Update app.py

app.py

@@ -20,23 +20,12 @@ from utils import (
     DialogueItem
 )
 from prompts import SYSTEM_PROMPT
-
-# NEW: For Q&A
+# Q&A
 from qa import transcribe_audio_deepgram, handle_qa_exchange
 
 MAX_QA_QUESTIONS = 5
 
 def parse_user_edited_transcript(edited_text: str, host_name: str, guest_name: str):
-    """
-    Looks for lines like:
-        **Angela**: Hello
-        **Dimitris**: Great topic...
-    We treat 'Angela' as the raw display_speaker, 'Hello' as text.
-    Then we map 'Angela' -> speaker='Jane' (if it matches host_name),
-    'Dimitris' -> speaker='John' (if it matches guest_name), etc.
-
-    Returns a list of DialogueItem.
-    """
     pattern = r"\*\*(.+?)\*\*:\s*(.+)"
     matches = re.findall(pattern, edited_text)
 
@@ -71,17 +60,11 @@ def parse_user_edited_transcript(edited_text: str, host_name: str, guest_name: s
     return items
 
 def regenerate_audio_from_dialogue(dialogue_items, custom_bg_music_path=None):
-    """
-    Re-generates multi-speaker audio from user-edited DialogueItems,
-    then mixes with background music or custom music.
-    Returns (audio_bytes, transcript_str).
-    """
     audio_segments = []
     transcript = ""
-    crossfade_duration = 50  # ms
+    crossfade_duration = 50
 
     for item in dialogue_items:
-        # Generate TTS for each line
         audio_file = generate_audio_mp3(item.text, item.speaker)
         seg = AudioSegment.from_file(audio_file, format="mp3")
         audio_segments.append(seg)
@@ -123,12 +106,6 @@ def generate_podcast(
     sponsor_style,
     custom_bg_music_path
 ):
-    """
-    Creates a multi-speaker podcast from PDF, URL, YouTube, or a research topic.
-    Ensures female voice for host (Jane), male voice for guest (John).
-    If sponsor_content is empty, we skip sponsor instructions entirely.
-    Returns (audio_bytes, transcript_str).
-    """
     sources = [bool(file), bool(url), bool(video_url), bool(research_topic_input)]
     if sum(sources) > 1:
         return None, "Provide only one input (PDF, URL, YouTube, or Topic)."
@@ -166,7 +143,6 @@ 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 = []
@@ -178,7 +154,7 @@ def generate_podcast(
     if user_specs.strip():
         extra_instructions.append(f"Additional User Instructions: {user_specs}")
 
-    # If sponsor_content is blank, skip sponsor instructions
+    # If user provided sponsor content, we pass it along; otherwise skip
     sponsor_instructions_present = False
     if sponsor_content.strip():
         extra_instructions.append(
@@ -189,14 +165,11 @@ def generate_podcast(
     from prompts import SYSTEM_PROMPT
     combined_instructions = "\n\n".join(extra_instructions).strip()
     full_prompt = SYSTEM_PROMPT
-
     if combined_instructions:
         full_prompt += f"\n\n# Additional Instructions\n{combined_instructions}\n"
 
     from utils import generate_script, generate_audio_mp3, mix_with_bg_music
     try:
-        # If sponsor is empty, we pass the sponsor_style anyway,
-        # but the model won't see sponsor instructions (since none were appended).
         script = generate_script(
             full_prompt,
             text,
@@ -204,7 +177,7 @@ def generate_podcast(
             f"{length_minutes} Mins",
             host_name=host_name or "Jane",
             guest_name=guest_name or "John",
-            sponsor_style=sponsor_style
+            sponsor_style=sponsor_style  # If sponsor is empty, no sponsor lines appended
         )
     except Exception as e:
         return None, f"Error generating script: {str(e)}"
@@ -273,7 +246,7 @@ def main():
         "conversational podcast.\n"
         "Select a tone and a duration range. The script will be on-topic, concise, and respect your chosen length.\n\n"
         "### How to use:\n"
-        "1. **Provide one source:** PDF Files, Website URL, YouTube videos, or a Topic to Research.\n"
+        "1. **Provide one source:** PDF Files, Website URL, YouTube link or a Topic to Research.\n"
         "2. **Choose the tone and the target duration.**\n"
         "3. **Click 'Generate Podcast'** to produce your podcast. After the audio is generated, "
         "   you can edit the transcript and re-generate the audio with your edits if needed.\n\n"
@@ -305,10 +278,7 @@ def main():
 
     user_specs = st.text_area("Any special instructions or prompts for the script? (Optional)", "")
     sponsor_content = st.text_area("Sponsored Content / Ad (Optional)", "")
-    sponsor_style = st.selectbox(
-        "Sponsor Integration Style",
-        ["Separate Break", "Blended"]
-    )
+    sponsor_style = st.selectbox("Sponsor Integration Style", ["Separate Break", "Blended"])
 
     custom_bg_music_file = st.file_uploader("Upload Custom Background Music (Optional)", type=["mp3", "wav"])
     custom_bg_music_path = None
@@ -411,7 +381,6 @@ def main():
                 st.session_state["transcript_original"],
                 edited_text
             )
-
             st.markdown("### **Edited Transcript Highlights**", unsafe_allow_html=True)
             st.markdown(highlighted_transcript, unsafe_allow_html=True)
 
@@ -459,9 +428,9 @@ def main():
                 st.markdown("### Updated Transcript")
                 st.markdown(new_transcript)
 
-        # -----------------------
-        # POST-PODCAST Q&A Logic
-        # -----------------------
+        # -------------------------------------------
+        # Post-Podcast Q&A using st.audio_input():
+        # -------------------------------------------
         st.markdown("## Post-Podcast Q&A")
         used_questions = st.session_state["qa_count"]
         remaining = MAX_QA_QUESTIONS - used_questions
@@ -470,8 +439,8 @@ def main():
             st.write(f"You can ask up to {remaining} more question(s).")
 
             typed_q = st.text_input("Type your follow-up question:")
-            # If on Streamlit >= 1.41.0, you could do st.audio_input for direct mic recordings
-            audio_q = st.file_uploader("Or upload an audio question (WAV, MP3)")
+            # Replacing file_uploader with st.audio_input (Streamlit >= 1.41)
+            audio_q = st.audio_input("Or record your voice question")
 
             if st.button("Submit Q&A"):
                 if used_questions >= MAX_QA_QUESTIONS:
@@ -479,11 +448,10 @@ def main():
                 else:
                     question_text = typed_q.strip()
                     if audio_q is not None:
-                        suffix = ".wav"
-                        with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tmp:
-                            tmp.write(audio_q.read())
+                        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
+                            tmp.write(audio_q.getvalue())
                             local_audio_path = tmp.name
-                        st.write("Transcribing your audio question...")
+                        st.write("Transcribing your audio question via Deepgram...")
                         audio_transcript = transcribe_audio_deepgram(local_audio_path)
                         if audio_transcript:
                             question_text = audio_transcript