Implement DashScope ASR provider and enhance ASR service architecture

- Added DashScope ASR service implementation for real-time streaming.
- Updated ASR provider logic to support DashScope alongside existing providers.
- Enhanced runtime metadata resolution to include DashScope as a valid ASR provider.
- Modified configuration files and documentation to reflect the addition of DashScope.
- Introduced tests to validate DashScope integration and ASR service behavior.
- Refactored ASR service factory to accommodate new provider options and modes.
This commit is contained in:
Xin Wang
2026-03-06 11:44:39 +08:00
parent 7e0b777923
commit e11c3abb9e
19 changed files with 940 additions and 44 deletions

View File

@@ -0,0 +1,67 @@
import asyncio
import pytest
from providers.asr.dashscope import DashScopeRealtimeASRService
@pytest.mark.asyncio
async def test_dashscope_asr_interim_event_emits_interim_transcript():
received = []
async def _on_transcript(text: str, is_final: bool) -> None:
received.append((text, is_final))
service = DashScopeRealtimeASRService(api_key="test-key", on_transcript=_on_transcript)
service._loop = asyncio.get_running_loop()
service._running = True
service._on_ws_event(
{
"type": "conversation.item.input_audio_transcription.text",
"stash": "你好世界",
}
)
await asyncio.sleep(0.05)
result = service._transcript_queue.get_nowait()
assert result.text == "你好世界"
assert result.is_final is False
assert received == [("你好世界", False)]
@pytest.mark.asyncio
async def test_dashscope_asr_final_event_emits_final_transcript_and_final_queue():
received = []
async def _on_transcript(text: str, is_final: bool) -> None:
received.append((text, is_final))
service = DashScopeRealtimeASRService(api_key="test-key", on_transcript=_on_transcript)
service._loop = asyncio.get_running_loop()
service._running = True
service._audio_sent_in_utterance = True
service._on_ws_event(
{
"type": "conversation.item.input_audio_transcription.completed",
"transcript": "最终识别结果",
}
)
await asyncio.sleep(0.05)
result = service._transcript_queue.get_nowait()
assert result.text == "最终识别结果"
assert result.is_final is True
assert service._final_queue.get_nowait() == "最终识别结果"
assert received == [("最终识别结果", True)]
@pytest.mark.asyncio
async def test_dashscope_wait_for_final_falls_back_to_latest_interim_on_timeout():
service = DashScopeRealtimeASRService(api_key="test-key")
service._audio_sent_in_utterance = True
service._last_interim_text = "部分结果"
text = await service.wait_for_final_transcription(timeout_ms=10)
assert text == "部分结果"