Update app.py

app.py

@@ -522,6 +522,9 @@ def setup_qa(db):
 
 #     return llm.invoke(prompt)
 
+import tempfile
+import os
+
 def process_answer(question, full_text):
     from langchain_community.document_loaders import TextLoader
     from langchain.text_splitter import RecursiveCharacterTextSplitter
@@ -530,49 +533,44 @@ def process_answer(question, full_text):
     from langchain.chains import RetrievalQA
     from langchain import HuggingFacePipeline
     from transformers import pipeline
-    import os
-    import shutil
 
-    # Save to temp file and load it as document
+    # Save the full_text to a temporary file
     with open("temp_text.txt", "w") as f:
         f.write(full_text)
 
     loader = TextLoader("temp_text.txt")
     docs = loader.load()
 
-    # Chunking the docs
+    # Chunk the documents
     text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=150)
     splits = text_splitter.split_documents(docs)
 
-    # Embeddings
+    # Load embeddings
     embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
 
-    # Clean up old DB if exists
-    if os.path.exists("chroma_db"):
-        shutil.rmtree("chroma_db")
+    # Create a temporary directory for ChromaDB
+    chroma_dir = os.path.join(tempfile.gettempdir(), "chroma_db")
+    if os.path.exists(chroma_dir):
+        import shutil
+        shutil.rmtree(chroma_dir)
 
-    db = Chroma.from_documents(splits, embeddings, persist_directory="chroma_db")
+    db = Chroma.from_documents(splits, embeddings, persist_directory=chroma_dir)
     retriever = db.as_retriever()
 
-    # Model pipeline
+    # Set up the model
     pipe = pipeline("text2text-generation", model="MBZUAI/LaMini-T5-738M", max_length=512)
     llm = HuggingFacePipeline(pipeline=pipe)
 
-    # Retrieval QA chain
-    qa_chain = RetrievalQA.from_chain_type(
-        llm=llm,
-        retriever=retriever,
-        return_source_documents=False
-    )
+    # RAG-style retrieval QA
+    qa_chain = RetrievalQA.from_chain_type(llm=llm, retriever=retriever)
 
-    # Check if question is about summarization
+    # Smart prompting
     if "summarize" in question.lower() or "summary" in question.lower() or "tl;dr" in question.lower():
-        prompt = f"Summarize the following document:\n\n{full_text[:3000]}"  # trimming to 3K chars for model
+        prompt = f"Summarize the following document:\n\n{full_text[:3000]}"
         summary = llm(prompt)
         return summary
     else:
-        answer = qa_chain.run(question)
-        return answer
+        return qa_chain.run(question)
 
 
 # --- UI Layout ---