ADAM

ADAM is the orchestration/runtime layer around ADAMAH for local LLM inference and diagnostics.

It includes:

• model loading and runtime profiles
• decode planning and execution paths
• profiling and diagnostics tooling
• chat/runtime integration utilities

Repository Layout

• adam/: Python runtime and model engine
• adamah-MAIN/: ADAMAH Vulkan compute backend
• tests/diagnostics/: validation and performance diagnostics
• tools/: helper scripts

Quick Start

python -m pip install -r requirements-runtime.txt

PYTHONUTF8=1 PYTHONPATH=. python -X utf8 tests/diagnostics/diag_inference.py gemma3-1b.gguf

Authorship and AI Assistance

• Concept, architecture, and product direction: Samuele Scuglia.
• Implementation support used during development: Claude, Gemini, and Codex.

License

• ADAM: LICENSE (CC BY-NC 4.0)
• ADAMAH backend: adamah-MAIN/LICENSE (CC BY-NC 4.0)