tests/test_speech_response.py

import numpy as np
from fastapi.testclient import TestClient

from app.main import app
from app.services.engine_registry import EngineRegistry, get_engine_registry
from app.services.supertonic_engine import SynthesisResult


class FakeEngine:
    def supports_voice(self, _: str) -> bool:
        return True

    def list_voice_bindings(self) -> list[dict[str, str]]:
        return []

    def synthesize(self, **_: object) -> SynthesisResult:
        wav = np.zeros((1, 2400), dtype=np.float32)
        return SynthesisResult(waveform=wav, sample_rate=24000, duration_seconds=0.1)


def fake_registry() -> EngineRegistry:
    engine = FakeEngine()
    return EngineRegistry(
        model_to_engine={
            "supertonic-2": engine,
            "tts-1": engine,
            "tts-1-hd": engine,
            "gpt-4o-mini-tts": engine,
            "kitten-tts-micro-0.8": engine,
            "kitten-tts-micro": engine,
            "kitten-tts": engine,
            "kitten": engine,
            "melo-en": engine,
            "kokoro": engine,
            "kokoro-onnx": engine,
            "kokoro-82m": engine,
        }
    )


def test_speech_returns_wav(monkeypatch) -> None:
    app.dependency_overrides[get_engine_registry] = fake_registry
    try:
        with TestClient(app) as client:
            response = client.post(
                "/v1/audio/speech",
                headers={"Authorization": "Bearer changeme"},
                json={
                    "model": "supertonic-2",
                    "input": "hello",
                    "voice": "alloy",
                    "response_format": "wav",
                },
            )
        assert response.status_code == 200
        assert response.headers["content-type"].startswith("audio/wav")
        assert len(response.content) > 44
    finally:
        app.dependency_overrides.clear()


def test_speech_returns_mp3(monkeypatch) -> None:
    app.dependency_overrides[get_engine_registry] = fake_registry
    try:
        with TestClient(app) as client:
            response = client.post(
                "/v1/audio/speech",
                headers={"Authorization": "Bearer changeme"},
                json={
                    "model": "supertonic-2",
                    "input": "hello",
                    "voice": "alloy",
                    "response_format": "mp3",
                },
            )
        assert response.status_code == 200
        assert response.headers["content-type"].startswith("audio/mpeg")
        assert response.content[:3] == b"ID3" or response.content[:2] == b"\xff\xfb"
    finally:
        app.dependency_overrides.clear()


def test_speech_accepts_quality_and_friendly_voice(monkeypatch) -> None:
    app.dependency_overrides[get_engine_registry] = fake_registry
    try:
        with TestClient(app) as client:
            response = client.post(
                "/v1/audio/speech",
                headers={"Authorization": "Bearer changeme"},
                json={
                    "model": "supertonic-2",
                    "input": "hello",
                    "voice": "atlas",
                    "quality": "high",
                    "response_format": "wav",
                },
            )
        assert response.status_code == 200
        assert response.headers["content-type"].startswith("audio/wav")
    finally:
        app.dependency_overrides.clear()


def test_kokoro_model_route(monkeypatch) -> None:
    app.dependency_overrides[get_engine_registry] = fake_registry
    try:
        with TestClient(app) as client:
            response = client.post(
                "/v1/audio/speech",
                headers={"Authorization": "Bearer changeme"},
                json={
                    "model": "kokoro",
                    "input": "hello",
                    "voice": "af_bella",
                    "response_format": "wav",
                },
            )
        assert response.status_code == 200
        assert response.headers["content-type"].startswith("audio/wav")
    finally:
        app.dependency_overrides.clear()


def test_kitten_model_route(monkeypatch) -> None:
    app.dependency_overrides[get_engine_registry] = fake_registry
    try:
        with TestClient(app) as client:
            response = client.post(
                "/v1/audio/speech",
                headers={"Authorization": "Bearer changeme"},
                json={
                    "model": "kitten-tts-micro-0.8",
                    "input": "hello",
                    "voice": "bella",
                    "response_format": "wav",
                },
            )
        assert response.status_code == 200
        assert response.headers["content-type"].startswith("audio/wav")
    finally:
        app.dependency_overrides.clear()


def test_melo_model_route(monkeypatch) -> None:
    app.dependency_overrides[get_engine_registry] = fake_registry
    try:
        with TestClient(app) as client:
            response = client.post(
                "/v1/audio/speech",
                headers={"Authorization": "Bearer changeme"},
                json={
                    "model": "melo-en",
                    "input": "hello",
                    "voice": "us",
                    "response_format": "wav",
                },
            )
        assert response.status_code == 200
        assert response.headers["content-type"].startswith("audio/wav")
    finally:
        app.dependency_overrides.clear()