Skip to content

aster

An antenna design bench — evidence first.

aster is a Python library, a command-line tool and an agent skill for simulation-driven antenna studies. You describe a parameter space and what "good" means, the library drives candidates through a guarded AEDT session, and every solve leaves its artifacts behind. When the requirement changes, you re-judge the history instead of re-solving it.

aster

Three ways to use it

form entry point best for
Library / CLI pip install aster-<version>-py3-none-any.whl[optimizer,surrogate], then aster doctor scripting your own loop, CI, notebooks
Agent skill aster skill install or copy skills/aster/ an AI agent running design studies
Docs + releases this site + GitHub Releases learning the contract, pinned wheels
Runtime safety safety.md + the guard's threat model real AEDT runs, agent safety, fail-closed policy

Start in one minute

python -m venv .venv
.\.venv\Scripts\python.exe -m pip install -U pip
.\.venv\Scripts\python.exe -m pip install "aster[optimizer,surrogate]"   # from a wheel
aster version
aster dry-demo --samples 64

dry-demo runs a complete study on an analytic objective — no AEDT required.

The loop

design, solve, keep, judge, re-target

  • Design — ParamSpec / ParamSpace for the knobs; Objective / Constraint for the target.
  • Solve — real AEDT runs through the ansys-aedt-mcp guard; explicit PID/port, budgeted, auditable.
  • Keep — S11 Touchstone, far-field grids and native report images stay with the run.
  • Judge — acceptance rules are code, so verdicts can be recomputed and compared.
  • Re-target — changed rules re-score the archive; the solver is not called again.

Next

MIT licensed. If you build a study on this bench, keep the artifacts — a bench without evidence is just furniture.