Get User Context

Max token size of returned Context

Max subtopic size of the same topic in returned Context

Set specific subtopic limits for topics in JSON, for example {"topic1": 3, "topic2": 5}. The limits in this param will override max_subtopic_size.

Profile event ratio of returned Context

Whether to require event summary in returned Context

Pass the recent chats to enable context search. Memobase will use those chats to search for relevant events. It's a list of chats in OpenAI Message format, for example: [{"role": "user", "content": "Hello"}, {"role": "assistant", "content": "Hi"}].

NOTICE

• It will increase your latency by 0.1-1 seconds, because Memobase will use Embedding to search for relevant profiles and events.
• It will cost your Memobase tokens, roughly 100~200 tokens per chat based on the profile size.
• The profiles in the context will not be searched by the chats_str.
• If you want also search profiles, see full_profile_and_only_search_event query parameter.

Event similarity threshold of returned Context

Only allow events within the past few days, default is 180

Customize context prompt template.

• use {profile_section} to refer to the profile section
• use {event_section} to refer to the event section

For example:

# Memory
Unless the user has relevant queries, do not actively mention those memories in the conversation.
## User Background:
{profile_section}

## Latest Events:
{event_section}

If you pass chats_str and set this to False, Memobase will search for relevant profiles and events at the same time. NOTICE

• It will increase your latency by 2-5(based on the profile size) seconds, because Memobase will use LLM and Embedding to search for relevant profiles and events.
• It will cost your Memobase tokens, roughly 100~1000 tokens per chat based on the profile size.

If set to True, Memobase will fill the token window with the rest events.