ovos-stt-plugin-server 0.1.5a6

Description

OpenVoiceOS companion plugin for OpenVoiceOS STT Server

Install

pip install ovos-stt-plugin-server

Configuration

  "stt": {
    "module": "ovos-stt-plugin-server",
    "ovos-stt-plugin-server": {
      "urls": ["https://0.0.0.0:8080/stt"],
      "verify_ssl": true
    },
 }

for audio language detection

  "listener": {
    "audio_transformers": {
        "ovos-audio-lang-server-plugin": {
          "urls": ["https://0.0.0.0:8080/lang_detect"],
          "verify_ssl": true
        }
    }
  }

Security warning

Please note that while you can set verify_ssl to false to disable SSL verification, this is not recommended and should only be used for testing purposes. Consider using a private CA or certificates signed using Let's Encrypt instead.

Self-hosting (recommended)

Run your own server. It keeps your audio on your own hardware, it does not depend on somebody else's uptime, and you choose the model.

pip install ovos-stt-http-server ovos-stt-plugin-onnx-asr[cpu]
ovos-stt-server --engine ovos-stt-plugin-onnx-asr

ovos-stt-plugin-onnx-asr is the recommended engine: it runs ONNX models on CPU, ships a best-model-per-language registry covering ~90 languages, and loads a model per request language, so one server can serve all of them.

The [cpu] extra installs the CPU ONNX runtime. If the box has a supported GPU, install [gpu] instead of [cpu], not in addition: pip does not remove the CPU runtime when you add the GPU one, and a venv with both installed is not a configuration we test.

Point the plugin at it:

  "stt": {
    "module": "ovos-stt-plugin-server",
    "ovos-stt-plugin-server": {
      "urls": ["https://your-server.example/stt"]
    }
  }

The server runs one engine at a time, and you choose the engine with --engine. This example serves Italian with the Vosk engine:

pip install ovos-stt-http-server ovos-stt-plugin-vosk
ovos-stt-server --engine ovos-stt-plugin-vosk

  "stt": {
    "module": "ovos-stt-plugin-server",
    "ovos-stt-plugin-server": {
      "urls": ["http://0.0.0.0:8080/stt"]
    },
    "ovos-stt-plugin-vosk": {
      "lang": "it-IT",
      "model": "https://alphacephei.com/vosk/models/vosk-model-small-it-0.22.zip"
    }
  }

In this example the server and the listener run on the same machine, so one mycroft.conf carries the whole block: the server reads the ovos-stt-plugin-vosk section, and the listener reads the ovos-stt-plugin-server section. If you move the server to another machine, that machine keeps the ovos-stt-plugin-vosk section, and the listener keeps only the ovos-stt-plugin-server section.

When the server is down, the listener can use a local STT plugin as a fallback. Set fallback_module and give that plugin its own config:

  "stt": {
    "module": "ovos-stt-plugin-server",
    "fallback_module": "ovos-stt-plugin-vosk",
    "ovos-stt-plugin-server": {
      "urls": ["https://your-server.example/stt"]
    },
    "ovos-stt-plugin-vosk": {
      "model": "https://alphacephei.com/vosk/models/vosk-model-small-it-0.22.zip"
    }
  }

Public servers

If you set no urls, the plugin falls back to public servers.

These are a community courtesy, not a service. They exist so you can try OVOS without setting anything up first. They are provided on a best-effort basis with no guarantees of uptime, latency, accuracy, privacy, or continued existence, and they can change or disappear without notice. They are meant for demos, evaluation and onboarding — not for production, and not for anything you would not want a third party to receive. Audio you send is processed on hardware you do not control.

For anything beyond trying it out, self-host.