Skip to content

Engines

Architecture
Engines

flowchart TB
  subgraph TTS["TTS engines"]
    Piper[Piper]
    ElevenLabs[ElevenLabs]
    GoogleTTS[Google]
    Rime[Rime]
  end

  subgraph STT["STT engines"]
    Vosk[Vosk]
    Deepgram[Deepgram]
    AssemblyAI[AssemblyAI]
    GoogleSTT[Google]
  end

  subgraph LLM["LLM engines"]
    OpenAI[OpenAI]
    Ollama[Ollama]
    Cerebras[Cerebras]
    Grok[Grok]
  end

  subgraph Services["Services (consumers)"]
    SpeechSyn[Speech Synthesis]
    SpeechRec[Speech Recognition]
    Assistant[Assistant]
  end

  TTS --> SpeechSyn
  TTS --> Assistant
  STT --> SpeechRec
  STT --> Assistant
  LLM --> Assistant
  SpeechSyn --> Assistant
  SpeechRec --> Assistant

Engines are pluggable backends for speech and language: text-to-speech (TTS), speech-to-text (STT), and language models (LLMs). They live under ubo_app/engines/ and are used by the assistant, speech synthesis, and speech recognition services.

What you see

  • Abstraction — engines/abstraction/:
  • engine.py — EngineMixin base with name, label, instance_label.
  • ai_provider_mixin.py, background_running_mixin.py, needs_setup_mixin.py, remote_mixin.py — Mixins for API-backed or on-device engines.
  • TTS — e.g. Piper (piper.py), ElevenLabs (elevenlabs.py), Google (google.py), Google Cloud (google_cloud.py), Rime (rime.py), Picovoice Orca (used via accessibility/speech synthesis).
  • STT — e.g. Vosk (vosk.py), Deepgram (deepgram.py), AssemblyAI (assemblyai.py), Google (speech recognition).
  • LLM — e.g. OpenAI (openai.py), Ollama (ollama.py, ollama_onprem.py), Cerebras (cerebras.py), Grok (grok.py), Google (for assistant).
  • Registry — The assistant and speech services use registries (e.g. engines_registry.py in the assistant service) to resolve which engine to use from settings/state.