Glossary

This glossary defines technical and domain terms used in Voice2Machine.

General Terms

Local-First

Design philosophy where data (audio, text) is processed and stored exclusively on the user's device, without relying on the cloud.

Daemon

Background process (written in Python) that manages recording, transcription, and communication with the frontend.

REST API

Communication mechanism between the Daemon (Python) and clients (scripts, frontends). We use FastAPI with standard HTTP endpoints and WebSocket for real-time events.

Technical Components

Whisper

Speech recognition model (ASR) developed by OpenAI. Voice2Machine uses faster-whisper, an optimized implementation with CTranslate2.

Workflows

Specialized coordination components that manage the complete lifecycle of a specific task (e.g., RecordingWorkflow, LLMWorkflow). They replace the old monolithic "Orchestrator" for better traceability and maintainability.

Features

Self-contained modules that group domain logic and its infrastructure adapters (audio, llm, transcription). They represent the system's core capabilities.

BackendProvider

Frontend component (React Context) that manages connection with the Daemon and distributes state to the UI.

TelemetryContext

Sub-context in React optimized for high-frequency updates (GPU metrics, audio levels) to avoid unnecessary re-renders of the main UI.

Modular Architecture

Evolution of Hexagonal Architecture that organizes code around business modules (Features) and execution flows (Workflows), minimizing coupling and maximizing clarity.