code update

app.py

@@ -31,21 +31,25 @@ def split_text(text):
 
 # FAISS 벡터 저장소 생성
 def create_knowledge_base(chunks):
-    model_name = "halyn/gemma2-2b-it-finetuned-paperqa"
+    model_name = "sentence-transformers/all-mpnet-base-v2"  # 임베딩 모델을 명시
     embeddings = HuggingFaceEmbeddings(model_name=model_name)
     return FAISS.from_texts(chunks, embeddings)
 
 # Hugging Face 모델 로드
 def load_model():
-    model_name = "halyn/gemma2-2b-it-finetuned-paperqa"
-    tokenizer = AutoTokenizer.from_pretrained(model_name, clean_up_tokenization_spaces=False)
+    model_name = "halyn/gemma2-2b-it-finetuned-paperqa"  # 텍스트 생성 모델 사용
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
     model = AutoModelForCausalLM.from_pretrained(model_name)
     return pipeline("text-generation", model=model, tokenizer=tokenizer, max_new_tokens=150, temperature=0.1)
 
 # QA 체인 설정
 def setup_qa_chain():
     global qa_chain
-    pipe = load_model()
+    try:
+        pipe = load_model()  
+    except Exception as e:
+        print(f"Error loading model: {e}")
+        return
     llm = HuggingFacePipeline(pipeline=pipe)
     qa_chain = load_qa_chain(llm, chain_type="stuff")
     
@@ -78,19 +82,19 @@ def main_page():
 
                 st.text_area("Preview of extracted text", text[:1000], height=200)
                 st.write(f"Total characters extracted: {len(text)}")
-
+                global knowledge_base
                 if st.button("Proceed with this file"):
                     chunks = split_text(text)
-                    global knowledge_base
                     knowledge_base = create_knowledge_base(chunks)
                     
                     if knowledge_base is None:
                         st.error("Failed to create knowledge base.")
                         return
                     
+                    setup_qa_chain()
+
                     st.session_state.paper_name = paper.name[:-4]
                     st.session_state.page = "chat"
-                    setup_qa_chain()
                     st.success("PDF successfully processed! You can now ask questions.")
 
             except Exception as e:

requirements.txt

@@ -1,5 +1,6 @@
 streamlit
 PyPDF2
+langchain-huggingface
 langchain==0.3.1
 langchain-community==0.3.1
 langchain-core==0.3.8
@@ -10,5 +11,4 @@ faiss-cpu==1.8.0.post1
 requests==2.32.3
 huggingface-hub==0.25.1
 sentence-transformers==3.1.1
-peft==0.2.0
-langchain-huggingface
+peft==0.2.0