Update app.py

app.py

@@ -21,12 +21,10 @@ from utils import (
 )
 from prompts import SYSTEM_PROMPT
 
-# The new Q&A with mic
-from qa import AudioBufferProcessor, handle_qa_exchange, transcribe_audio_deepgram
+# We are no longer importing streamlit-webrtc or mic-based Q&A
+# from qa import AudioBufferProcessor, handle_qa_exchange, transcribe_audio_deepgram
 
-from streamlit_webrtc import webrtc_streamer, WebRtcMode, RTCConfiguration
-
-MAX_QA_QUESTIONS = 5
+MAX_QA_QUESTIONS = 5  # Up to 5 typed follow-up questions
 
 def parse_user_edited_transcript(edited_text: str, host_name: str, guest_name: str):
     pattern = r"\*\*(.+?)\*\*:\s*(.+)"
@@ -65,7 +63,7 @@ def parse_user_edited_transcript(edited_text: str, host_name: str, guest_name: s
 def regenerate_audio_from_dialogue(dialogue_items, custom_bg_music_path=None):
     audio_segments = []
     transcript = ""
-    crossfade_duration = 50
+    crossfade_duration = 50  # ms
 
     for item in dialogue_items:
         audio_file = generate_audio_mp3(item.text, item.speaker)
@@ -146,7 +144,6 @@ def generate_podcast(
         except Exception as e:
             return None, f"Error researching topic: {str(e)}"
 
-    from utils import truncate_text
     text = truncate_text(text)
 
     extra_instructions = []
@@ -164,10 +161,9 @@ def generate_podcast(
         )
 
     from prompts import SYSTEM_PROMPT
-    from utils import generate_script, generate_audio_mp3, mix_with_bg_music
-    combined_instructions = "\n\n".join(extra_instructions).strip()
     full_prompt = SYSTEM_PROMPT
-    if combined_instructions:
+    if extra_instructions:
+        combined_instructions = "\n\n".join(extra_instructions).strip()
         full_prompt += f"\n\n# Additional Instructions\n{combined_instructions}\n"
 
     try:
@@ -185,7 +181,7 @@ def generate_podcast(
 
     audio_segments = []
     transcript = ""
-    crossfade_duration = 50
+    crossfade_duration = 50  # ms
 
     try:
         for item in script.dialogue:
@@ -204,7 +200,6 @@ def generate_podcast(
 
         final_mix = mix_with_bg_music(combined_spoken, custom_bg_music_path)
 
-        import tempfile
         with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as temp_audio:
             final_mix.export(temp_audio.name, format="mp3")
             final_mp3_path = temp_audio.name
@@ -299,7 +294,6 @@ def main():
     if "transcript_original" not in st.session_state:
         st.session_state["transcript_original"] = None
 
-    # For Q&A
     if "qa_count" not in st.session_state:
         st.session_state["qa_count"] = 0
     if "conversation_history" not in st.session_state:
@@ -367,6 +361,7 @@ def main():
             st.session_state["qa_count"] = 0
             st.session_state["conversation_history"] = ""
 
+    # Display generated audio and transcript if present
     if st.session_state["audio_bytes"]:
         st.audio(st.session_state["audio_bytes"], format='audio/mp3')
         st.download_button(
@@ -383,9 +378,8 @@ def main():
             height=300
         )
 
-        from difflib import SequenceMatcher
         def highlight_differences(original: str, edited: str) -> str:
-            matcher = SequenceMatcher(None, original.split(), edited.split())
+            matcher = difflib.SequenceMatcher(None, original.split(), edited.split())
             highlighted = []
             for opcode, i1, i2, j1, j2 in matcher.get_opcodes():
                 if opcode == 'equal':
@@ -449,8 +443,8 @@ def main():
                 st.markdown("### Updated Transcript")
                 st.markdown(new_transcript)
 
-        # ----------- POST-PODCAST Q&A with Microphone -----------
-        st.markdown("## Post-Podcast Q&A (Using Microphone)")
+        # ----------- POST-PODCAST Q&A (Text-based) -----------
+        st.markdown("## Post-Podcast Q&A (Text-based)")
 
         used_questions = st.session_state["qa_count"]
         remaining = MAX_QA_QUESTIONS - used_questions
@@ -458,66 +452,27 @@ def main():
         if remaining > 0:
             st.write(f"You can ask up to {remaining} more question(s).")
 
-            st.write("### Record Your Follow-Up Question:")
-
-            # EXPLICIT STUN SERVER
-            # So we can confirm ICE candidates are gathered
-            new_rtc_config = RTCConfiguration(
-                {
-                    "iceServers": [
-                        {"urls": ["stun:stun.l.google.com:19302"]}
-                    ]
-                }
-            )
-
-            webrtc_ctx = webrtc_streamer(
-                key="qna-audio-stream",
-                mode=WebRtcMode.SENDONLY,
-                rtc_configuration=new_rtc_config,  # <--- STUN server explicitly set
-                media_stream_constraints={"audio": True, "video": False},
-                audio_processor_factory=AudioBufferProcessor
-            )
-
-            if "audio-processor" not in st.session_state:
-                st.session_state["audio-processor"] = None
-
-            # If the stream is currently playing, store the processor
-            if webrtc_ctx.state.playing and webrtc_ctx.audio_processor:
-                st.session_state["audio-processor"] = webrtc_ctx.audio_processor
-
-                # Debug print: how many frames have arrived?
-                st.write("Frames so far:", webrtc_ctx.audio_processor.frame_count)
-
-            if not webrtc_ctx.state.playing:
-                st.write("Recording Stopped. You may now submit your question.")
+            typed_q = st.text_input("Type your follow-up question here:")
 
             if st.button("Submit Q&A"):
                 if used_questions >= MAX_QA_QUESTIONS:
                     st.warning("You have reached the Q&A limit.")
                 else:
-                    processor = st.session_state.get("audio-processor")
-                    if not processor or not getattr(processor, "frames", None):
-                        st.warning("No recorded audio found. Please record your question first.")
+                    question_text = typed_q.strip()
+                    if not question_text:
+                        st.warning("No question found. Please type something.")
                     else:
-                        local_wav_path = processor.finalize_wav()
-                        if not local_wav_path:
-                            st.warning("No audio frames found. Please record again.")
-                        else:
-                            st.write("Transcribing your voice question via Deepgram...")
-                            question_text = transcribe_audio_deepgram(local_wav_path)
-                            if not question_text.strip():
-                                st.warning("No transcript found. Please try again.")
-                            else:
-                                st.write(f"**You asked**: {question_text}")
-
-                                conversation_so_far = st.session_state["conversation_history"]
-                                ans_audio, ans_text = handle_qa_exchange(conversation_so_far, question_text)
-                                if ans_audio:
-                                    st.audio(ans_audio, format="audio/mp3")
-                                    st.markdown(f"**John**: {ans_text}")
-                                    st.session_state["qa_count"] += 1
-                                else:
-                                    st.warning("No response could be generated.")
+                        st.write(f"**You asked**: {question_text}")
+
+                        # We'll just store the question + a mock response for now
+                        # or you can do an LLM call
+                        # For example, let's do a minimal approach:
+                        fake_answer = "That's a great question! I'd love to answer, but I'm currently text-based only."
+                        st.write(f"**John**: {fake_answer}")
+
+                        # Update conversation
+                        st.session_state["conversation_history"] += f"\nUser: {question_text}\nJohn: {fake_answer}\n"
+                        st.session_state["qa_count"] += 1
         else:
             st.write("You have used all 5 Q&A opportunities.")