The SimAPI Journal

SimAPI checks simulation output against physical laws before anyone trusts it. You send the results of a run — drag coefficients, stresses, temperatures, joint torques — and it returns a verdict: passed, warning, or failed, with the specific values and bounds that were violated.

It is not a solver and it does not replace engineering judgment. It is the automated gate that catches the obvious-in-hindsight failures — a diverged run, a unit-conversion slip, a saturated sensor, an efficiency above 1.0 — that otherwise slip into a design review or an ML dataset.

• Simulation engineers who run sweeps and want a fast sanity gate before sharing results.
• ML teams building surrogate models on simulation data, who need to exclude physically invalid trials from training sets.
• Platform / infra teams adding a validation step to a CI pipeline so bad runs fail the build instead of reaching production.
• Reviewers who want an auditable, reproducible second opinion rather than eyeballing spreadsheets.

1. Run your solver as usual and export results (CSV, JSON, VTK, NumPy, OpenFOAM).
2. Call POST /v1/validate — from a CI step, a data job, the CLI, or the in-browser validator on the dashboard.
3. Branch on the verdict: block the merge or exclude trials when the status is failed; surface warnings for review.
4. Feed only the training_ready trials into downstream design decisions or ML pipelines.

• Aerospace: validate CFD drag/lift sweeps against physical envelopes before they enter design review or a surrogate model.
• Robotics: gate controller simulations on joint-torque, tracking-error, and stability checks before hardware deployment.
• Automotive: screen aero and thermal runs nightly; block regressions from reaching the vehicle program.
• Energy: verify combustion, heat-exchanger, and structural results against conservation laws.
• Scientific computing: catch solver divergence and unit-conversion errors before results are published or reused.
• Digital twins: continuously validate live simulation streams so the twin never drifts from physical plausibility.

• Try it with zero setup: run a validation in the browser from the dashboard.
• Python: pip install simapi, then from simapi import SimAPI.
• Node/CLI: npm install -g simapi-cli, then simapi login and simapi validate run.json.
• Keep your key in the SIMAPI_API_KEY environment variable — never commit it. In CI, set it as a secret.
• Column names are normalized automatically (Cd, cd, and drag_coefficient all map to the same quantity), so you rarely need to rename anything.