Tools for supporting very simple orchestration for AI workload, inspired by UNIX pipes, but with a Pythonic bent
Our philosophy with Arkestra is definitely not to wrap everything. There are many libraries out there for different capabilities needed in LLMOps pipelines.
We'll use the demos directory for a cookbook. Elements to incorporate:
- Uritemplate—Implementation of RFC 6570 URI Templates
- KinOS - AI Team Orchestration System
- mlops-python-package - "Kickstart your MLOps initiative with a flexible, robust, and productive Python package."
- RAG_Techniques (see also their Prompt_Engineering repo)
- Prompt Engineering Guide
- RAG All-in-one—Basically just links to other projects