Update app.py

app.py

@@ -9,7 +9,7 @@ from langchain.document_loaders import PDFPlumberLoader
 from langchain_experimental.text_splitter import SemanticChunker
 from langchain_huggingface import HuggingFaceEmbeddings
 from langchain_chroma import Chroma
-from prompts import rag_prompt, relevancy_prompt, relevant_context_picker_prompt, response_synth
+from prompts import rag_prompt
 
 # Set API Keys
 os.environ["GROQ_API_KEY"] = st.secrets.get("GROQ_API_KEY", "")
@@ -57,7 +57,7 @@ if pdf_source == "Upload a PDF file":
 
 elif pdf_source == "Enter a PDF URL":
     pdf_url = st.text_input("Enter PDF URL:", value="https://arxiv.org/pdf/2406.06998")
-    if pdf_url and st.session_state.pdf_path is None:
+    if pdf_url and not st.session_state.get("pdf_loaded", False):
         with st.spinner("Downloading PDF..."):
             try:
                 response = requests.get(pdf_url)
@@ -75,28 +75,31 @@ elif pdf_source == "Enter a PDF URL":
                 st.error(f"Error downloading PDF: {e}")
 
 # Step 2: Process PDF
-if st.session_state.pdf_path and not st.session_state.pdf_loaded:
+if st.session_state.pdf_path and not st.session_state.get("pdf_loaded", False):
     with st.spinner("Loading and processing PDF..."):
         loader = PDFPlumberLoader(st.session_state.pdf_path)
         docs = loader.load()
         st.session_state.documents = docs
-        st.session_state.pdf_loaded = True
+        st.session_state.pdf_loaded = True  # ✅ Prevent re-loading
         st.success(f"✅ **PDF Loaded!** Total Pages: {len(docs)}")
 
 # Step 3: Chunking
-if st.session_state.pdf_loaded and not st.session_state.chunked and st.session_state.documents:
+if st.session_state.get("pdf_loaded", False) and not st.session_state.get("chunked", False):
     with st.spinner("Chunking the document..."):
         model_name = "nomic-ai/modernbert-embed-base"
         embedding_model = HuggingFaceEmbeddings(model_name=model_name, model_kwargs={'device': 'cpu'}, encode_kwargs={'normalize_embeddings': False})
         text_splitter = SemanticChunker(embedding_model)
         documents = text_splitter.split_documents(st.session_state.documents)
-        st.session_state.documents = documents  # Store chunked docs
-        st.session_state.chunked = True
+        st.session_state.documents = documents  # ✅ Store chunked docs
+        st.session_state.chunked = True  # ✅ Prevent re-chunking
         st.success(f"✅ **Document Chunked!** Total Chunks: {len(documents)}")
 
 # Step 4: Setup Vectorstore
-if st.session_state.chunked and not st.session_state.vector_created:
+if st.session_state.get("chunked", False) and not st.session_state.get("vector_created", False):
     with st.spinner("Creating vector store..."):
+        model_name = "nomic-ai/modernbert-embed-base"
+        embedding_model = HuggingFaceEmbeddings(model_name=model_name, model_kwargs={'device': 'cpu'}, encode_kwargs={'normalize_embeddings': False})
+
         vector_store = Chroma(
             collection_name="deepseek_collection",
             collection_metadata={"hnsw:space": "cosine"},
@@ -106,13 +109,13 @@ if st.session_state.chunked and not st.session_state.vector_created:
         vector_store.add_documents(st.session_state.documents)
         num_documents = len(vector_store.get()["documents"])
         st.session_state.vector_store = vector_store
-        st.session_state.vector_created = True
+        st.session_state.vector_created = True  # ✅ Prevent re-creating vector store
         st.success(f"✅ **Vector Store Created!** Total documents stored: {num_documents}")
 
-# Step 5: Query Input
-if st.session_state.vector_created and st.session_state.vector_store:
+# Step 5: Query Input (this should not trigger previous steps!)
+if st.session_state.get("vector_created", False) and st.session_state.get("vector_store", None):
     query = st.text_input("🔍 Enter a Query:")
-    
+
     if query:
         with st.spinner("Retrieving relevant contexts..."):
             retriever = st.session_state.vector_store.as_retriever(search_type="similarity", search_kwargs={"k": 5})
@@ -130,4 +133,4 @@ if st.session_state.vector_created and st.session_state.vector_store:
             final_response = response_chain.invoke({"query": query, "context": context_texts})
 
         st.subheader("🟥 RAG Final Response")
-        st.success(final_response['final_response'])
+        st.success(final_response['final_response'])