Closed bshikin closed 11 months ago
Please note: this PR is on hold for a better implementation of the context lookup.
The idea is to load known samples into an in-memory DB and only attach ones that match the query best.
No longer needed, https://github.com/ProjectUnifree/unifree/pull/47 is a more generic solution
The idea is to create a "fake" conversation that mimics GPT model learning various language rules. The rules are defined under:
Example LLM input looks like: