OSAF Analysis Workflows

Use this skill when working with OpenStudio Server / OSAF analysis JSON, support ZIPs, sampling sweeps, optimization, calibration, or server smoke tests.

Defaults That Matter

• Omit output_variables when creating OSA JSON unless the user explicitly asks for a custom set — the tools then include the foundational OpenStudio Results outputs (EUI, site energy, peak demand, unmet hours, per-fuel end uses) with export+visualize enabled so OSAF plots work. Use openstudio_analysis_default_output_variables to inspect or extend the default payload.
• The OSA seed weather file is an EPW. DDY/STAT files are support files — never set one as the analysis weather file.

Discover Algorithms

For “list OSAF algorithms” or “what is this algorithm best for”, call:

openstudio_analysis_algorithms

Filter by category or analysis type when useful:

openstudio_analysis_algorithms category="sampling"
openstudio_analysis_algorithms category="lhs"

The tool returns analysis types, categories, best-fit use cases, caveats, typical algorithm keys, and the correct start sequence.

Algorithm Selection

Common choices:

Need	Prefer
Check server/config/package with one run	single_run
One-variable or general parameter sweep	lhs
Full-factorial style experiment with 2+ variables	doe
Screening many variables cheaply	morris
Formal variance-based sensitivity	sobol
Multi-objective Pareto calibration/design	nsga_nrel or spea_nrel
Continuous calibration/search	pso or rgenoud
Debugging repeatability or bounds	repeat_run or preflight

Do not use DOE for only one real variable. Use LHS for one-variable sweeps.

Validate Existing Configs

For existing OpenStudio-server/spec OSA JSON configs, call:

openstudio_analysis_validate_osa_json

Raw validation is schema-focused by default and accepts legacy server config patterns. Use require_foundational_measures=true when validating an MCP-generated analysis that is about to be packaged/submitted.

Create and Submit MCP Analyses

Create OSA JSON from measures:

openstudio_analysis_create_osa_json_from_measures

Before uploading a support ZIP:

openstudio_analysis_prepare_package
openstudio_analysis_validate_package

Submit and upload:

openstudio_analysis_submit

Generated analysis JSON files and support ZIP packages always include the foundational measures:

• view_model
• openstudio_results
• generic_qaqc

Default output variables are exported and visualized so OSAF plots can show OpenStudio Results whole-building metrics and electricity/natural gas end uses.

Start Order

OSAF sampled analyses need two start actions:

1. Start the algorithm action first, such as lhs, sobol, morris, pso, or nsga_nrel.
2. Start batch_run to simulate the generated datapoints.

Prefer:

openstudio_analysis_start_sampled_run

For a single-run smoke test, start single_run first, then batch_run.