Guide

The guide covers the runtime surface one feature at a time. The tutorials teach the ideas in order; these pages go deeper and stand alone, so read whichever one answers your question.

• Install and first steps installs the package and runs the first program.
• Atoms, operators, and terms builds every atom kind in Python, operators included.
• Where code runs separates staged term construction, live evaluation, and compiled Python bodies.
• Run and query explains run for source, eval for terms, and query for joins over facts.
• Python functions in MeTTa registers callables as MeTTa functions, with generators as nondeterminism.
• Write MeTTa in Python compiles Python function bodies into equations with @m.define.
• Spaces selects, creates, pools, and drops named spaces.
• The contract explains how backends attach by declaration: fidelity, sources, errors, writes, annotations, and explain.
• Threads, tasks, and pickling states, per type, what locks, what is atomic, and what crosses a boundary.
• Observability answers nine questions about a running program.
• Jupyter notebooks walks the executed notebook tour.
• Pettorch connects PyTorch through the public integration surface.

Stability follows one rule: a surface removal warns first. A name on its way out raises DeprecationWarning through warnings.warn(..., stacklevel=2) for one release before it goes, so a breaking change is always signaled a release ahead, and the changelog names it both times.