Skip to content

Sequence/Parallel filterset

Laurian Gridinoc requested to merge parallel-filterset into main

WIP fixes to deal with larger than LLM's context size filtersets request.

Includes:

  • create LLM call chain per query/filter pair
  • experimental parallel execution of the chains (works with OpenAI gpt-4o, needs fixes for Groq Llama 3.1)
  • experimental sequence execution of the chains (works with the in-house LLM endpoint running Hermes 2 Pro Mistral 8B)
  • system prompt with instructions
  • potential insert point for dynamic few-shot examples
Edited by Laurian Gridinoc

Merge request reports