Hugging Face's logo

FerrellSyntheticIntelligence
/
aide-sovereign-workbench

Text Generation
offline
ide
local-ai
coding
llama.cpp
software-engineering
Model card Files
xet
 Community
aide-sovereign-workbench / models /PACKS.md
Ferrell Synthetic Intelligence
Add curated multi-model pack registry
47a24c7
preview code
|
Raw
History Blame Contribute Delete
991 Bytes

AIDE Model Packs

AIDE ships the registry for three optional public model packs. The application is model-neutral; users can install one, two, or all three depending on device memory.

Pack Role Approx. file License
SmolLM2 360M Q8_0 Fast chat/planning 386 MB Apache-2.0
Qwen2.5-Coder 0.5B Q4_K_M Autocomplete/light edits 491 MB Apache-2.0
Qwen2.5-Coder 1.5B Q4_K_M Main coding/build lane 1.12 GB Apache-2.0

The registry records the official source repository and expected file. Before a pack becomes ready, AIDE must verify the download checksum, load it through the selected runtime, query /v1/models, run a short generation smoke test, and record the result. A model that downloads successfully but fails the smoke test remains pending.

The user's unfinished Liquid model is intentionally not included. It can later be added as a separate pack after evaluation, model-card completion, and license confirmation.