model loads twice sometimes wiht st.cache_resources

app.py

@@ -28,8 +28,6 @@ if "messages" not in st.session_state:
 if "doc" not in st.session_state:
     st.session_state.doc = None
 
-if "refresh" not in st.session_state:
-    st.session_state.refresh = True 
 if "refresh" not in st.session_state:
     st.session_state.refresh = True 
 
@@ -42,19 +40,18 @@ if prompt := st.chat_input("Search Here insetad of Google"):
     st.chat_message("user").markdown(prompt)
     st.session_state.messages.append({"role": "user", "content": prompt})
 
+    configs = st.session_state.configs
     if st.session_state.refresh:
         st.session_state.refresh = False 
         search(prompt)
 
     s = SemanticSearch(
         st.session_state.doc,
-        st.session_state.st.session_state.configs["model"]["embeding_model"],
-        st.session_state.st.session_state.configs["model"]["device"],
+        configs["model"]["embeding_model"],
+        configs["model"]["device"],
     )
     topk, u = s.semantic_search(query=prompt, k=32)
     output = st.session_state.model.answer_query(query=prompt, topk_items=topk)
-    topk, u = s.semantic_search(query=prompt, k=32)
-    output = st.session_state.model.answer_query(query=prompt, topk_items=topk)
     response = output
     with st.chat_message("assistant"):
         st.markdown(response)

model.py

@@ -68,13 +68,13 @@ class RAGModel:
 
 if __name__ == "__main__":
     configs = load_configs(config_file="rag.configs.yml")
-    query = "Explain F1 racing for a beginer"
-    g = GoogleSearch(query)
-    data = g.all_page_data
-    d = Document(data, 512)
-    doc_chunks = d.doc()
-    s = SemanticSearch(doc_chunks, "all-mpnet-base-v2", "mps")
-    topk, u = s.semantic_search(query=query, k=32)
+    query = "The height of burj khalifa is 1000 meters and it was built in 2023. What is the height of burgj khalifa"
+    # g = GoogleSearch(query)
+    # data = g.all_page_data
+    # d = Document(data, 512)
+    # doc_chunks = d.doc()
+    # s = SemanticSearch(doc_chunks, "all-mpnet-base-v2", "mps")
+    # topk, u = s.semantic_search(query=query, k=32)
     r = RAGModel(configs)
-    output = r.answer_query(query=query, topk_items=topk)
+    output = r.answer_query(query=query, topk_items=[""])
     print(output)

rag.configs.yml

@@ -4,5 +4,4 @@ document:
 model: 
   embeding_model: all-mpnet-base-v2
   genration_model: google/gemma-7b-it
-  device : cuda
-  
+  device : cuda