Skip to content

README.md

Latest commit

 

History

History
57 lines (37 loc) · 1.8 KB

README.md

File metadata and controls

57 lines (37 loc) · 1.8 KB

logo

GELLI \ˈgel-ē\ ● proper noun — fr Modern English gel ("to congeal") fr Old French gelee (“to congeal, frost”) fr Latin gelare (“to freeze”) fr root gelū (“frost”) ● Also known as the acronymn "General Edge Local Llama Instances" — but only in marketing for extreme techno-nerds.

  1. A deep freeze container for AI models without the Amazon rainforest in tow...
  2. Especially for use with AI in contrained spaces.
  3. Easily train LoRAs on your corpus without fuss.
  4. Intended to avoid leakage or other AI security mishaps.

Compile, quantize, bake, train, import and export AI artifact with the casual flair...

curl -fsSL https://github.com/frinknet/gelli/raw/main/install.sh | sh

Or you can install a different version (even develop or preview) by adding a version:

curl -fsSL https://github.com/frinknet/gelli/raw/main/install.sh | sh -s -- $VER

Once installed, keeping things up to date is easy by running: (Note you can change version numbers but it defaults to latest)

gelli update $VER

Managing Models & LoRAs

Internally GELLI manages its own models. It is really meant to be a general purpose runner for just about anything you can think of...

TODO - Write about DL from HF or OL or MR TODO - write about importting models and loras

Running your Models

TODO Explain prompt one shot TODO Explain serve models TODO Explain using aLoRAs

Creating Agentic Systems

TODO Explain agent workflow TODO Explain adding more tools TODO Explain adding agent memory TODO Explain adding interfaces TODO Explain adding endpoints

Developing your Own Models

TODO - Write about training LoRAs TODO - Write about testing your LoRA TODO - Write about merging into base models TODO - Write about exporting models and loras