Update app.py

app.py

@@ -51,20 +51,28 @@ def predict(message, history):
     question = message
     context = ""  # Add context if
     
-# Create a RetrievalQA instance
-chain = RetrievalQA.from_chain_type(
-    llm=llm,
-    chain_type="stuff",
-    retriever=retriever,
-    return_source_documents=True,
-    chain_type_kwargs={"prompt": QA_CHAIN_PROMPT}
-)
+    # Create a RetrievalQA instance
+    chain = RetrievalQA.from_chain_type(
+        llm=llm,
+        chain_type="stuff",
+        retriever=retriever,
+        return_source_documents=True,
+        chain_type_kwargs={"prompt": QA_CHAIN_PROMPT}
+    )
 
-# Execute the query
-result = chain({"query": question})
+    # Execute the query
+    result = chain({"query": question})
 
-# Stream the response
-partial_message = ""
-for chunk in result['result']:
-    partial_message += chunk
-    yield partial_message
+    # Stream the response
+    partial_message = ""
+    for chunk in result['result']:
+        partial_message += chunk
+        yield partial_message
+        
+gr.Interface(
+    fn=predict, 
+    inputs=[gr.inputs.Textbox(label="Your Message"), gr.inputs.State(label="History")],
+    outputs='text',
+    live=True,
+    allow_flagging="never"
+).launch()