Update app.py

app.py

@@ -8,7 +8,7 @@ from trace import wandb_trace
 
 _ = load_dotenv(find_dotenv())
 
-RAG_BATCH = False # document loading, splitting, storage
+RAG_BATCH = True # document loading, splitting, storage
 
 config = {
     "chunk_overlap": 150,       # document splitting
@@ -29,6 +29,9 @@ def invoke(openai_api_key, rag_option, prompt):
         raise gr.Error("Retrieval Augmented Generation is required.")
     if (prompt == ""):
         raise gr.Error("Prompt is required.")
+
+    if (RAG_BATCH):
+        rag_batch(config)
     
     chain = None
     completion = ""
@@ -49,9 +52,6 @@ def invoke(openai_api_key, rag_option, prompt):
 
             llm_output = completion.llm_output
         else:
-            if (RAG_BATCH):
-                rag_batch(config)
-            
             completion, chain = rag_chain(config, openai_api_key, rag_option, prompt)
             result = completion["result"]
     except Exception as e: