Assistant Retains Knowledge of Previous Conversati...
# support
v
Hi @SANAVA, I'm an AI assistant for Vapi. I'll analyze available sources and get back to you shortly.
s
@Sahil This is the call where I told the assistant my name is Emma: 7e8cefae-e4c7-4b00-8431-13db99a6e4a0 And this is the second call where it knew my name without me mentioning it: 5c803ed2-2db4-4d39-8587-cc029be206f1 This happens although the assistant is not connected to any other external source and does not store any information. Why does it have knowledge of previous conversations?
v
----------------------------- **While I generate an answer, here are some sources that might be helpful**: * [List Calls]() * [Get Assistant]() * [Prompting Guide]() * [Glossary]() -----------------------------
### Answer The behavior you're observing where the assistant retains knowledge from previous conversations typically relates to how the assistant's memory and context management are configured. Vapi does allow developers to manage the state and context between sessions, which can be used to create a more seamless and personalized user experience. In Vapi, this can be handled by storing certain information across calls or setting up the assistant to manage context more dynamically. The configuration details for managing conversation states or context are not directly covered in the provided documentation, but these capabilities generally depend on the design of the assistant and how it is programmed to handle session data. If you're looking to manage or modify how the assistant handles user data across different sessions, you might need to look into the assistant's configuration settings or the logic implemented for handling session data. This typically involves programming the assistant to either forget certain information after a call ends or to maintain it based on specific criteria. For more specific guidance on configuring these aspects, you might need to consult the specific implementation details of your assistant or the platform documentation for handling state and context.
Got it. Tagging @User for additional help.
m
During the call, there is a parameter like « semantic cache », this can be true or false. Default is true, If I remember well. I think this could be the thing
s
Set semanticCachingEnabled: False it will fix this issue.
As per I can see in the assistant logs this option is set to True
2 Views