Update app.py

app.py

@@ -7,6 +7,14 @@ import scipy.io.wavfile as wavfile # For writing WAV data to in-memory file
 import numpy as np # To handle the audio array
 import datetime # For logging timestamps if needed (not directly used in this version, but good practice)
 
+# --- Fetch API Key from Environment Variable ---
+# This is the SECURE way to handle API keys in Hugging Face Spaces.
+# You MUST set an environment variable named OPENAI_API_KEY in your Space's settings.
+OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
+
+# --- Define the OpenAI Model to use ---
+OPENAI_MODEL = "gpt-3.5-turbo" # Changed from gpt-4o to gpt-3.5-turbo
+
 system_prompt = """
     You are a sophisticated AI voice bot representing Krishnavamshi Thumma. Your persona should be that of a highly skilled, professional, and engaging Generative AI and Data Engineering enthusiast. When responding to questions, embody the following detailed professional identity:
 
@@ -21,13 +29,13 @@ system_prompt = """
 
     **Technical Skills:**
     Your robust technical skill set includes:
-    * **Languages:** Python, SQL, JavaScript (Node.js) 
-    * **GenAI/ML:** OpenAI GPT-4o, LangChain, Transformers Architecture, LLMs, RAG 
-    * **Vector Databases:** FAISS, MongoDB Vector Search 
-    * **Data Engineering Tools:** Apache Airflow, AWS Lambda, REST APIs, Pandas, PyPDF2, BeautifulSoup, FastAPI, Streamlit 
-    * **Cloud & Infrastructure:** AWS S3, GCP, Docker, Terraform 
-    * **Version Control:** Git, GitHub 
-    * **Other Relevant Skills:** Data Structures & Algorithms (DSA), Content-Based Retrieval, Prompt Engineering 
+    * **Languages:** Python, SQL, JavaScript (Node.js)
+    * **GenAI/ML:** OpenAI GPT-4o, LangChain, Transformers Architecture, LLMs, RAG
+    * **Vector Databases:** FAISS, MongoDB Vector Search
+    * **Data Engineering Tools:** Apache Airflow, AWS Lambda, REST APIs, Pandas, PyPDF2, BeautifulSoup, FastAPI, Streamlit
+    * **Cloud & Infrastructure:** AWS S3, GCP, Docker, Terraform
+    * **Version Control:** Git, GitHub
+    * **Other Relevant Skills:** Data Structures & Algorithms (DSA), Content-Based Retrieval, Prompt Engineering
 
     **Key Projects & Expertise Areas:**
 
@@ -46,24 +54,23 @@ system_prompt = """
 r = sr.Recognizer()
 
 # Modified function to accept audio as a numpy array and samplerate
-def transcribe_audio_and_chat(audio_tuple, history, api_key):
-    # 'history' now directly contains the list of message dictionaries
-    if not api_key:
-        raise gr.Error("❌ Please enter your OpenAI API key.")
-
-    if audio_tuple is None:
-        # If no audio is received, add an assistant message to history and reset audio input
-        return history, history, None
+def transcribe_audio_and_chat(audio_tuple, history): # Removed api_key from function arguments
+    # Check if API key is available in environment
+    if not OPENAI_API_KEY:
+        raise gr.Error("❌ OpenAI API key not found. Please set OPENAI_API_KEY as a Space Secret.")
 
-    # Ensure history is a list, even if it somehow became None (defensive programming)
+    # Handle cases where history might be None (defensive programming)
     if history is None:
         history = []
 
+    if audio_tuple is None:
+        # If no audio, raise a Gradio Error directly instead of adding to chat history
+        raise gr.Error("No audio received. Please speak into the microphone.")
+
     samplerate, audio_np_array = audio_tuple
 
     try:
         # Convert the NumPy array to a format speech_recognition can handle (in-memory WAV)
-        # Ensure the array is int16 as it's a common format for audio samples and expected by scipy.io.wavfile
         if audio_np_array.dtype != np.int16:
              audio_np_array = audio_np_array.astype(np.int16)
 
@@ -88,16 +95,16 @@ def transcribe_audio_and_chat(audio_tuple, history, api_key):
             return history, history, None # Reset audio input after error
 
         # --- Proceed with OpenAI chat ---
-        client = OpenAI(api_key=api_key)
+        # Use the global OPENAI_API_KEY
+        client = OpenAI(api_key=OPENAI_API_KEY)
 
         # Create the full messages list for OpenAI, starting with the system prompt
-        # and then appending the current chat history, followed by the new user input.
         messages_for_openai = [{"role": "system", "content": system_prompt}] + history
         messages_for_openai.append({"role": "user", "content": user_input})
 
-        # Get response from OpenAI
+        # Get response from OpenAI using the specified model
         response = client.chat.completions.create(
-            model="gpt-4o",
+            model=OPENAI_MODEL, # Use the global OPENAI_MODEL
             messages=messages_for_openai, # Pass the correctly formatted messages
             temperature=0.7
         )
@@ -105,7 +112,6 @@ def transcribe_audio_and_chat(audio_tuple, history, api_key):
         bot_reply = response.choices[0].message.content
         
         # Append both the user input and bot reply to the *Gradio* history (state)
-        # in the 'messages' format that Gradio's chatbot expects.
         history.append({"role": "user", "content": user_input})
         history.append({"role": "assistant", "content": bot_reply})
         
@@ -116,7 +122,6 @@ def transcribe_audio_and_chat(audio_tuple, history, api_key):
 
     except Exception as e:
         print(f"An unexpected error occurred: {e}") # Log the error for debugging
-        # If an unexpected error occurs, still try to reset the audio input
         raise gr.Error(f"❌ An unexpected error occurred: {str(e)}")
 
 
@@ -149,7 +154,7 @@ with gr.Blocks(title="Voice Bot: Krishnavamshi Thumma") as demo:
         #audioInputComponent {
             margin-top: 20px;
         }
-        .key-status {
+        .key-status { /* No longer strictly needed but keeping for style consistency if other status messages arise */
             padding: 5px;
             margin-top: 5px;
             border-radius: 4px;
@@ -165,13 +170,9 @@ with gr.Blocks(title="Voice Bot: Krishnavamshi Thumma") as demo:
     </style>
     """)
 
-    api_key = gr.Textbox(
-    label="🔐 OpenAI API Key (GPT-4o required)",
-    placeholder="Paste your OpenAI GPT-4o API key here (not GPT-3/3.5-turbo)",
-    type="password",
-    elem_id="apiKeyInput"
-    )
-    key_status = gr.HTML("<div id='keyStatus'></div>")
+    # Removed the API key textbox and its status display as it's now handled by Space Secrets
+    # api_key = gr.Textbox(...)
+    # key_status = gr.HTML(...)
     
     # Chatbot component to display messages
     chatbot = gr.Chatbot(elem_id="chatBox", type="messages", height=400)
@@ -192,27 +193,16 @@ with gr.Blocks(title="Voice Bot: Krishnavamshi Thumma") as demo:
     # Event handler: When audio input is recorded and submitted (by stopping recording)
     audio_input.change(
         fn=transcribe_audio_and_chat,
-        inputs=[audio_input, state, api_key],
+        inputs=[audio_input, state], # Removed api_key from inputs as it's global
         # Outputs: 1. chatbot display, 2. state (updated history), 3. audio_input (to clear it)
         outputs=[chatbot, state, audio_input] 
     )
 
-    # JavaScript for API key input and status display
+    # Removed JavaScript related to API key input
     gr.HTML("""
     <script>
-        document.getElementById("apiKeyInput").addEventListener("input", function() {
-            const apiKey = this.value.trim();
-            const keyStatus = document.getElementById("keyStatus");
-            
-            if (apiKey) {
-                keyStatus.innerHTML = '<div class="key-status success">API Key saved successfully!</div>';
-            } else {
-                keyStatus.innerHTML = '<div class="key-status error">Please enter a valid API key</div>';
-            }
-        });
-
-        // Focus on API key input on page load for convenience
-        document.querySelector("#apiKeyInput input").focus();
+        // No specific API key JS needed anymore as it's handled by secrets
+        // You can add other useful JS here if needed in the future
     </script>
     """)