Update app.py

app.py

@@ -45,6 +45,17 @@ from langchain_core.output_parsers import StrOutputParser
 from langchain_core.runnables import RunnableBranch, RunnableLambda, RunnableParallel, RunnablePassthrough
 from serpapi.google_search import GoogleSearch
 
+#Parler TTS v1 Modules 
+
+import os
+import re
+import tempfile
+import soundfile as sf
+from string import punctuation
+from pydub import AudioSegment
+from transformers import AutoTokenizer, AutoFeatureExtractor
+
+
 #API AutoDate Fix Up
 def get_current_date1():
     return datetime.now().strftime("%Y-%m-%d")
@@ -75,7 +86,7 @@ embeddings = OpenAIEmbeddings(api_key=os.environ['OPENAI_API_KEY'])
 from pinecone import Pinecone
 pc = Pinecone(api_key=os.environ['PINECONE_API_KEY'])
 
-index_name = "radardata08152024"
+index_name = "radardata08172024"
 vectorstore = PineconeVectorStore(index_name=index_name, embedding=embeddings)
 retriever = vectorstore.as_retriever(search_kwargs={'k': 5})
 
@@ -133,9 +144,9 @@ QA_CHAIN_PROMPT_1 = PromptTemplate(input_variables=["context", "question"], temp
 QA_CHAIN_PROMPT_2 = PromptTemplate(input_variables=["context", "question"], template=template2)
 
 # Neo4j setup
-graph = Neo4jGraph(url="neo4j+s://bcbd420c.databases.neo4j.io",
+graph = Neo4jGraph(url="neo4j+s://6457770f.databases.neo4j.io",
                     username="neo4j",
-                    password="znKCoED1qIoYrkmZHqVY3sGWxmUvSdI8EqG5hRNlROI"
+                    password="Z10duoPkKCtENuOukw3eIlvl0xJWKtrVSr-_hGX1LQ4"
                     )
 # Avoid pushing the graph documents to Neo4j every time
 # Only push the documents once and comment the code below after the initial push
@@ -335,7 +346,7 @@ def bot(history, choice, tts_choice, retrieval_mode):
 
 def add_message(history, message):
     history.append((message, None))
-    return history, gr.Textbox(value="", interactive=True, placeholder="Ask More Question...", show_label=False)
+    return history, gr.Textbox(value="", interactive=True, placeholder="Ask More Questions...", show_label=False)
 
 def print_like_dislike(x: gr.LikeData):
     print(x.index, x.value, x.liked)
@@ -573,14 +584,13 @@ def generate_audio_elevenlabs(text):
         return None
 
 
-repo_id = "parler-tts/parler-tts-mini-expresso"
+repo_id = "parler-tts/parler-tts-mini-v1"
 
 parler_model = ParlerTTSForConditionalGeneration.from_pretrained(repo_id).to(device)
 parler_tokenizer = AutoTokenizer.from_pretrained(repo_id)
 parler_feature_extractor = AutoFeatureExtractor.from_pretrained(repo_id)
 
 SAMPLE_RATE = parler_feature_extractor.sampling_rate
-SEED = 42
 
 def preprocess(text):
     number_normalizer = EnglishNumberNormalizer()
@@ -597,7 +607,7 @@ def preprocess(text):
     abbreviations = re.findall(abbreviations_pattern, text)
     for abv in abbreviations:
         if abv in text:
-            text is text.replace(abv, separate_abb(abv))
+            text = text.replace(abv, separate_abb(abv))
     return text
 
 def chunk_text(text, max_length=250):
@@ -621,20 +631,19 @@ def chunk_text(text, max_length=250):
     return chunks
 
 def generate_audio_parler_tts(text):
-    description = "Thomas speaks with emphasis and excitement at a moderate pace with high quality."
+    description = "A female speaker delivers a slightly expressive and animated speech with a moderate speed and pitch. The recording is of very high quality, with the speaker's voice sounding clear and very close up."
     chunks = chunk_text(preprocess(text))
     audio_segments = []
 
     for chunk in chunks:
-        inputs = parler_tokenizer(description, return_tensors="pt").to(device)
-        prompt = parler_tokenizer(chunk, return_tensors="pt").to(device)
+        input_ids = parler_tokenizer(description, return_tensors="pt").input_ids.to(device)
+        prompt_input_ids = parler_tokenizer(chunk, return_tensors="pt").input_ids.to(device)
 
-        set_seed(SEED)
-        generation = parler_model.generate(input_ids=inputs.input_ids, prompt_input_ids=prompt.input_ids)
+        generation = parler_model.generate(input_ids=input_ids, prompt_input_ids=prompt_input_ids)
         audio_arr = generation.cpu().numpy().squeeze()
 
         temp_audio_path = os.path.join(tempfile.gettempdir(), f"parler_tts_audio_{len(audio_segments)}.wav")
-        write_wav(temp_audio_path, SAMPLE_RATE, audio_arr)
+        sf.write(temp_audio_path, audio_arr, parler_model.config.sampling_rate)
         audio_segments.append(AudioSegment.from_wav(temp_audio_path))
 
     combined_audio = sum(audio_segments)
@@ -643,6 +652,7 @@ def generate_audio_parler_tts(text):
 
     logging.debug(f"Audio saved to {combined_audio_path}")
     return combined_audio_path
+    
 
 # Load the MARS5 model
 mars5, config_class = torch.hub.load('Camb-ai/mars5-tts', 'mars5_english', trust_repo=True)
@@ -1055,7 +1065,7 @@ with gr.Blocks(theme='Pijush2023/scikit-learn-pijush') as demo:
 
             gr.Markdown("<h1 style='color: red;'>Talk to RADAR</h1>", elem_id="voice-markdown")
 
-            chat_input = gr.Textbox(show_copy_button=True, interactive=True, show_label=False, label="ASK Radar !!!",placeholder="Hey Radar...!!")
+            chat_input = gr.Textbox(show_copy_button=True, interactive=True, show_label=False, label="ASK Radar !!!", placeholder="Hey Radar...!!")
             tts_choice = gr.Radio(label="Select TTS System", choices=["Alpha", "Beta", "Gamma"], value="Alpha")
             retriever_button = gr.Button("Retriever")
 
@@ -1121,4 +1131,3 @@ demo.launch(share=True)
 
 
 
-