Update app.py

app.py

@@ -47,10 +47,14 @@ def transcribe(audio_path):
         text = model.stt_file(audio_file)[0]
     return text
 
+# Inference function using Hugging Face InferenceClient
 # Inference function using Hugging Face InferenceClient
 @spaces.GPU(duration=120)
 def model(text):
-    system_instructions = "[SYSTEM] You are CrucialCoach, an AI-powered conversational coach. Guide the user through challenging workplace situations using the principles from 'Crucial Conversations'. Ask one question at a time and provide step-by-step guidance.\n\n[USER]"
+    system_instructions = (
+        "[SYSTEM] You are OralCoach, an AI-powered conversational coach. Guide the student through their oral responses "
+        "using the principles from their English curriculum. Ask one question at a time and provide step-by-step guidance.\n\n[USER]"
+    )
     generate_kwargs = dict(
         temperature=0.7,
         max_new_tokens=512,
@@ -76,7 +80,7 @@ async def generate_audio_feedback(feedback_text):
         await communicate.save(tmp_path)
     return tmp_path
 
-# Generating feedback for the Oral Coach
+#generate feedback
 async def generate_feedback(user_id, question_choice, strategy_choice, message, feedback_level):
     current_question_index = thinkingframes.questions.index(question_choice)
     strategy, explanation = thinkingframes.strategy_options[strategy_choice]
@@ -115,7 +119,9 @@ async def generate_feedback(user_id, question_choice, strategy_choice, message,
         "content": message
     }]
 
-    response = model(conversation)
+    user_message = conversation[1]["content"]  # Extract the user message from the conversation
+    response = model(user_message)
+
     chat_history = []  # Initialize chat history outside the loop
     full_feedback = ""  # Accumulate the entire feedback message
     try:
@@ -168,6 +174,15 @@ async def predict(question_choice, strategy_choice, feedback_level, audio):
         chat_history.append(("Oral Coach  ⚡ ϞϞ(๑⚈ ․̫ ⚈๑)∩ ⚡", "Transcription complete. Generating feedback. Please continue listening to your oral response while waiting ..."))
         yield chat_history, current_audio_output
 
+        moderation_response = client.moderations.create(input=student_response)
+        flagged = any(result.flagged for result in moderation_response.results)
+        if flagged:
+            moderated_message = "The message has been flagged. Please see your teacher to clarify."
+            questionNo = thinkingframes.questions.index(question_choice) + 1
+            add_submission(int(user_state.value), moderated_message, "", int(0), "", questionNo)
+            yield chat_history, current_audio_output
+            return
+
         accumulated_feedback = ""  # Variable to store the accumulated feedback
 
         async for feedback_chunk in generate_feedback(int(user_state.value), question_choice, strategy_choice, student_response, feedback_level):
@@ -187,7 +202,6 @@ async def predict(question_choice, strategy_choice, feedback_level, audio):
         logging.error(f"An error occurred: {str(e)}", exc_info=True)
         yield [("Oral Coach  ⚡ ϞϞ(๑⚈ ․̫ ⚈๑)∩ ⚡", "An error occurred. Please try again or seek assistance.")], current_audio_output
 
-
 with gr.Blocks(title="Oral Coach powered by ZeroGPU⚡ϞϞ(๑⚈ ․̫ ⚈๑)∩ ⚡ and Meta AI 🦙 (LLama3)", theme=theme, css="footer {visibility: hidden}textbox{resize:none}") as demo:    
     with gr.Tab("Oral Coach  ⚡ ϞϞ(๑⚈ ․̫ ⚈๑)∩ ⚡"):
         gr.Markdown("## Student Information")