🔥 Create a LLM From Our Documentation With Chatbot For Support

[jasonBirchall]

We've recently embarked on an innovative journey to enhance our support system by integrating a Language Learning Model (LLM) chatbot derived from our own documentation. The goal is to provide immediate, intelligent support by leveraging our internal knowledge base.

During our initial phase, we attempted to use bedrock as a foundation for our model. Our aim was to extend it with our documentation, essentially creating a tailored version of a foundational model. However, we encountered a significant roadblock: the advanced customisation and extension features we needed were not available in the region where we have bedrock enabled.

We also explored an alternative path by running a localgpt client, hoping to train and validate a model exclusively on our documentation. Unfortunately, the results fell short of our expectations and were, frankly, unusable.

The Path Forward:
This firebreak ticket is a call to action for innovative solutions and ideas.