Add SmallestAI Pulse STT and Lightning TTS to pipecat cascade pipeline

.env.example

@@ -69,7 +69,7 @@ EVA_MODEL__LLM=gpt-5.2
 # --- LLM mode: STT ---
 #i STT provider for the voice pipeline.
 #d enum
-#e assemblyai,cartesia,cartesia-multilingual,deepgram,deepgram-flux,elevenlabs,nvidia,nvidia-baseten,openai
+#e assemblyai,cartesia,cartesia-multilingual,deepgram,deepgram-flux,elevenlabs,nvidia,nvidia-baseten,openai,smallest
 #x pipeline_mode=LLM
 EVA_MODEL__STT=cartesia
 
@@ -87,7 +87,7 @@ EVA_MODEL__STT_PARAMS='{"api_key": "your_cartesia_api_key", "model": "ink-2"}'
 # --- TTS (LLM and AudioLLM modes) ---
 #i TTS provider for the voice pipeline.
 #d enum
-#e cartesia,chatterbox,elevenlabs,gemini,kokoro,nvidia-baseten,openai,xtts
+#e cartesia,chatterbox,elevenlabs,gemini,kokoro,nvidia-baseten,openai,smallest,xtts
 #x pipeline_mode=LLM,AudioLLM
 EVA_MODEL__TTS=cartesia
 

src/eva/__init__.py

@@ -7,7 +7,7 @@
 
 # Bump simulation_version when changes affect benchmark outputs (agent code,
 # user simulator, orchestrator, simulation prompts, agent configs, tool mocks).
-simulation_version = "2.0.1"
+simulation_version = "2.0.2"
 
 # Bump metrics_version when changes affect metric computation (metrics code,
 # judge prompts, pricing tables, postprocessor).

src/eva/assistant/pipeline/services.py

@@ -37,6 +37,8 @@
 from pipecat.services.openai.realtime.llm import OpenAIRealtimeLLMService
 from pipecat.services.openai.stt import OpenAISTTService
 from pipecat.services.openai.tts import OpenAITTSService
+from pipecat.services.smallest.stt import SmallestSTTService
+from pipecat.services.smallest.tts import SmallestTTSService
 from pipecat.services.stt_service import STTService
 from pipecat.services.tts_service import TTSService
 from pipecat.services.xai.stt import XAISTTService
@@ -279,6 +281,21 @@ def create_stt_service(
             )
         return stt_service
 
+    elif model_lower == "smallest":
+        logger.info(f"Using Smallest Pulse STT: {params['model']}")
+        smallest_stt_settings_kwargs = {
+            k: params[k] for f in dataclasses.fields(SmallestSTTService.Settings) if (k := f.name) in params
+        }
+        return SmallestSTTService(
+            api_key=api_key,
+            base_url=url or "wss://api.smallest.ai",
+            sample_rate=params.get("sample_rate", SAMPLE_RATE),
+            settings=SmallestSTTService.Settings(
+                language=_to_language_enum(language_code),
+                **smallest_stt_settings_kwargs,
+            ),
+        )
+
     elif model_lower == "xai":
         logger.info("Using xAI STT")
         xai_settings_kwargs = {
@@ -298,7 +315,7 @@ def create_stt_service(
 
     else:
         raise ValueError(
-            f"Unknown STT model: {model}. Available: assemblyai, cartesia, cartesia-multilingual, cohere, deepgram, deepgram-flux, elevenlabs, nvidia, nvidia-baseten, openai, xai"
+            f"Unknown STT model: {model}. Available: assemblyai, cartesia, cartesia-multilingual, cohere, deepgram, deepgram-flux, elevenlabs, nvidia, nvidia-baseten, openai, smallest, xai"
         )
 
 
@@ -500,6 +517,25 @@ def create_tts_service(
 
         return openai_tts
 
+    elif model_lower == "smallest":
+        logger.info(f"Using Smallest Lightning TTS: {params['model']}")
+        smallest_tts_settings_kwargs = {
+            k: params[k] for f in dataclasses.fields(SmallestTTSService.Settings) if (k := f.name) in params
+        }
+        if "voice_id" in params:
+            # Omit "voice" entirely when unset so pipecat's per-model default (e.g. "sophia" for
+            # lightning_v3.1, "meher" for lightning_v3.1_pro) applies instead of a hardcoded EVA default.
+            smallest_tts_settings_kwargs["voice"] = params["voice_id"]
+        return SmallestTTSService(
+            api_key=api_key,
+            base_url=url or "wss://api.smallest.ai",
+            sample_rate=SAMPLE_RATE,
+            settings=SmallestTTSService.Settings(
+                language=_to_language_enum(language_code),
+                **smallest_tts_settings_kwargs,
+            ),
+        )
+
     elif model_lower == "voxtral":
         logger.info(f"Using Voxtral TTS: {params['model']}")
         voxtral_tts = OpenAITTSService(
@@ -557,7 +593,7 @@ def create_tts_service(
 
     else:
         raise ValueError(
-            f"Unknown TTS model: {model}. Available: cartesia, chatterbox, deepgram, elevenlabs, gemini, kokoro, nvidia-baseten, openai, xai, xtts"
+            f"Unknown TTS model: {model}. Available: cartesia, chatterbox, deepgram, elevenlabs, gemini, kokoro, nvidia-baseten, openai, smallest, voxtral, xai, xtts"
         )
 
 

tests/unit/assistant/test_services.py

@@ -151,6 +151,19 @@ def test_xai_stt_defaults_and_forwards(self):
         assert svc._settings.endpointing == 42  # Explicit override wins over the Eva default
         assert svc._settings.interim_results is True  # EVA default preserved
 
+    def test_smallest_returns_smallest_service(self):
+        svc = create_stt_service("smallest", params={"api_key": "k", "model": "pulse"})
+        assert "Smallest" in type(svc).__name__
+        assert svc._settings.model == "pulse"
+
+    def test_smallest_forwards_optional_settings(self):
+        svc = create_stt_service(
+            "smallest",
+            params={"api_key": "k", "model": "pulse", "word_timestamps": True, "diarize": True},
+        )
+        assert svc._settings.word_timestamps is True
+        assert svc._settings.diarize is True
+
     def test_cartesia_is_ink2_turns_service(self):
         svc = create_stt_service("cartesia", params={"api_key": "k", "model": "ink-2"})
         assert "Turns" in type(svc).__name__
@@ -256,6 +269,23 @@ def test_elevenlabs_forwards_voice_tuning(self):
         assert svc._settings.stability == 0.7
         assert svc._settings.speed == 1.1
 
+    def test_smallest_returns_smallest_service(self):
+        svc = create_tts_service("smallest", params={"api_key": "k", "model": "lightning_v3.1"})
+        assert "Smallest" in type(svc).__name__
+
+    def test_smallest_forwards_voice_id_and_speed(self):
+        svc = create_tts_service(
+            "smallest",
+            params={"api_key": "k", "model": "lightning_v3.1", "voice_id": "sophia", "speed": 1.2},
+        )
+        assert svc._settings.voice == "sophia"
+        assert svc._settings.speed == 1.2
+
+    def test_smallest_defaults_voice_per_model_when_unset(self):
+        """Omitting voice_id must not override pipecat's per-model default voice."""
+        svc = create_tts_service("smallest", params={"api_key": "k", "model": "lightning_v3.1_pro"})
+        assert svc._settings.voice == "meher"
+
     def test_openai_respects_voice_param(self):
         svc = create_tts_service("openai", params={"api_key": "k", "model": "tts-1", "voice": "nova"})
         assert svc._settings.voice == "nova"