Update app.py

app.py

@@ -3,4 +3,39 @@ from transformers.pipelines.audio_utils import ffmpeg_microphone_live
 import torch
 
 asr_model = "openai/whisper-tiny.en"
-nlp_model = "MoritzLaurer/DeBERTa-v3-base-mnli-fever-anli"
+nlp_model = "MoritzLaurer/DeBERTa-v3-base-mnli-fever-anli"
+
+pipe = pipeline("automatic-speech-recognition", model=model_id, device=device)
+sampling_rate = pipe.feature_extractor.sampling_rate
+
+chunk_length_s = 10  # how often returns the text
+stream_chunk_s = 1  # how often the microphone is checked for new audio
+mic = ffmpeg_microphone_live(
+    sampling_rate=sampling_rate,
+    chunk_length_s=chunk_length_s,
+    stream_chunk_s=stream_chunk_s,
+)
+
+def listen_print_loop(responses):
+    for response in responses:
+        if response["text"]:
+            print(response["text"], end="\r")
+            return response["text"]
+        if not response["partial"]:
+            print("")
+
+
+classifier = pipeline("zero-shot-classification", model=nlp_model)
+candidate_labels = ["dim the light", "turn on light fully", "turn off light fully", "raise the light", "nothing about light"]
+
+
+while True:
+    context = listen_print_loop(pipe(mic))
+    print(context)
+    output = classifier(context, candidate_labels, multi_label=False)
+    top_label = output['labels'][0]
+    top_score = output['scores'][0]
+    print(f"Top Prediction: {top_label} with a score of {top_score:.2f}")
+
+
+