update settings

app.py

@@ -48,41 +48,43 @@ st.markdown(" ")
 # Sidebar for settings
 with st.sidebar:
     st.header("Settings")
-    # with st.container(border=True):
-    # Embedding model
-
-    model_name = st.selectbox("Choose content embedding model", [
-        "text-embedding-3-small",
-        # "text-embedding-3-large",
-        # "all-MiniLM-L6-v2", 
-        # "all-mpnet-base-v2"
-    ], 
-    # help="""
-    # Select the embedding model to use for encoding the retrieved text data. 
-    # Options include OpenAI's `text-embedding-3` models and two widely
-    # used SentenceTransformers models.
-    # """
-    )
     
-    with st.container(border=True):
-        st.write('**Video lectures**')
-        yt_token_choice = st.select_slider("Token per content", [256, 512, 1024], value=256, help="Larger values lead to an increase in the length of each retrieved piece of content", key="yt_token_len")
-        yt_chunk_tokens = yt_token_choice
-        yt_max_content = {128: 32, 256: 16, 512: 8, 1024: 4}[yt_chunk_tokens]
-        top_k_YT = st.slider("Number of relevant content pieces to retrieve", 0, yt_max_content, 4, key="yt_token_num")
-        yt_overlap_tokens = yt_chunk_tokens // 4
-
-    # st.divider()
-    with st.container(border=True):
-        st.write('**Textbook**')
-        show_textbook = False
-        # show_textbook = st.toggle("Show Textbook Content", value=False)
-        latex_token_choice = st.select_slider("Token per content", [128, 256, 512, 1024], value=256, help="Larger values lead to an increase in the length of each retrieved piece of content", key="latex_token_len")
-        latex_chunk_tokens = latex_token_choice
-        latex_max_content = {128: 32, 256: 16, 512: 8, 1024: 4}[latex_chunk_tokens]
-        top_k_Latex = st.slider("Number of relevant content pieces to retrieve", 0, latex_max_content, 4, key="latex_token_num")
-        # latex_overlap_tokens = latex_chunk_tokens // 4
-        latex_overlap_tokens = 0
+    with st.expander('Embedding model',expanded=True):
+        # with st.container(border=True):
+        # Embedding model
+
+        model_name = st.selectbox("Choose content embedding model", [
+            "text-embedding-3-small",
+            # "text-embedding-3-large",
+            # "all-MiniLM-L6-v2", 
+            # "all-mpnet-base-v2"
+        ], 
+        # help="""
+        # Select the embedding model to use for encoding the retrieved text data. 
+        # Options include OpenAI's `text-embedding-3` models and two widely
+        # used SentenceTransformers models.
+        # """
+        )
+        
+        with st.container(border=True):
+            st.write('**Video lectures**')
+            yt_token_choice = st.select_slider("Token per content", [256, 512, 1024], value=256, help="Larger values lead to an increase in the length of each retrieved piece of content", key="yt_token_len")
+            yt_chunk_tokens = yt_token_choice
+            yt_max_content = {128: 32, 256: 16, 512: 8, 1024: 4}[yt_chunk_tokens]
+            top_k_YT = st.slider("Number of relevant content pieces to retrieve", 0, yt_max_content, 4, key="yt_token_num")
+            yt_overlap_tokens = yt_chunk_tokens // 4
+
+        # st.divider()
+        with st.container(border=True):
+            st.write('**Textbook**')
+            show_textbook = False
+            # show_textbook = st.toggle("Show Textbook Content", value=False)
+            latex_token_choice = st.select_slider("Token per content", [128, 256, 512, 1024], value=256, help="Larger values lead to an increase in the length of each retrieved piece of content", key="latex_token_len")
+            latex_chunk_tokens = latex_token_choice
+            latex_max_content = {128: 32, 256: 16, 512: 8, 1024: 4}[latex_chunk_tokens]
+            top_k_Latex = st.slider("Number of relevant content pieces to retrieve", 0, latex_max_content, 4, key="latex_token_num")
+            # latex_overlap_tokens = latex_chunk_tokens // 4
+            latex_overlap_tokens = 0
 
     st.write(' ')
     with st.expander('Expert model', expanded=False):
@@ -220,8 +222,7 @@ if submit_button_placeholder.button("AI Answer", type="primary"):
             initial_max_k = int(0.1 * context_embeddings_YT.shape[0])
             idx_YT = fixed_knn_retrieval(question_embedding, context_embeddings_YT, top_k=top_k_YT, min_k=0)
             idx_Latex = fixed_knn_retrieval(question_embedding, context_embeddings_Latex, top_k=top_k_Latex, min_k=0)
-
-        with st.spinner("Answering the question..."):
+            
             relevant_contexts_YT = sorted([text_data_YT[i] for i in idx_YT], key=lambda x: x['order'])
             relevant_contexts_Latex = sorted([text_data_Latex[i] for i in idx_Latex], key=lambda x: x['order'])
 
@@ -250,6 +251,7 @@ if submit_button_placeholder.button("AI Answer", type="primary"):
                 for context_item in contexts:
                     context += context_item['text'] + '\n\n'
             
+        with st.spinner("Answering the question..."):
             #-------------------------
             # getting expert answer
             #-------------------------
@@ -301,11 +303,11 @@ if submit_button_placeholder.button("AI Answer", type="primary"):
             #-------------------------
             if st.session_state.synthesis_model in ["LLaMA-3.2-3B", "LLaMA-3.2-11B"]:
 
-                if st.session_state.expert_model == "LLaMA-3.2-11B":
+                if st.session_state.synthesis_model == "LLaMA-3.2-11B":
                     model_s = st.session_state.llama_model
                     tokenizer_s = st.session_state.llama_tokenizer
                     
-                elif st.session_state.expert_model == "LLaMA-3.2-3B":
+                elif st.session_state.synthesis_model == "LLaMA-3.2-3B":
                     model_s = st.session_state.llama_model_3B
                     tokenizer_s = st.session_state.llama_tokenizer_3B