diff --git a/src/suql/agent.py b/src/suql/agent.py index 7c836c1..1cb8f69 100644 --- a/src/suql/agent.py +++ b/src/suql/agent.py @@ -230,7 +230,7 @@ def parse_execute_sql(dlgHistory, user_query, prompt_file="prompts/parser_suql.p """ generated_suql, generated_sql_time = llm_generate( template_file=prompt_file, - engine="gpt-3.5-turbo-0613", + engine="gpt-3.5-turbo-0125", stop_tokens=["Agent:"], max_tokens=300, temperature=0, @@ -362,7 +362,7 @@ def convert_opening_hours_query(opening_hours_query): response, _ = llm_generate( "prompts/opening_hours.prompt", {"opening_hours_query": opening_hours_query}, - engine="gpt-3.5-turbo-0613", + engine="gpt-3.5-turbo-0125", max_tokens=200, temperature=0.0, stop_tokens=["\n"], @@ -407,7 +407,7 @@ def compute_next_turn( continuation, first_classification_time = llm_generate( template_file="prompts/if_db_classification.prompt", prompt_parameter_values={"dlg": dlgHistory}, - engine="gpt-3.5-turbo-0613", + engine="gpt-3.5-turbo-0125", max_tokens=50, temperature=0.0, stop_tokens=["\n"], @@ -436,7 +436,7 @@ def compute_next_turn( response, final_response_time = llm_generate( template_file="prompts/yelp_response_no_results.prompt", prompt_parameter_values={"dlg": dlgHistory}, - engine="gpt-3.5-turbo-0613", + engine="gpt-3.5-turbo-0125", max_tokens=400, temperature=0.0, stop_tokens=[], @@ -456,7 +456,7 @@ def compute_next_turn( response, final_response_time = llm_generate( template_file="prompts/yelp_response_SQL.prompt", prompt_parameter_values={"dlg": dlgHistory}, - engine="gpt-3.5-turbo-0613", + engine="gpt-3.5-turbo-0125", max_tokens=400, temperature=0.0, stop_tokens=[], diff --git a/src/suql/free_text_fcns_server.py b/src/suql/free_text_fcns_server.py index 3d5b810..71b59b0 100644 --- a/src/suql/free_text_fcns_server.py +++ b/src/suql/free_text_fcns_server.py @@ -67,7 +67,7 @@ def _answer( return {"result": continuation} def start_free_text_fncs_server( - host="127.0.0.1", port=8500, k=5, max_input_token=3800, engine="gpt-3.5-turbo-0613" + host="127.0.0.1", port=8500, k=5, max_input_token=3800, engine="gpt-4o-mini" ): """ Set up a free text functions server for the free text