Update app.py

app.py

@@ -118,6 +118,13 @@ class TogetherEmbeddings(Embeddings):
 def get_pdf_index():
     with st.spinner('📄 در حال پردازش فایل PDF...'):
         loader = [PyPDFLoader('test1.pdf')]
+        pages = loader.load()
+        full_text = "\n".join([page.page_content for page in pages])
+        text_splitter = RecursiveCharacterTextSplitter(
+            chunk_size=2048,  
+            chunk_overlap=256 
+        )
+        texts = text_splitter.split_text(full_text)
         embeddings = TogetherEmbeddings(
             model_name="togethercomputer/m2-bert-80M-8k-retrieval",
             api_key="0291f33aee03412a47fa5d8e562e515182dcc5d9aac5a7fb5eefdd1759005979"
@@ -126,10 +133,10 @@ def get_pdf_index():
 
 
         )
-        return VectorstoreIndexCreator(
-            embedding=embeddings,
-            text_splitter=RecursiveCharacterTextSplitter(chunk_size=300, chunk_overlap=0)
-        ).from_loaders(loader)
+
+        vectorstore = FAISS.from_texts(texts, embeddings)
+
+        return vectorstore
 
 index = get_pdf_index()
 
@@ -168,7 +175,7 @@ if st.session_state.pending_prompt:
         thinking = st.empty()
         thinking.markdown("🤖 در حال فکر کردن...")
 
-        response = chain.run(f'question:پاسخ را فقط به زبان فارسی جواب بده {st.session_state.pending_prompt}')
+        response = chain.run(f'لطفاً فقط به زبان فارسی پاسخ بده: {st.session_state.pending_prompt}')
         answer = response.split("Helpful Answer:")[-1].strip()
         if not answer:
             answer = "متأسفم، اطلاعات دقیقی در این مورد ندارم."