"""Voxi Lite server checks (in-process, no network): every endpoint, format and error path, and the WebSocket (stream, cancel, error, two requests on one socket). python tests/test_server.py """ import io import json import sys from pathlib import Path import soundfile as sf from fastapi.testclient import TestClient sys.path.insert(0, str(Path(__file__).resolve().parents[1])) import server # noqa: E402 LINE = "नमस्ते, आपका ऑर्डर कल शाम तक पहुँच जाएगा। क्या मैं और कोई मदद कर सकती हूँ?" LONG = " ".join(["आपका ऑर्डर कल शाम तक पहुँच जाएगा, और ट्रैकिंग लिंक एसएमएस पर भेज दिया गया है।"] * 6) results = [] def check(name, ok, detail=""): results.append(ok) print(("PASS " if ok else "FAIL ") + name + (f" ({detail})" if detail else ""), flush=True) def main(): with TestClient(server.app) as c: r = c.get("/health"); j = r.json() check("health", r.status_code == 200 and j["status"] == "ok" and j["voices"] >= 4, json.dumps(j)) v = c.get("/v1/voices").json()["voices"] check("voices", all(x in v for x in ("payal", "arjun", "sanjana", "vishnu", "payal@happy", "arjun@sad-soft", "harper", "ethan", "amelia", "oliver")), f"{len(v)} names") r = c.post("/v1/tts", json={"text": LINE, "lang": "hi", "voice": "payal"}) y, sr = sf.read(io.BytesIO(r.content)) check("tts wav 24k", r.status_code == 200 and r.headers["content-type"].startswith("audio/wav") and sr == 24000 and len(y) > sr, f"{len(y) / sr:.2f} s") dur = len(y) / sr r = c.post("/v1/tts", json={"text": LINE, "voice": "arjun", "format": "pcm16", "sample_rate": 16000}) check("tts pcm16 16k", r.status_code == 200 and "rate=16000" in r.headers["content-type"] and len(r.content) % 2 == 0 and len(r.content) / 2 / 16000 > 1, f"{len(r.content) / 32000:.2f} s") r = c.post("/v1/tts", json={"text": LINE, "voice": "payal@happy", "format": "mulaw", "sample_rate": 8000, "stream": True}) check("tts mulaw 8k streamed", r.status_code == 200 and r.headers["content-type"].startswith("audio/basic") and len(r.content) / 8000 > 1, f"{len(r.content) / 8000:.2f} s") r = c.post("/v1/tts", json={"text": "Your refund has been processed.", "lang": "en", "voice": "sanjana", "format": "wav", "sample_rate": 8000}) y8, sr8 = sf.read(io.BytesIO(r.content)); check("tts wav 8k", r.status_code == 200 and sr8 == 8000 and len(y8) > 4000) r = c.post("/v1/tts", json={"text": "Aapka bill kal tak pay kar dijiye.", "lang": "hinglish", "voice": "vishnu"}) check("tts hinglish", r.status_code == 200) r = c.post("/v1/tts", json={"text": "Unga order naalaikku varum, link SMS-la anuppirukom.", "lang": "tanglish", "voice": "payal"}) check("tts tanglish", r.status_code == 200 and len(sf.read(io.BytesIO(r.content))[0]) > 24000) for voice in ("harper", "amelia"): r = c.post("/v1/tts", json={"text": "Your refund has been processed.", "lang": "en", "voice": voice, "format": "mulaw", "sample_rate": 8000, "stream": True}) check(f"tts english {voice} mulaw streamed", r.status_code == 200 and len(r.content) / 8000 > 1, f"{len(r.content) / 8000:.2f} s") r = c.post("/v1/tts", json={"text": "नमस्ते", "lang": "hi", "voice": "oliver"}) check("400 english voice with hi", r.status_code == 400 and "English only" in r.json().get("error", ""), r.text[:80]) r = c.post("/v1/tts", json={"text": "", "lang": "hi"}) check("tts empty text -> empty audio", r.status_code == 200 and len(sf.read(io.BytesIO(r.content))[0]) == 0) for name, body, frag in [("bad voice", {"text": "hi", "voice": "nobody"}, "unknown voice"), ("bad lang", {"text": "hi", "lang": "xx"}, "unsupported lang"), ("bad format", {"text": "hi", "format": "mp3"}, "unsupported format"), ("bad rate", {"text": "hi", "sample_rate": 44100}, "sample_rate"), ("text too long", {"text": "क" * 2001}, "longer than"), ("wav stream", {"text": "hi", "stream": True}, "stream=true")]: r = c.post("/v1/tts", json=body) check(f"400 {name}", r.status_code == 400 and frag in r.json().get("error", ""), r.text[:80]) r = c.post("/v1/audio/speech", json={"model": "voxi-lite", "input": LINE, "voice": "payal", "response_format": "wav"}) check("openai speech wav", r.status_code == 200 and sf.read(io.BytesIO(r.content))[1] == 24000) r = c.post("/v1/audio/speech", json={"input": "Hello there, how can I help?", "voice": "arjun", "response_format": "pcm", "language": "en"}) check("openai speech pcm", r.status_code == 200 and len(r.content) / 48000 > 0.5) r = c.post("/v1/audio/speech", json={"input": "hi", "response_format": "mp3"}) check("openai speech 400", r.status_code == 400) with c.websocket_connect("/v1/tts/ws") as ws: ws.send_text(json.dumps({"text": LINE, "lang": "hi", "voice": "payal", "format": "mulaw", "sample_rate": 8000, "request_id": "a"})) n, frames = 0, 0 while True: m = ws.receive() if m.get("bytes") is not None: n += len(m["bytes"]); frames += 1 else: done = json.loads(m["text"]); break check("ws stream", done["event"] == "done" and done["request_id"] == "a" and done["status"] == "ok" and frames >= 2 and abs(n / 8000 - done["audio_s"]) < 0.1, f"{frames} frames, {n / 8000:.2f} s, first {done['first_audio_ms']} ms") ws.send_text(json.dumps({"text": LONG, "lang": "hi", "voice": "arjun", "format": "pcm16", "sample_rate": 16000, "request_id": "b"})) first = ws.receive() ws.send_text(json.dumps({"event": "cancel"})) while True: m = ws.receive() if m.get("text") is not None: done = json.loads(m["text"]); break check("ws cancel", first.get("bytes") is not None and done["request_id"] == "b" and done["status"] == "cancelled" and done["audio_s"] < 30, json.dumps(done)) ws.send_text(json.dumps({"text": "hi", "voice": "nobody", "request_id": "c"})) e = json.loads(ws.receive_text()) check("ws error", e["event"] == "error" and e["request_id"] == "c" and "unknown voice" in e["error"]) ws.send_text(json.dumps({"text": "जी, बिल्कुल।", "voice": "sanjana", "request_id": "d"})) while True: m = ws.receive() if m.get("text") is not None: done = json.loads(m["text"]); break check("ws second request", done["request_id"] == "d" and done["status"] == "ok") print(f"SERVER_TESTS {sum(results)}/{len(results)} passed", flush=True) sys.exit(0 if all(results) else 1) if __name__ == "__main__": main()