Update app.py

app.py

@@ -43,7 +43,7 @@ from langchain_core.pydantic_v1 import BaseModel, Field
 from langchain_core.messages import AIMessage, HumanMessage
 from langchain_core.output_parsers import StrOutputParser
 from langchain_core.runnables import RunnableBranch, RunnableLambda, RunnableParallel, RunnablePassthrough
-
+from serpapi.google_search import GoogleSearch
 
 # Set environment variables for CUDA
 os.environ['PYTORCH_USE_CUDA_DSA'] = '1'
@@ -277,11 +277,48 @@ def generate_answer(message, choice, retrieval_mode):
     else:
         return "Invalid retrieval mode selected.", []
 
+# def bot(history, choice, tts_choice, retrieval_mode):
+#     if not history:
+#         return history
+    
+#     response, addresses = generate_answer(history[-1][0], choice, retrieval_mode)
+#     history[-1][1] = ""
+
+#     with concurrent.futures.ThreadPoolExecutor() as executor:
+#         if tts_choice == "Alpha":
+#             audio_future = executor.submit(generate_audio_elevenlabs, response)
+#         elif tts_choice == "Beta":
+#             audio_future = executor.submit(generate_audio_parler_tts, response)
+#         elif tts_choice == "Gamma":
+#             audio_future = executor.submit(generate_audio_mars5, response)
+
+#         for character in response:
+#             history[-1][1] += character
+#             time.sleep(0.05)
+#             yield history, None
+
+#         audio_path = audio_future.result()
+#         yield history, audio_path
+
+#     history.append([response, None])  # Ensure the response is added in the correct format
+
+
 def bot(history, choice, tts_choice, retrieval_mode):
     if not history:
         return history
     
-    response, addresses = generate_answer(history[-1][0], choice, retrieval_mode)
+    user_message = history[-1][0]
+    response = ""
+    addresses = []
+
+    # Check if the question is about restaurants
+    if "restaurant" in user_message.lower():
+        # Use the agent to get the response
+        response = agent.run(user_message)
+    else:
+        # Otherwise, use the existing logic
+        response, addresses = generate_answer(user_message, choice, retrieval_mode)
+    
     history[-1][1] = ""
 
     with concurrent.futures.ThreadPoolExecutor() as executor:
@@ -302,6 +339,7 @@ def bot(history, choice, tts_choice, retrieval_mode):
 
     history.append([response, None])  # Ensure the response is added in the correct format
 
+
 def add_message(history, message):
     history.append((message, None))
     return history, gr.Textbox(value="", interactive=True, placeholder="Enter message or upload file...", show_label=False)
@@ -426,83 +464,9 @@ def fetch_local_news():
     else:
         return "<p>Failed to fetch local news</p>"
 
-# import numpy as np
-# import torch
-# from transformers import pipeline, AutoModelForSpeechSeq2Seq, AutoProcessor
-
-# model_id = 'openai/whisper-large-v3'
-# device = "cuda:0" if torch.cuda.is_available() else "cpu"
-# torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
-# model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id, torch_dtype=torch_dtype).to(device)
-# processor = AutoProcessor.from_pretrained(model_id)
-
-# pipe_asr = pipeline("automatic-speech-recognition", model=model, tokenizer=processor.tokenizer, feature_extractor=processor.feature_extractor, max_new_tokens=128, chunk_length_s=15, batch_size=16, torch_dtype=torch_dtype, device=device, return_timestamps=True)
-
-# base_audio_drive = "/data/audio"
-
-#Normal Code with sample rate is 44100 Hz
-
-# def transcribe_function(stream, new_chunk):
-#     try:
-#         sr, y = new_chunk[0], new_chunk[1]
-#     except TypeError:
-#         print(f"Error chunk structure: {type(new_chunk)}, content: {new_chunk}")
-#         return stream, "", None
-
-#     y = y.astype(np.float32) / np.max(np.abs(y))
-
-#     if stream is not None:
-#         stream = np.concatenate([stream, y])
-#     else:
-#         stream = y
-
-#     result = pipe_asr({"array": stream, "sampling_rate": sr}, return_timestamps=False)
-
-#     full_text = result.get("text","")
-    
-#     return stream, full_text, result
-
-# Resampling code with 16000 Hz
-
-# import numpy as np
-# from scipy.signal import resample
-
-# def transcribe_function(stream, new_chunk):
-#     try:
-#         sr, y = new_chunk[0], new_chunk[1]
-#     except TypeError:
-#         print(f"Error chunk structure: {type(new_chunk)}, content: {new_chunk}")
-#         return stream, "", None
-
-#     # Resample to 16000 Hz
-#     target_sr = 16000
-#     if sr != target_sr:
-#         num_samples = int(len(y) * float(target_sr) / sr)
-#         y = resample(y, num_samples)
-#         sr = target_sr
-
-#     y = y.astype(np.float32) / np.max(np.abs(y))
-
-#     if stream is not None:
-#         stream = np.concatenate([stream, y])
-#     else:
-#         stream = y
-
-#     result = pipe_asr({"array": stream, "sampling_rate": sr}, return_timestamps=False)
-
-#     full_text = result.get("text", "")
-    
-#     return stream, full_text, result
-
-#Resample part -1
-
 import numpy as np
 import torch
 from transformers import pipeline, AutoModelForSpeechSeq2Seq, AutoProcessor
-from scipy.signal import resample  # Import resample from scipy.signal
-import base64
-import io
-from pydub import AudioSegment
 
 model_id = 'openai/whisper-large-v3'
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
@@ -512,20 +476,9 @@ processor = AutoProcessor.from_pretrained(model_id)
 
 pipe_asr = pipeline("automatic-speech-recognition", model=model, tokenizer=processor.tokenizer, feature_extractor=processor.feature_extractor, max_new_tokens=128, chunk_length_s=15, batch_size=16, torch_dtype=torch_dtype, device=device, return_timestamps=True)
 
-def is_base64_audio(data):
-    try:
-        if isinstance(data, str):
-            base64.b64decode(data)
-            return True
-        return False
-    except Exception:
-        return False
-
-def base64_to_float32(base64_str):
-    audio_bytes = base64.b64decode(base64_str)
-    audio_segment = AudioSegment.from_file(io.BytesIO(audio_bytes), format="wav")
-    samples = np.array(audio_segment.get_array_of_samples())
-    return audio_segment.frame_rate, samples.astype(np.float32)
+base_audio_drive = "/data/audio"
+
+#Normal Code with sample rate is 44100 Hz
 
 def transcribe_function(stream, new_chunk):
     try:
@@ -534,17 +487,6 @@ def transcribe_function(stream, new_chunk):
         print(f"Error chunk structure: {type(new_chunk)}, content: {new_chunk}")
         return stream, "", None
 
-    # Check if input is base64 and convert to float32 if necessary
-    if is_base64_audio(y):
-        sr, y = base64_to_float32(y)
-
-    # Ensure the sample rate is 16000 Hz
-    target_sr = 16000
-    if sr != target_sr:
-        num_samples = int(len(y) * float(target_sr) / sr)
-        y = resample(y, num_samples)
-        sr = target_sr
-
     y = y.astype(np.float32) / np.max(np.abs(y))
 
     if stream is not None:
@@ -554,12 +496,14 @@ def transcribe_function(stream, new_chunk):
 
     result = pipe_asr({"array": stream, "sampling_rate": sr}, return_timestamps=False)
 
-    full_text = result.get("text", "")
+    full_text = result.get("text","")
     
     return stream, full_text, result
 
 
 
+
+
 def update_map_with_response(history):
     if not history:
         return ""
@@ -746,6 +690,35 @@ def update_images():
     image_3 = generate_image(hardcoded_prompt_3)
     return image_1, image_2, image_3
 
+
+
+#Crew AI Integration
+
+from crewai import Agent
+from langchain.agents import Tool
+from langchain.utilities import GoogleSerperAPIWrapper
+
+# Setup API keys
+os.environ["SERP_API"] = "Your Key"
+
+search = GoogleSerperAPIWrapper()
+
+# Create and assign the search tool to an agent
+serper_tool = Tool(
+    name="Intermediate Answer",
+    func=search.run,
+    description="Useful for search-based queries",
+)
+
+agent = Agent(
+    role='Research Analyst',
+    goal='Provide up-to-date market analysis',
+    backstory='An expert analyst with a keen eye for market trends.',
+    tools=[serper_tool]
+)
+
+
+
 def fetch_local_events():
     api_key = os.environ['SERP_API']
     url = f'https://serpapi.com/search.json?engine=google_events&q=Events+in+Birmingham&hl=en&gl=us&api_key={api_key}'
@@ -902,6 +875,90 @@ def handle_retrieval_mode_change(choice):
         return gr.update(interactive=True), gr.update(interactive=True)
 
 
+def fetch_yelp_restaurants():
+    from serpapi.google_search import GoogleSearch
+    import os
+
+    params = {
+      "engine": "yelp",
+      "find_desc": "Restaurant",
+      "find_loc": "Birmingham, AL, USA",
+      "api_key": os.getenv("SERP_API")
+    }
+
+    search = GoogleSearch(params)
+    results = search.get_dict()
+    organic_results = results.get("organic_results", [])
+
+    yelp_html = """
+    <h2 style="font-family: 'Georgia', serif; color: #ff0000; background-color: #f8f8f8; padding: 10px; border-radius: 10px;">Top Restaurants</h2>
+    <style>
+        table {
+            font-family: 'Verdana', sans-serif;
+            color: #333;
+            border-collapse: collapse;
+            width: 100%;
+        }
+        th, td {
+            border: 1px solid #fff !important;
+            padding: 8px;
+        }
+        th {
+            background-color: #f2f2f2;
+            color: #333;
+            text-align: left;
+        }
+        tr:hover {
+            background-color: #f5f5f5;
+        }
+        .restaurant-link {
+            color: #1E90FF;
+            text-decoration: none;
+        }
+        .restaurant-link:hover {
+            text-decoration: underline;
+        }
+    </style>
+    <table>
+        <tr>
+            <th>Name</th>
+            <th>Rating</th>
+            <th>Reviews</th>
+            <th>Phone</th>
+            <th>Snippet</th>
+            <th>Services</th>
+        </tr>
+    """
+    for result in organic_results:
+        name = result.get("title", "No name")
+        rating = result.get("rating", "No rating")
+        reviews = result.get("reviews", "No reviews")
+        phone = result.get("phone", "Not Available")
+        snippet = result.get("snippet", "No Available")
+        services = result.get("service_options", "Not Known")
+
+        if isinstance(services, list):
+            services = ", ".join(services)
+        elif isinstance(services, dict):
+            services = ", ".join([f"{key}: {value}" for key, value in services.items()])
+
+        link = result.get("link", "#")
+
+        yelp_html += f"""
+        <tr>
+            <td><a class='restaurant-link' href='{link}' target='_blank'>{name}</a></td>
+            <td>{rating}</td>
+            <td>{reviews}</td>
+            <td>{phone}</td>
+            <td>{snippet}</td>
+            <td>{services}</td>
+        </tr>
+        """
+    yelp_html += "</table>"
+    return yelp_html
+
+
+
 with gr.Blocks(theme='Pijush2023/scikit-learn-pijush') as demo:
     with gr.Row():
         with gr.Column():
@@ -958,7 +1015,8 @@ with gr.Blocks(theme='Pijush2023/scikit-learn-pijush') as demo:
         with gr.Column():
             weather_output = gr.HTML(value=fetch_local_weather())
             news_output = gr.HTML(value=fetch_local_news())
-            events_output = gr.HTML(value=fetch_local_events())            
+            events_output = gr.HTML(value=fetch_local_events())
+            restaurant_output=gr.HTML(value=fetch_yelp_restaurants())
         
 
         with gr.Column():