Merge pull request #241 from nulib/deploy/staging

Deploy to production

chat/src/handlers/chat.py

@@ -13,7 +13,8 @@
 RESPONSE_TYPES = {
     "base": ["answer", "ref"],
     "debug": ["answer", "attributes", "azure_endpoint", "deployment_name", "is_superuser", "k", "openai_api_version", "prompt", "question", "ref", "temperature", "text_key", "token_counts"],
-    "log": ["answer", "deployment_name", "is_superuser", "k", "openai_api_version", "prompt", "question", "ref", "size", "source_documents", "temperature", "token_counts"]
+    "log": ["answer", "deployment_name", "is_superuser", "k", "openai_api_version", "prompt", "question", "ref", "size", "source_documents", "temperature", "token_counts"],
+    "error": ["question", "error", "source_documents"]
 }
 
 def handler(event, context):
@@ -37,7 +38,12 @@ def handler(event, context):
         config.setup_llm_request()
         response = Response(config)
         final_response = response.prepare_response()
-        config.socket.send(reshape_response(final_response, 'debug' if config.debug_mode else 'base'))
+        if "error" in final_response:
+            logging.error(f'Error: {final_response["error"]}')
+            config.socket.send({"type": "error", "message": "Internal Server Error"})
+            return {"statusCode": 500, "body": "Internal Server Error"}
+        else:
+            config.socket.send(reshape_response(final_response, 'debug' if config.debug_mode else 'base'))
 
     log_group = os.getenv('METRICS_LOG_GROUP')
     log_stream = context.log_stream_name

chat/src/helpers/prompts.py

@@ -2,7 +2,7 @@
 
 
 def prompt_template() -> str:
-    return """Please provide a brief answer to the question based on the documents provided. Include specific details from the documents that support your answer. Keep your answer concise. Each document is identified by a 'title' and a unique 'source' UUID:
+    return """Please provide a brief answer to the question based on the documents provided. Include specific details from the documents that support your answer. Keep your answer concise and keep reading time under 45 seconds. Each document is identified by a 'title' and a unique 'source' UUID. If the documents do not answer the question, please respond that you can't seem to find enough information about [keyword] that in the Digital Collection and suggest they search NUL's catalog and construct a search link using this format: https://search.library.northwestern.edu/discovery/search?field=any&query=any,contains,keyword&institution=01NWU&vid=01NWU_INST:NULVNEW&search_scope=MyInst_and_CI&tab=Everything&mode=Basic&displayMode=full&bulkSize=10&highlight=true&dum=true&displayField=all&pcAvailabiltyMode=true&facet=rtype,exclude,reviews,lk:
 
     Documents:
     {context}

chat/test/helpers/test_metrics.py

@@ -99,10 +99,10 @@ def test_token_usage(self):
 
         expected_result = {
             "answer": 12,
-            "prompt": 314,
+            "prompt": 462,
             "question": 5,
             "source_documents": 527,
-            "total": 858
+            "total": 1006
         }
 
         self.assertEqual(result, expected_result)