Guides¶

Each guide shows how to do one thing with a scenario or with the Python API.

Guide	What it covers
Working with AI assistants	The `homeostat` command line: describe what exists, validate a scenario with all problems at once, run it and read a summary; JSON output, errors and warnings, the schema.
Checking a run against what you meant	The `expect` section: statistics, shifts, saturation, settling and fault visibility stated in the scenario and checked with evidence by `homeostat diagnose`.
Measuring how well an assistant writes scenarios	The eval set: requests to write a scenario, scored by what the scenario does; the protocol, the measures, and the Claude Code skill.
Scenario files	The structure of a scenario: units, ports and wiring, measurements, custom operators, the initial state, output.
Changes and faults	Setpoint and mode changes, parameter changes, faults, ramps, and changes on some lanes only.
Plans, degradation and maintenance	Regimes, production transitions, slow degradation, maintenance tasks, and the plan's timeline.
Self-running plans	Production, maintenance and faults described as random processes and expanded into explicit plans.
Plant families	Distributions in scenarios, drawing plants, and running a family of plants as one batch.
Labels	When a fault becomes visible, how long loops take to settle, whether a plant is feasible, and where a sample's material came from.
Twin runs and lanes	Several lanes in one run, shared noise, and experiments that isolate one cause.
Multi-rate plants	Operators and controllers that update less often than the base step.
Using the core in Python	Building graphs from operators directly, and driving the engine step by step.