DAN-51 Model warm up on start (#27)

Also added a minor prompt update

backend/danswer/direct_qa/qa_prompts.py

@@ -10,7 +10,8 @@ SYSTEM_ROLE = "You are a Question Answering system that answers queries based on
 
 BASE_PROMPT = (
     f"Answer the query based on provided documents and quote relevant sections. "
-    f"Respond with a json containing a concise answer and up to three most relevant quotes from the documents.\n"
+    f"Respond with a json containing a concise answer and up to three most relevant quotes from the documents. "
+    f"The quotes must be EXACT substrings from the documents.\n"
 )
 
 UNABLE_TO_FIND_JSON_MSG = (

backend/danswer/main.py

@@ -86,6 +86,15 @@ def get_application() -> FastAPI:
         RequestValidationError, validation_exception_handler
     )
 
+    @application.on_event("startup")
+    async def startup_event() -> None:
+        from danswer.semantic_search.semantic_search import (
+            warm_up_models,
+        )  # To avoid circular imports
+
+        warm_up_models()
+        logger.info("Semantic Search models are ready.")
+
     return application
 
 

backend/danswer/semantic_search/semantic_search.py

@@ -66,6 +66,11 @@ def get_default_reranking_model() -> CrossEncoder:
     return _RERANK_MODEL
 
 
+def warm_up_models() -> None:
+    get_default_embedding_model().encode("Danswer is so cool")
+    get_default_reranking_model().predict(("What is Danswer", "Enterprise QA"))  # type: ignore
+
+
 @log_function_time()
 def semantic_reranking(
     query: str,

backend/danswer/utils/text_processing.py

@@ -26,5 +26,6 @@ def shared_precompare_cleanup(text: str) -> str:
     text = text.replace(".", "")
     text = text.replace(":", "")
     text = text.replace(",", "")
+    text = text.replace("-", "")
 
     return text