Being able to have conversations with a documentation site is the future. Awesome that you are tackling it!
The other major LLM improvement I expect is multi-round code execution. Let the model iterate on bug fixing, like us. We need the extra steps, the LLM does too.