Hello, I want to create a chatbot. This bot will work as a salesperson and have long conversations with customers. It needs to have memory, but the problem is that all the sources I see do it using "prompt" and we reach the token limit after a certain time. Is there a better way? Thank you
#Which langchain memory best for customer selling agent bot?
simple nexus
wicked ibex
Hello, I want to create a chatbot. This bot will work as a salesperson and have ...
Remove previous messages based on tokens
simple nexus
but i needd all history i found redus database solutions but still not sure
wicked ibex
but i needd all history i found redus database solutions but still not sure
Well you have 2 options: manage it through a token count or use a vector database for storing the conversation history but that has the downside of needing the context to match the query
The model is stateless and you can only fit a certain amount of tokens into the prompt
There is no way around it