Skip to content
michael edited this page Nov 3, 2026 · 6 revisions

HermiVoice

HermiVoice Hauptbildschirm mit Chat

Willkommen bei HermiVoice – deiner persönlichen Sprachsteuerung für den Hermes-Agent. HermiVoice ist eine Android-App in nativem Kotlin, die du direkt als APK installierst (Sideload, ohne Play Store). Sie läuft auf Android 7.0 oder neuer (minSdk 24). Du kommunizierst mit Hermes per Sprache (Push-to-Talk) – oder ganz klassisch per Tastatur.

HermiVoice is your personal voice control for the Hermes agent. It is a native Kotlin Android app that you install directly as an APK (sideload, no Play Store). It runs on Android 7.0 or newer (minSdk 24). You talk to Hermes by voice (push-to-talk) – or simply by keyboard.

English Version below.

Seitenübersicht

Page Index

So funktioniert es:

  1. Sprechtaste gedrückt halten und sprechen – oder tippe deine Nachricht ins Eingabefeld.
  2. Die Spracherkennung läuft vollständig offline auf deinem Gerät.
  3. Deine Nachricht geht an die Hermes-API (OpenAI-kompatible Chat-Completions).
  4. Die Antwort erscheint live in der App, während sie generiert wird (Streaming).

Highlights:

  • Push-to-Talk mit Live-Zustandsfeedback: Bereit → Hört → Erkennt → Sendet → Gesendet/Fehler.
  • Tastatureingabe: Nachrichten tippen statt sprechen – jederzeit umschaltbar (Startmodus wählbar).
  • Offline-Spracherkennung mit deutschen und englischen Modellen (~57 MB einmaliger Download je Modell).
  • Mehrere Chat-Sessions, Favoriten, durchsuchbare Historie, Markdown-Export.
  • Mehrere Hermes-Bots und Cloud-KI-Anbieter (OpenRouter, OpenAI, Groq, Mistral, DeepSeek) mit eigenem API-Key, zweisprachige Oberfläche (DE/EN), verschlüsselte API-Keys.

In wenigen Minuten verbunden: Siehe „Einrichtung: App mit dem Hermes-Agent verbinden“.

English

Here is how it works:

  1. Press and hold the talk button and speak – or type your message into the input field.
  2. Speech recognition runs fully offline on your device.
  3. Your message is sent to the Hermes API (OpenAI-compatible chat completions).
  4. The answer appears live in the app while it is generated (streaming).

Highlights:

  • Push-to-talk with live status feedback: Ready → Listening → Recognizing → Sending → Sent/Error.
  • Keyboard input: Type your messages instead of speaking – switchable at any time (start mode selectable).
  • Offline speech recognition with German and English models (~57 MB one-time download per model).
  • Multiple chat sessions, favorites, searchable history, Markdown export.
  • Multiple Hermes bots and cloud AI providers (OpenRouter, OpenAI, Groq, Mistral, DeepSeek) with their own API key, bilingual interface (DE/EN), encrypted API keys.

Connected in a few minutes: See „Setup: Connecting the App to the Hermes Agent“.

Clone this wiki locally