digiops/call-center
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(voice): add streaming asr sidecar for v2 duplex

Browse Source
This commit is contained in:
Yera All
2026-04-12 01:38:30 +05:00
parent b75b76fae6
commit 9386b8f9b9
14 changed files with 1203 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
scripts/local_stack.py
+9
View File
@@ -88,6 +88,7 @@ def build_service_specs(env: dict[str, str] | None = None) -> list[dict[str, Any
{"name": "telegram", "port": 8007, "module": "services.telegram_adapter_service.app:app"},
{"name": "whatsapp", "port": 8019, "module": "services.whatsapp_adapter_service.app:app"},
{"name": "ai", "port": 8017, "module": "services.ai_orchestrator_service.app:app"},
{"name": "streaming-asr-sidecar", "port": 8021, "module": "services.streaming_asr_sidecar_service.app:app"},
{"name": "ai-voice-runtime", "port": 8018, "module": "services.ai_voice_runtime_service.app:app"},
{"name": "webchat", "port": 8011, "module": "services.webchat_adapter_service.app:app"},
{"name": "email", "port": 8012, "module": "services.email_adapter_service.app:app"},
@@ -317,6 +318,12 @@ def spawn_service(spec: dict[str, Any], runtime_dir: Path, data_dir: Path, base_
env["AI_VOICE_V2_DUPLEX_ENABLED"] = env.get("AI_VOICE_V2_DUPLEX_ENABLED", "1")
env["AI_VOICE_V2_STREAMING_ASR_BACKEND"] = env.get("AI_VOICE_V2_STREAMING_ASR_BACKEND", "local_sidecar")
env["AI_VOICE_V2_STREAMING_ASR_BASE_URL"] = env.get("AI_VOICE_V2_STREAMING_ASR_BASE_URL", "http://127.0.0.1:8021")
env["AI_VOICE_V2_STREAMING_ASR_TIMEOUT_SECONDS"] = env.get("AI_VOICE_V2_STREAMING_ASR_TIMEOUT_SECONDS", "4")
env["AI_VOICE_V2_STREAMING_ASR_MODEL"] = env.get("AI_VOICE_V2_STREAMING_ASR_MODEL", "base")
env["AI_VOICE_V2_STREAMING_ASR_COMPUTE_TYPE"] = env.get("AI_VOICE_V2_STREAMING_ASR_COMPUTE_TYPE", "int8")
env["AI_VOICE_V2_STREAMING_ASR_DEVICE"] = env.get("AI_VOICE_V2_STREAMING_ASR_DEVICE", "cpu")
env["AI_VOICE_V2_STREAMING_ASR_CACHE_DIR"] = env.get("AI_VOICE_V2_STREAMING_ASR_CACHE_DIR", str(ROOT / ".models" / "faster_whisper"))
env["AI_VOICE_V2_STREAMING_ASR_SUPPORTED_LANGUAGES"] = env.get("AI_VOICE_V2_STREAMING_ASR_SUPPORTED_LANGUAGES", "ru")
env["AI_VOICE_V2_PREBAKED_ACK_ENABLED"] = env.get("AI_VOICE_V2_PREBAKED_ACK_ENABLED", "1")
env["AI_VOICE_V2_PREBAKED_ACK_DIR"] = env.get("AI_VOICE_V2_PREBAKED_ACK_DIR", str(DATA_DIR / "voice_v2_ack_bank"))
env["AI_VOICE_V2_STREAMING_TTS"] = env.get("AI_VOICE_V2_STREAMING_TTS", "1")
@@ -338,6 +345,8 @@ def spawn_service(spec: dict[str, Any], runtime_dir: Path, data_dir: Path, base_
if spec["name"] == "recording":
env["CC_RECORDINGS_DIR"] = str((data_dir / "recordings").resolve())
env["RECORDING_MAX_BYTES"] = env.get("RECORDING_MAX_BYTES", "26214400")
if spec["name"] == "streaming-asr-sidecar":
Path(env["AI_VOICE_V2_STREAMING_ASR_CACHE_DIR"]).mkdir(parents=True, exist_ok=True)
if spec["name"] == "gateway":
env.update(build_gateway_env(env))