- Do Assistants have the ability to reference past call logs to improve performance/context over time? Or is each call isolated, with no persistent memory? - If no persistent learning: is there a recommended architecture for simulating memory (e.g. storing transcripts in our own DB and injecting into the prompt/context on the next call)? - Is “AI smartness” something that’s expected to be solved via external integrations e.g. using n8n or Make and connecting to our own DB, or will Vapi introduce native conversation memory?

@A.G | Ant requires third party.

Yes, to echo kitten you will need some kind of data storing service that is separate from VAPI due to our data retention policies. There are a few Youtube videos on accomplishing this task.