How-To Guides #

Task recipes for specific jobs. Each group below mirrors the same axis used by concepts and reference: the recipe you want is at how-to/<category>/<task>.md, the explanation is at concepts/<category>/<topic>.md, the lookup is at reference/<category>/<item>.md.

If you're new to Litmus, work through the tutorial first — these guides assume you can already run a simple test.

Overview #

Orientation map of the platform's surfaces.

• Tour of the Operator UI — what each of the 14 sidebar entries does, with cross-links to per-screen reference
• AI-assisted test development via MCP — register the MCP server with Claude Code, Cursor, Copilot, Cline, Claude Desktop

Configuration #

Get hardware on the bench and wired into a station.

• Configuring stations — station YAML, instruments, environments
• Custom instrument drivers — bring your own driver (PyVISA / PyMeasure / vendor)
• Mock mode — --mock-instruments, station mock_config, per-test mocks

Execution #

Author and run tests.

• Writing tests — pytest classes, sidecar YAML, the verify pattern
• Test limits — limit shapes, condition-indexed bands, comparator semantics
• Test vectors & sweeps — sidecar sweeps:, @parametrize, the vectors fixture
• Spec-driven testing — derive limits from the product YAML
• Read and write the test context — what the context fixture knows and how to use it from inside a test
• Profiles — named config sets — select which tests run and how
• Managing sessions — connect/disconnect lifecycle for instrument usage
• Multi-DUT testing — subprocess-per-slot, shared instruments
• Measurement traceability — ATML / IEEE 1671 metadata captured automatically
• Operator prompts — design guide for the litmus_prompts marker and the prompt fixture

Data #

Pull data out of the runtime — query, debug, export, dashboard.

• Querying historical events — MCP tool, HTTP API, Python
• Querying channel data — time-series data plane
• Find flaky tests — use Metrics → Retest and Results detail to spot intermittents
• Compare two runs — diff known-good vs failing with two tabs + a DuckDB query
• Export results — litmus show -f for reports (HTML/PDF/JSON/CSV) and litmus export -f for interchange (STDF/HDF5/TDMS/MDF4/ATML)
• Query runs and metrics via MCP — litmus_runs / litmus_steps / litmus_metrics recipes
• Debug failures via MCP — chained investigative workflow when a run fails
• Grafana dashboards — pre-built dashboards for results, events, channels

Catalog #

AI-assisted authoring against the capability catalog.

• Datasheet → tests with Claude Code — end-to-end walkthrough of the datasheet-to-test workflow