Update app.py

app.py

@@ -650,7 +650,101 @@ def create_zip_of_files(md_files, mp3_files, wav_files, input_question):
 # 7. MAIN AI LOGIC: LOOKUP & TAB HANDLERS
 # ─────────────────────────────────────────────────────────
 
-def perform_ai_lookup(
+
+def perform_ai_lookup(q, vocal_summary=True, extended_refs=False, 
+                     titles_summary=True, full_audio=False, useArxiv=False):
+    """Main routine that uses Anthropic (Claude) + Gradio ArXiv RAG pipeline."""
+    start = time.time()
+    ai_constitution = """
+    You are a medical and machine learning review board expert and streamlit python and html5 expert. You are tasked with creating a streamlit app.py and requirements.txt for a solution that answers the questions with a working app to demonstrate. You are to use the paper list below to answer the question thinking through step by step how to create a streamlit app.py and requirements.txt for the solution that answers the questions with a working app to demonstrate.
+    """
+
+    # --- 1) Claude API
+    client = anthropic.Anthropic(api_key=anthropic_key)
+    user_input = q
+    response = client.messages.create(
+        model="claude-3-sonnet-20240229",
+        max_tokens=1000,
+        messages=[
+            {"role": "user", "content": user_input}
+        ])
+    st.write("Claude's reply 🧠:")
+    st.markdown(response.content[0].text)
+
+    # Save & produce audio
+    result = response.content[0].text
+    create_file(q, result) 
+    md_file, audio_file = save_qa_with_audio(q, result)
+    st.subheader("📝 Main Response Audio")
+    play_and_download_audio(audio_file, st.session_state['audio_format'])
+
+
+    #useArxiv = st.checkbox("Search Arxiv for Research Answer", value=False)
+    if useArxiv:
+        # --- 2) Arxiv RAG
+        #st.write("Arxiv's AI this Evening is Mixtral 8x7B...")
+        st.write('Running Arxiv RAG with Claude inputs.')
+        client = Client("awacke1/Arxiv-Paper-Search-And-QA-RAG-Pattern")
+        refs = client.predict(
+            q, 
+            10, 
+            "Semantic Search", 
+            "mistralai/Mixtral-8x7B-Instruct-v0.1",
+            api_name="/update_with_rag_md"
+        )[0]
+
+        #r2 = client.predict(
+        #    q, 
+        #    "mistralai/Mixtral-8x7B-Instruct-v0.1", 
+        #    True, 
+        #    api_name="/ask_llm"
+        #)
+
+        # --- 3) Claude API with arxiv list of papers to app.py
+        client = anthropic.Anthropic(api_key=anthropic_key)
+        user_input = q + '\n\n' + 'Use the paper list below to answer the question thinking through step by step how to create a streamlit app.py and requirements.txt for the solution that answers the questions with a working app to demonstrate.'+ '\n\n'
+        response = client.messages.create(
+            model="claude-3-sonnet-20240229",
+            max_tokens=1000,
+            messages=[
+                {"role": "user", "content": user_input}
+            ])
+        r2 = response.content[0].text
+        st.write("Claude's reply 🧠:")
+        st.markdown(r2)
+        
+        #result = f"### 🔎 {q}\n\n{r2}\n\n{refs}"
+        result = f"🔎 {r2}\n\n{refs}"
+        md_file, audio_file = save_qa_with_audio(q, result)
+        st.subheader("📝 Main Response Audio")
+        play_and_download_audio(audio_file, st.session_state['audio_format'])
+
+        # --- 3) Parse + handle papers
+        papers = parse_arxiv_refs(refs)
+        if papers:
+            # Create minimal links page first
+            paper_links = create_paper_links_md(papers)
+            links_file = create_file(q, paper_links, "md")
+            st.markdown(paper_links)
+
+            # Then create audio for each paper
+            create_paper_audio_files(papers, input_question=q)
+            display_papers(papers, get_marquee_settings())
+            display_papers_in_sidebar(papers)
+        else:
+            st.warning("No papers found in the response.")
+
+    elapsed = time.time() - start
+    st.write(f"**Total Elapsed:** {elapsed:.2f} s")
+    return result
+
+
+
+
+
+
+
+def perform_ai_lookup_old(
     q,
     vocal_summary=True,
     extended_refs=False,
@@ -698,11 +792,6 @@ def perform_ai_lookup(
         end = time.time()
         st.write(f"**Elapsed:** {end - start:.2f}s")
 
-
-        # Try async run from top
-        st.markdown('# Async run:')
-        asyncio.run(process_voice_input(q + result_text))
-        
     return result_text
 
 async def process_voice_input(text):