Update app.py

app.py

@@ -74,20 +74,23 @@ if uploaded_file is not None:
     pdf_vectors = embedding.encode(texts)
     vector_count = len(documents)
     example_data_generator = map(lambda i: (f'id-{i}', pdf_vectors[i], {"text": texts[i]}), range(vector_count))
-    #if len(index.describe_index_stats()['namespaces'])!=0:
-    #    index.delete(delete_all=True)
+    # Update the Pinecone index with new vectors
+    if 'ns1' in index.describe_index_stats()['namespaces']:
+        index.delete(delete_all=True, namespace='ns1')
     for ids_vectors_chunk in chunks(example_data_generator, batch_size=100):
-        index.upsert(vectors=ids_vectors_chunk)
-        index.upsert(vectors=ids_vectors_chunk,namespace='ns1')
-
-# Search query related context
-sample_query = st.text_input("Stellen Sie eine Frage zu dem PDF: (Ask a question related to the PDF:)")
-if st.button("Abschicken (Submit)"):
+        index.upsert(vectors=ids_vectors_chunk, namespace='ns1')
+        index.upsert(vectors=ids_vectors_chunk) # creating a second namespace could fix the index0 bug!
+
+# Input for the search query
+with st.form(key='my_form'):
+    sample_query = st.text_input("Stellen Sie eine Frage zu dem PDF: (Ask a question related to the PDF:)")
+    submit_button = st.form_submit_button(label='Abschicken (Submit)')
+if submit_button:
     if uploaded_file is not None and sample_query:
         query_vector = embedding.encode(sample_query).tolist()
         query_search = index.query(vector=query_vector, top_k=5, include_metadata=True)
 
-        matched_contents = [match["metadata"]["text"] for match in query_search["matches"]]
+        matched_contents = [match["metadata"]["text"] for match in query_search["matches"], namespace='ns1']
 
         # Rerank
         rerank_model = "BAAI/bge-reranker-v2-m3"