Added download button

Had to add a download button because the built in st.audio() portion is failing the download part.

hype_pack/streamlit_app.py

@@ -167,13 +167,10 @@ def main():
                             )
                             
                             st.session_state.interview_state = interview_state
-                            if interview_state.audio_bytes:
-                                st.audio(interview_state.audio_bytes, format='audio/mp3')
-                            
                             st.session_state.stage = 'results'
                             st.rerun()
 
-    # Results Stage
+        # Results Stage
     elif st.session_state.stage == 'results':
         if (st.session_state.interview_state and 
             st.session_state.interview_state.transcript):
@@ -184,7 +181,26 @@ def main():
                 # Audio player section
                 st.markdown("#### Listen to Your Hype Speech 🎧")
                 if st.session_state.interview_state.audio_bytes:
-                    st.audio(st.session_state.interview_state.audio_bytes, format='audio/mp3')
+                    # Create columns for audio player and download button
+                    col1, col2 = st.columns([3, 1])
+                    
+                    with col1:
+                        # Display audio player
+                        st.audio(
+                            st.session_state.interview_state.audio_bytes,
+                            format='audio/mp3'
+                        )
+                    
+                    with col2:
+                        # Add download button with unique filename
+                        unique_filename = f"hype_speech_{uuid.uuid4().hex[:8]}.mp3"
+                        st.download_button(
+                            label="💾 Download",
+                            data=st.session_state.interview_state.audio_bytes,
+                            file_name=unique_filename,
+                            mime="audio/mpeg",
+                            help="Download your hype speech as an MP3 file"
+                        )
                 
                 # Collapsible transcript
                 with st.expander("View Speech Transcript 📝", expanded=False):

hype_pack/utils/nodes.py

@@ -9,19 +9,23 @@ from dotenv import load_dotenv
 from lmnt.api import Speech
 import time
 import tempfile
+from langchain_core.tracers.context import tracing_v2_enabled
 
 load_dotenv()
 
-
+# At the top of your file, after imports
+os.environ["LANGCHAIN_TRACING_V2"] = "true"
+os.environ["LANGCHAIN_PROJECT"] = "hypecast_generator"
 
 def build_reference_material_node(interview_state: InterviewState) -> InterviewState:
     """
     Analyzes candidate background to generate material for motivational speeches.
     """
-    llm = ChatOpenAI(
-        model="gpt-4o-mini",
-        temperature=0.1
-    ).with_structured_output(ReferenceMaterial)
+    with tracing_v2_enabled(tags=["reference_material"]):
+        llm = ChatOpenAI(
+            model="gpt-4o-mini",
+            temperature=0.1
+        ).with_structured_output(ReferenceMaterial)
 
     prompt = ChatPromptTemplate.from_messages([
         ("system", """You are an expert at identifying compelling personal narratives 
@@ -75,54 +79,55 @@ def generate_questions_node(interview_state: InterviewState) -> InterviewState:
     """
     Generates questions and manages the question history.
     """
-    llm = ChatOpenAI(
-        model="gpt-4o-mini",
-        temperature=0.35
-    ).with_structured_output(QuestionList)
-
-    # Get existing question texts to avoid duplicates
-    existing_questions = set()
-    if interview_state.qa_history is None:
-        interview_state.qa_history = QuestionList(questions=[])
-    
-    for q in interview_state.qa_history.questions:
-        existing_questions.add(q.question_text)
-
-    prompt = ChatPromptTemplate.from_messages([
-        ("system", """Generate 2-3 focused questions that reveal what motivates 
-        this person. Each question should have 3 distinct choices."""),
-        ("human", """
-        Reference Material:
-        {reference_material}
-
-        Previous Questions Asked:
-        {previous_questions}
-
-        Create new questions that:
-        - Are different from previous questions
-        - Focus on motivation and confidence
-        - Connect to their background
-        """)
-    ])
-
-    new_questions = llm.invoke(prompt.format_messages(
-        reference_material=interview_state.reference_material,
-        previous_questions="\n".join([
-            f"Q: {q.question_text}" 
-            for q in (interview_state.qa_history.questions or [])
+    with tracing_v2_enabled(tags=["questions"]):
+        llm = ChatOpenAI(
+            model="gpt-4o-mini",
+            temperature=0.35
+        ).with_structured_output(QuestionList)
+
+        # Get existing question texts to avoid duplicates
+        existing_questions = set()
+        if interview_state.qa_history is None:
+            interview_state.qa_history = QuestionList(questions=[])
+        
+        for q in interview_state.qa_history.questions:
+            existing_questions.add(q.question_text)
+
+        prompt = ChatPromptTemplate.from_messages([
+            ("system", """Generate 2-3 focused questions that reveal what motivates 
+            this person. Each question should have 3 distinct choices."""),
+            ("human", """
+            Reference Material:
+            {reference_material}
+
+            Previous Questions Asked:
+            {previous_questions}
+
+            Create new questions that:
+            - Are different from previous questions
+            - Focus on motivation and confidence
+            - Connect to their background
+            """)
         ])
-    ))
-
-    # Filter out any duplicate questions and append new ones
-    unique_new_questions = [
-        q for q in new_questions.questions
-        if q.question_text not in existing_questions
-    ]
 
-    # Update qa_history, initializing if None
-    if interview_state.qa_history is None:
-        interview_state.qa_history = None
-    interview_state.qa_history.questions.extend(unique_new_questions)
+        new_questions = llm.invoke(prompt.format_messages(
+            reference_material=interview_state.reference_material,
+            previous_questions="\n".join([
+                f"Q: {q.question_text}" 
+                for q in (interview_state.qa_history.questions or [])
+            ])
+        ))
+
+        # Filter out any duplicate questions and append new ones
+        unique_new_questions = [
+            q for q in new_questions.questions
+            if q.question_text not in existing_questions
+        ]
+
+        # Update qa_history, initializing if None
+        if interview_state.qa_history is None:
+            interview_state.qa_history = None
+        interview_state.qa_history.questions.extend(unique_new_questions)
 
     return interview_state
 
@@ -130,10 +135,11 @@ def generate_transcript_node(interview_state: InterviewState, speaker_profile: d
     """
     Generates a concise, TTS-friendly motivational speech.
     """
-    llm = ChatOpenAI(
-        model="gpt-4o-mini",
-        temperature=0.6
-    ).with_structured_output(HypeCastTranscript)
+    with tracing_v2_enabled(tags=["transcript"]):
+        llm = ChatOpenAI(
+            model="gpt-4o-mini",
+            temperature=0.6
+        ).with_structured_output(HypeCastTranscript)
 
     prompt = ChatPromptTemplate.from_messages([
     ("system", f"""You are speaking directly TO the candidate about why they should be excited about THIS specific opportunity.