# ModelBake

ModelBake is a free, Apache-2.0 local CLI for recording and comparing
Llama-to-GGUF model builds.

Website: https://modelbake.dev/
Real build record: https://modelbake.dev/real-evidence.html
Wheel: https://modelbake.dev/downloads/modelbake_ai-0.1.0-py3-none-any.whl
Source archive: https://modelbake.dev/downloads/modelbake_ai-0.1.0.tar.gz

## What it does

ModelBake records the local checkpoint, recipe, llama.cpp tools, generated
commands, output hashes, cache reuse and scoped runner results behind a GGUF
release. A later build can be compared with the previously accepted release.

## Quick start

Download the wheel, then run:

    python -m pip install ./modelbake_ai-0.1.0-py3-none-any.whl
    modelbake tour

The tour uses a deterministic private fixture format. It does not require a
model and does not claim GGUF or model-runtime compatibility.

## Current scope

- Python 3.11 or newer
- macOS and Linux
- local Llama-family safetensors checkpoints
- explicitly configured local llama.cpp tools
- GGUF F16, Q4_0 and Q8_0 targets
- no account and no model upload

## Evidence limits

ModelBake can identify recorded build inputs, commands and output bytes. It
does not prove model quality, numerical equivalence, safety, licence compliance,
future reproducibility or production readiness.