SonusAI ASR Faster Whisper

Faster-Whisper ASR tools for SonusAI.

This package provides a SonusAI-compatible Automatic Speech Recognition (ASR) plugin backed by faster-whisper.

Overview

• Integrates Faster-Whisper transcription into SonusAI workflows.
• Exposes ASR functions under sonusai_asr_faster_whisper/asr_functions/.
• Returns SonusAI-native ASRResult objects for downstream processing.

Requirements

• Python >=3.13, <3.15
• Dependencies are managed via uv and defined in pyproject.toml

Installation

Install dependencies for development:

uv sync

Usage

Core exported functions:

• faster_whisper_validate(**config)
• faster_whisper(audio, **config)

Minimum config requirement:

• model: Faster-Whisper model name/path

Common optional config keys:

• device (default: "cpu")
• cpu_threads (default: 1)
• compute_type (default: "int8")
• beam_size (default: 5)

Development

Quality and validation commands:

ruff check .
ruff format .
pytest

Additional project context and standards are documented in:

• CLAUDE.md
• docs/internal/overview.md
• docs/internal/architecture.md
• docs/internal/cli_commands.md
• docs/internal/standards.md