update

app.py

@@ -7,12 +7,23 @@ import requests
 import tempfile
 from neon_tts_plugin_coqui import CoquiTTS
 from datasets import load_dataset
-#import whisper 
+import random
+
+dataset = load_dataset("ysharma/short_jokes", split="train")
+
+# Model 2: Sentence Transformer
+API_URL = "https://api-inference.huggingface.co/models/sentence-transformers/msmarco-distilbert-base-tas-b"
+HF_TOKEN = os.environ["HF_TOKEN"]
+headers = {"Authorization": f"Bearer {HF_TOKEN}"}
+
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
+
 
-dataset = load_dataset("ysharma/short_jokes")
 
 # Language common in both the multilingual models - English, Chinese, Spanish, and French etc
-# /model 1: Whisper: Speech-to-text
+# Model 1: Whisper: Speech-to-text
 model = whisper.load_model("base")
 #model_med = whisper.load_model("medium")
 # Languages covered in Whisper - (exhaustive list) : 
@@ -48,11 +59,25 @@ print(f"Languages for Coqui are: {LANGUAGES}")
 # nl - dutch, fi - finnish, sl - slovenian, lv - latvian, ga - ??  
 
 
+  
 # Driver function
 def driver_fun(audio) : 
   translation, lang = whisper_stt(audio)  # older : transcribe, translation, lang 
   #text1 = model.transcribe(audio)["text"]
   
+  random_val = random.randrange(0,231657)
+  if random_val < 226657:
+    lower_limit = random_val
+    upper_limit = random_val + 5000 
+  else:
+    lower_limit = random_val - 5000
+    upper_limit = random_val 
+  print(f"lower_limit : upper_limit = {lower_limit} : {upper_limit}")  
+  dataset_subset = dataset['Joke'][lower_limit : upper_limit]
+  data = query({"inputs": {"source_sentence": "That is a happy person","sentences": dataset_subset} } )
+  max_match_score = max(data)
+  indx_score = data.index(max_match_score)
+  joke = max_match_score[indx_score]
   #if translation 
   #For now only taking in English text for Bloom prompting as inference model is not high spec
   #text_generated = lang_model_response(transcribe, lang)
@@ -61,8 +86,8 @@ def driver_fun(audio) :
   #if lang in ['es', 'fr']:
   #  speech = tts(transcribe, lang)
   #else:
-  speech = tts(translation, 'en') #'en')
-  return translation, speech #transcribe, 
+  speech = tts(joke, 'en') #'en' # translation
+  return translation, joke, speech #transcribe, 
 
 
 # Whisper - speech-to-text
@@ -117,9 +142,9 @@ with demo:
       #out_translation_en = gr.Textbox(label= 'English Translation of audio using OpenAI Whisper')
     with gr.Column():
       out_audio = gr.Audio(label='Audio response form CoquiTTS')  
-      #out_generated_text = gr.Textbox(label= 'AI response to your query in your preferred language using Bloom! ')
+      out_generated_joke = gr.Textbox(label= 'Joke returned! ')
       #out_generated_text_en = gr.Textbox(label= 'AI response to your query in English using Bloom! ')
     
-      b1.click(driver_fun,inputs=[in_audio], outputs=[out_transcript, out_audio]) #out_translation_en, out_generated_text,out_generated_text_en, 
+      b1.click(driver_fun,inputs=[in_audio], outputs=[out_transcript, out_generated_joke, out_audio]) #out_translation_en, out_generated_text,out_generated_text_en, 
     
 demo.launch(enable_queue=True, debug=True)