Currenty, we generate keywords from the conversation as a start point to classify the conversation. The classification of conversation can have several downstream uses, such as support allocation, prioritization of emergencies, and understanding what topics drive your support, etc.
LLMs are also good at classification, and we can take advantage of langchain's implementation for tagging documents. The ideia is to allow the user to specify a prompt and tags description, and then at each request an async task will tag the conversion as we already do here.
Currenty, we generate keywords from the conversation as a start point to classify the conversation. The classification of conversation can have several downstream uses, such as support allocation, prioritization of emergencies, and understanding what topics drive your support, etc.
LLMs are also good at classification, and we can take advantage of langchain's implementation for tagging documents. The ideia is to allow the user to specify a prompt and tags description, and then at each request an async task will tag the conversion as we already do here.