History & Query API

FFAI maintains four parallel history stores:

Query methods

ffai.get_all_interactions()
ffai.get_latest_interaction()
ffai.get_latest_interaction_by_prompt_name("analysis")
ffai.get_last_n_interactions(5)
ffai.get_interaction(sequence_number=3)
ffai.get_model_interactions("mistral-small-latest")
ffai.get_interactions_by_prompt_name("summary")
ffai.get_prompt_history()
ffai.get_response_history()
ffai.get_model_usage_stats()
ffai.get_prompt_name_usage_stats()
ffai.get_prompt_dict()
ffai.get_latest_responses_by_prompt_names(["a", "b"])
ffai.get_formatted_responses(["analysis", "summary"])

DataFrame export

df = ffai.history_to_dataframe()
df = ffai.clean_history_to_dataframe()
df = ffai.ordered_history_to_dataframe()
df = ffai.search_history(text="error", model="gpt-4o")
df = ffai.get_model_stats_df()
df = ffai.get_prompt_name_stats_df()
df = ffai.get_response_length_stats()
df = ffai.interaction_counts_by_date()

Persistence

ffai.persist_all_histories()  # write Parquet to configured directory

Client conversation history

ffai.get_client_conversation_history()
ffai.set_client_conversation_history(history)
ffai.add_client_message("user", "Hello")
ffai.clear_conversation()