feat(voice): add elevenlabs realtime streaming asr

tests/test_ai_voice_asr_provider.py

@@ -1,5 +1,7 @@
 from __future__ import annotations
 
+import json
+import queue
 import time
 
 import httpx
@@ -301,6 +303,102 @@ def test_yandex_buffered_streaming_provider_buffers_pcm_until_finalize():
     assert calls == [{"audio_bytes": b"\x10\x00\x20\x00", "language_hint": "ru"}]
 
 
+def test_elevenlabs_realtime_streaming_provider_returns_partial_and_final():
+    calls: list[dict] = []
+
+    class _FakeRealtimeWebSocket:
+        def __init__(self) -> None:
+            self.sent_payloads: list[dict] = []
+            self.incoming: queue.Queue[str | None] = queue.Queue()
+            self.incoming.put(
+                json.dumps(
+                    {
+                        "message_type": "session_started",
+                        "session_id": "session-test",
+                        "config": {"sample_rate": 16000, "audio_format": "pcm_16000"},
+                    }
+                )
+            )
+
+        def send(self, raw_payload: str) -> None:
+            payload = json.loads(raw_payload)
+            self.sent_payloads.append(payload)
+            if payload.get("commit"):
+                self.incoming.put(
+                    json.dumps(
+                        {
+                            "message_type": "committed_transcript",
+                            "text": "нужно узнать график работы",
+                            "language_code": "ru",
+                        }
+                    )
+                )
+            else:
+                self.incoming.put(
+                    json.dumps(
+                        {
+                            "message_type": "partial_transcript",
+                            "text": "нужно узнать график",
+                            "language_code": "ru",
+                        }
+                    )
+                )
+
+        def recv(self) -> str:
+            item = self.incoming.get(timeout=1)
+            if item is None:
+                raise RuntimeError("closed")
+            return item
+
+        def close(self) -> None:
+            self.incoming.put(None)
+
+    websocket = _FakeRealtimeWebSocket()
+
+    def _factory(url: str, *, header: list[str], timeout: float):
+        calls.append({"url": url, "header": header, "timeout": timeout})
+        return websocket
+
+    provider = asr_module.ElevenLabsRealtimeStreamingASRProvider(
+        api_base="https://api.elevenlabs.example",
+        api_key="asr-key",
+        timeout_seconds=3,
+        finalize_timeout_seconds=1,
+        websocket_factory=_factory,
+    )
+    stream_id = provider.open_stream("session-1", language_hint="ru")
+    provider.push_pcm(stream_id, b"\x01\x00" * 160)
+
+    partial = None
+    for _ in range(20):
+        partial = provider.poll_partial(stream_id)
+        if partial is not None:
+            break
+        time.sleep(0.02)
+
+    assert partial is not None
+    assert partial.text == "нужно узнать график"
+    assert partial.is_final is False
+
+    final = provider.finalize(stream_id)
+    provider.close_stream(stream_id)
+
+    assert final.text == "нужно узнать график работы"
+    assert final.language == "ru"
+    assert calls[0]["url"].startswith("wss://api.elevenlabs.example/v1/speech-to-text/realtime?")
+    assert "model_id=scribe_v2_realtime" in calls[0]["url"]
+    assert "audio_format=pcm_16000" in calls[0]["url"]
+    assert "language_code=ru" in calls[0]["url"]
+    assert calls[0]["header"] == ["xi-api-key: asr-key"]
+    assert calls[0]["timeout"] == 3.0
+    assert len(websocket.sent_payloads) == 2
+    assert websocket.sent_payloads[0]["message_type"] == "input_audio_chunk"
+    assert websocket.sent_payloads[0]["commit"] is False
+    assert websocket.sent_payloads[0]["sample_rate"] == 16000
+    assert websocket.sent_payloads[0]["audio_base_64"]
+    assert websocket.sent_payloads[1]["commit"] is True
+
+
 def test_yandex_grpc_streaming_provider_returns_partial_and_final():
     calls: list[dict] = []
 
@@ -424,6 +522,10 @@ def test_yandex_asr_builders():
     assert isinstance(asr_module.build_asr_provider("elevenlabs"), asr_module.ElevenLabsASRProvider)
     assert isinstance(asr_module.build_asr_provider("yandex"), asr_module.YandexSpeechKitASRProvider)
     assert isinstance(asr_module.build_asr_provider("speechkit"), asr_module.YandexSpeechKitASRProvider)
+    assert isinstance(
+        asr_module.build_streaming_asr_provider("elevenlabs_realtime"),
+        asr_module.ElevenLabsRealtimeStreamingASRProvider,
+    )
     assert isinstance(
         asr_module.build_streaming_asr_provider("yandex_speechkit"),
         asr_module.YandexSpeechKitGrpcStreamingASRProvider,