From 228b8a46506c29d6daf69f4863b4b114a1ab937b Mon Sep 17 00:00:00 2001
From: admin <m@b.de>
Date: Mon, 8 Jun 2026 10:11:20 +0000
Subject: [PATCH] =?UTF-8?q?README.md=20hinzugef=C3=BCgt?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 README.md | 151 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 151 insertions(+)
 create mode 100644 README.md

diff --git a/README.md b/README.md
new file mode 100644
index 0000000..23e508a
--- /dev/null
+++ b/README.md
@@ -0,0 +1,151 @@
+# Installation
+## Docker Install
+1) `curl -fsSL https://get.Docker.com -o get-Docker.sh`
+2) `sudo sh get-Docker.sh`
+3) `sudo usermod -aG docker $USER`
+4) `newgrp docker`
+
+## Pfade
+1) Anlegen des Ordners: `mkdir -p ~/vosk-mqtt`
+2) Reinwechseln: `cd /home/pi/vosk-mqtt/`
+3) Dockerfile anlegen: `nano Dockerfile` und mit folgendem Inhalt füllen:
+```
+FROM python:3.11-slim
+
+ENV DEBIAN_FRONTEND=noninteractive
+
+# Audio-Bibliotheken für sounddevice/PortAudio + ALSA-Tools
+# UND libatomic1 für Vosk (fehlt im slim-Image)
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    portaudio19-dev \
+    alsa-utils \
+    libatomic1 \
+    && rm -rf /var/lib/apt/lists/*
+
+# Python-Pakete: Vosk, Sounddevice, Paho-MQTT
+RUN pip install --no-cache-dir \
+    vosk \
+    sounddevice \
+    paho-mqtt
+
+WORKDIR /app
+
+# Standard-Umgebungsvariablen
+ENV MQTT_HOST=mosquitto \
+    MQTT_PORT=1883 \
+    MQTT_TOPIC=klassengelaber \
+    VOSK_LANG=de \
+    DEVICE_ID=-1
+
+COPY stt_mqtt.py .
+
+CMD ["python", "stt_mqtt.py"]
+```
+4) Python-Skript anlegen: `nano stt_mqtt.py` und mit folgendem Inhalt füllen
+```
+#!/usr/bin/env python3
+import os
+import sys
+import json
+import queue
+
+import sounddevice as sd
+from vosk import Model, KaldiRecognizer
+import paho.mqtt.client as mqtt
+
+# --- Konfiguration aus Umgebungsvariablen ---
+MQTT_HOST = os.getenv("MQTT_HOST", "localhost")
+MQTT_PORT = int(os.getenv("MQTT_PORT", "1883"))
+MQTT_TOPIC = os.getenv("MQTT_TOPIC", "klassengelaber")
+VOSK_LANG = os.getenv("VOSK_LANG", "de")
+DEVICE_ID = int(os.getenv("DEVICE_ID", "-1"))
+
+# Ideale Sample-Rate für Vosk (Pi-schnell)
+VOSK_SAMPLERATE = 16000
+
+q = queue.Queue()
+
+
+def audio_callback(indata, frames, time, status):
+    """Callback wird von sounddevice für jeden Audio-Block aufgerufen."""
+    if status:
+        print(status, file=sys.stderr)
+    q.put(bytes(indata))
+
+
+def main():
+    # --- MQTT-Client aufsetzen ---
+    client = mqtt.Client()
+    client.connect(MQTT_HOST, MQTT_PORT, 60)
+    client.loop_start()
+
+    # --- Audio-Device bestimmen ---
+    if DEVICE_ID >= 0:
+        device = DEVICE_ID
+    else:
+        device = None  # default device
+
+    device_info = sd.query_devices(device, "input")
+    native_samplerate = int(device_info["default_samplerate"])
+    print(f"Native Sample-Rate des Mikrofon: {native_samplerate} Hz")
+
+    # Wir nutzen VOSK_SAMPLERATE (16000 Hz) für Vosk.
+    # sounddevice konvertiert automatisch, wenn wir samplerate=VOSK_SAMPLERATE setzen.
+    samplerate = VOSK_SAMPLERATE
+
+    print(f"Nutze Eingabegerät: {device_info['name']} @ {samplerate} Hz (für Vosk)")
+
+    # --- Vosk-Modell laden ---
+    print(f"Lade Vosk-Modell für Sprache: {VOSK_LANG}")
+    model = Model(lang=VOSK_LANG)
+
+    rec = KaldiRecognizer(model, samplerate)
+
+    # --- Audio-Stream öffnen mit angepassten Parametern ---
+    # blocksize=4000 -> 4000 Samples @ 16000 Hz = 250 ms
+    # latency="high" -> mehr Buffer, weniger Overflow
+    with sd.RawInputStream(
+        samplerate=samplerate,
+        blocksize=4000,
+        device=device,
+        dtype="int16",
+        channels=1,
+        latency="high",
+        callback=audio_callback,
+    ):
+        print("#" * 80)
+        print("Live-Transkription läuft. Strg+C zum Beenden.")
+        print("#" * 80)
+
+        try:
+            while True:
+                data = q.get()
+                if rec.AcceptWaveform(data):
+                    # Finales Ergebnis
+                    res = json.loads(rec.Result())
+                    text = res.get("text", "").strip()
+                    if text:
+                        print(f"FINAL: {text}")
+                        client.publish(MQTT_TOPIC, payload=text, qos=0, retain=False)
+                else:
+                    # Optional: Teil-Ergebnis
+                    pres = json.loads(rec.PartialResult())
+                    ptext = pres.get("partial", "").strip()
+                    if ptext:
+                        print(f"PARTIAL: {ptext}", end="\r")
+        except KeyboardInterrupt:
+            print("\nStoppe…")
+        finally:
+            client.loop_stop()
+
+
+if __name__ == "__main__":
+    main()
+```
+## Docker build
+Bauen des Container-Images imd Projektordner `docker build -t vosk-mqtt .`
+## Docker run
+Ausführen des Docker-Containers (ggfs. später per Systemdienst starten!)
+```
+docker run --rm   --name vosk-mqtt   --device /dev/snd:/dev/snd   --group-add audio   -e MQTT_HOST=192.168.213.12   -e MQTT_PORT=1883   -e MQTT_TOPIC=fi24b/klassengelaber   -e VOSK_LANG=de   -e DEVICE_ID=-1 vosk-mqtt:latest
+```
\ No newline at end of file