feat: add audio test

tests/integrations/openai_tracker/test_audio.py

@@ -0,0 +1,119 @@
+import pytest
+
+from greptimeai import collector
+from . import sync_client
+from ..database.db import truncate_tables, get_trace_data_with_retry
+
+
+@pytest.fixture
+def _truncate_tables():
+    truncate_tables()
+    yield
+
+
+def test_audio_speech(_truncate_tables):
+    model = "tts-1"
+    text = "你好"
+    audio_file = "hello.mp3"
+    speed = 0.8
+    resp = sync_client.audio.speech.create(
+        input=text,
+        voice="alloy",
+        model=model,
+        speed=speed,
+    )
+    with open(audio_file, "wb") as f:
+        f.write(resp.content)
+
+    collector.otel._force_flush()
+
+    trace = get_trace_data_with_retry(user_id="", span_name="openai_speech", retry=3)
+
+    assert trace is not None
+
+    assert "greptimeai" == trace.get("resource_attributes", {}).get("service.name")
+    assert "openai" == trace.get("span_attributes", {}).get("source")
+
+    assert ["client.audio.speech.create", "end"] == [
+        event.get("name") for event in trace.get("span_events", [])
+    ]
+
+    events = trace.get("span_events")
+    assert events
+    attrs = events[0].get("attributes")
+    assert attrs
+    assert attrs["input"] == text
+    assert attrs["model"] == model
+    assert attrs["voice"] == "alloy"
+    assert attrs["speed"] == speed
+
+
+def test_audio_transcription(_truncate_tables):
+    model = "whisper-1"
+    audio_file = "hello.mp3"
+    language = "zh"
+
+    resp = sync_client.audio.transcriptions.create(
+        file=open(audio_file, "rb"),
+        model=model,
+        language=language,
+    )
+
+    assert "你好" in resp.text.lower()
+
+    collector.otel._force_flush()
+
+    trace = get_trace_data_with_retry(
+        user_id="", span_name="openai_transcription", retry=3
+    )
+
+    assert trace is not None
+
+    assert "greptimeai" == trace.get("resource_attributes", {}).get("service.name")
+    assert "openai" == trace.get("span_attributes", {}).get("source")
+
+    assert ["client.audio.transcriptions.create", "end"] == [
+        event.get("name") for event in trace.get("span_events", [])
+    ]
+    for event in trace.get("span_events", []):
+        if event.get("name") == "client.audio.transcriptions.create":
+            assert audio_file in event["attributes"]["file"]
+            assert event["attributes"]["model"] == model
+            assert event["attributes"]["language"] == language
+
+        elif event.get("name") == "end":
+            assert event["attributes"]["text"] == resp.text
+
+
+def test_audio_translation(_truncate_tables):
+    model = "whisper-1"
+    audio_file = "hello.mp3"
+
+    resp = sync_client.audio.translations.create(
+        file=open(audio_file, "rb"),
+        model=model,
+    )
+
+    assert "ni hao" or "你好" in resp.text.lower()
+
+    collector.otel._force_flush()
+
+    trace = get_trace_data_with_retry(
+        user_id="", span_name="openai_translation", retry=3
+    )
+
+    assert trace is not None
+
+    assert "greptimeai" == trace.get("resource_attributes", {}).get("service.name")
+    assert "openai" == trace.get("span_attributes", {}).get("source")
+
+    assert ["client.audio.translations.create", "end"] == [
+        event.get("name") for event in trace.get("span_events", [])
+    ]
+    for event in trace.get("span_events", []):
+        if event.get("name") == "client.audio.translations.create":
+            assert audio_file in event["attributes"]["file"]
+            assert event["attributes"]["model"] == model
+
+        elif event.get("name") == "end":
+            assert event["attributes"]["text"] == resp.text