after copieng it into the project root
⚠️ Warning: if you are using pipewire it may not work you can find a fix in pipewire-fix.md ⚠️
java -jar [JARFILE]
Cofigure via setup window

The fields in the window should be selfexplaining.

'Overlay position' definines the coordinates for the overlay (should only be necessary, if you have multiple screens)
'CPU temp file' defines the file, in which the cpu temperature is logged by the kernel (if you don't have the overlay enabled, this is not used at all)
any MPRIS settings are just defining, which player you are targeting in the dbus interface (you can start, pause and skip titles with voicecommands)#

Running with recognition server:

Clone the repository:

git clone https://github.com/TheRedstoneDEV-DE/VoiceAssistant

and get the voice model for vosk from (the small one is recommended)

VOSK Models unzip it, rename the folder inside it to 'model' and put it into the project root on the server.
Build

mvn install
Copy the Jar file from the 'target' folder in the project root and run the client standalone for setup.
⚠️ Warning: if you are using pipewire it may not work you can find a fix in pipewire-fix.md ⚠️
```
`java -jar [JARFILE]`
```
Copy the built jar from the target folder, the 'model' folder and the 'keys' folder to the server device (here it really could be Windows and should'n affect anything)
Generate the Keystore and Truststore for Client and Server on the Server. To do that, go into the 'keys' folder and execute the bash file. Next you will be asked some questions, because of the certificate and a password for the key (you should remember it, because you need it when running)

bash genkeys.sh
Move the 'client-trust.jks' from the 'keys' folder of the server to the 'keys' folder of the client
Run! First run the server.

KEY_PASSPHRASE=[Password of the Keystore] java -jar [JARFILE] --server

Then the client (you could even use multiple with one server).

KEY_PASSPHRASE=[password of the Truststore] java -jar [JARFILE] --client [server IP address]

List of commands

command	short explanation
hey javi	just the activation command, has to be said, before any other voice command
set volume to [number between 10-90] percent	sets the volume of the main audio output via amixer
open [application name]	opens the named application in another "screen"
reconfigure	opens the setup ui again (in case you missconfigured something)
reconfigure programs	opens the setup ui for adding programs to the open [application name] command
[hide/show] overlay	hides or shows the overlay
pause/resume/next/previous	plays/pauses a mediaplayer, skippes one title forward/backward
start/stop artificial intelligence	enters/exits the ai-mode if LLMs are enabled

List of Environment variables

Env variable	explanation
`OUTPUT_GAIN`	adjusts the output volume of the tts system (in case you are using it with some audio server like alsa)
`AUDIODEV`	sets the audio device for input and ouput, but only if it is supported, if it is not supported the voice assistant won't have any audio

Run with Mimic 3

Just start the VoiceAssistant with the following arguments:

--mimic3 [HOST]:[PORT]

⚠️ NOTE: you need to give both: HOST and PORT ⚠️

Enabling the LLM integration (activateable by saying: “start artificial intelligence”)

Setup the VoiceAssistant, if you havent already and go through initial setup

Append this to the config/VoiceAssistant.properties config file and edit it to your liking:

llm-enabled=true
llm-host=localhost:4891
llm-model=GPT4All Falkon
llm-api-key=NOT NEEDED
llm-max-tokens=150

Start the VoiceAssistant and try it out!

This site is open source. Improve this page.