Closed rchan26 closed 12 months ago
Need to consider whether or not this constitutes a new class, or we can add functionality within LlamaIndex
to specify whether a chat engine is preferred over a query. It feels like the latter would be nicer to use, but I wonder if it might get messy...
For the multiple chat instances. When we receive a message from Slack, we do also get their user id, so this might be useful
From #64 and #65, we have been able to replace the LLM with a quantized version llama-2 which uses far less memory than the standard and still seems to provide good performance. Further, we have looked at implementing a chat engine variant (rather than a query engine) of the model. Lets add this model to slack.