Goal
Define a Llama Stack UI experience similar to the Llama API UI (from the LlamaCon keynote). This UI has some of the same building blocks as the IBM Cloud InstructLab UI we use for field POCs, but it is more flexible.

Persona
This new UI is not for an SME, it is for a more technical persona that has good understanding of the industry concepts of SDG/Training/Eval/Inference/Chat/Agents (not the InstructLab 1.x version of the concepts).

What problems are we trying to solve?
1. This persona may go to this UI and start a flow at any stage, can move from one step to the other while having access to the most relevant knobs for that step.

– They may bring a custom pipeline definition for the SDG
– They may bring a training dataset created offline and then run various training techniques

Possible training techniques
– fine tuning to learn knowledge,
– preference tuning,
– enabling reasoning in the model

2. This persona is looking for meaningful evaluations of the model
– can my model reason?
– how well?
– how well can my model use tools?
– how well has it learned the knowledge?

Resources
LlamaCon
LLama API
add resources here