Skip to content

Drive it from VS Code

A walkthrough of generating a mission in the editor: install the extension, point it at your engine, draft from a description, review the diff, and apply it. The full reference is the VS Code extension page; this is the end-to-end path.

Set up

  1. Install the engine into a Python environment, with the provider extra you intend to use:

    pip install "gmat-copilot[anthropic]"   # or [openai] / [ollama]; GitHub Models needs no extra

  2. Install the extension from the Marketplace or Open VSX. VS Code also installs the GMAT Script extension it depends on, which provides .script highlighting and on-type linting.

  3. Point the extension at the engine. Set gmatCopilot.server.pythonPath to the interpreter from step 1 (or put gmat-copilot-worker on your PATH), and make sure that environment has your provider credential (e.g. ANTHROPIC_API_KEY).

Pick a model

There is no default model. Run GMAT Copilot: Select the Provider and Model… from the Command Palette; the quick-pick lists only the providers your credentials can reach. Choosing one sets gmatCopilot.model to a provider:model selector — for example anthropic:claude-sonnet-4-6.

Draft a mission

Open (or create) a .script file so the GMAT Copilot commands are active, then run GMAT Copilot: Draft a Mission from a Description… and type a request:

A 500 km circular Earth orbit at 51.6 degrees inclination; propagate one day
and report altitude and semi-major axis.

The extension generates the script in the worker, lints it, and opens the result as a diff against the active file. Review it, then accept to replace the file's contents or dismiss to leave the file untouched. Nothing is written until you accept.

A handy variant: write the description as a comment in the file, select it, and run Draft a Mission from the Selection to use the selection as the prompt.

Read the diagnostics

Lint findings appear in the Problems panel and as inline squiggles, tagged gmat-copilot. In strict mode (the default) a draft that does not lint clean is not applied — the diagnostics tell you why. Turn off gmatCopilot.strict for permissive mode if you want the best-effort draft applied with its diagnostics attached.

After editing a script by hand, run GMAT Copilot: Re-validate the Active Script to refresh the findings without regenerating.

Close the loop in the editor

Two optional settings mirror the CLI's dynamic-validation flags:

  • gmatCopilot.repair — regenerate a failing draft up to N times, feeding the diagnostics back each round. The loop runs before the diff is shown, so you review the best draft it produced.
  • gmatCopilot.dryRun — after linting, load (and run, where a solver is present) the draft in a real GMAT to catch runtime errors a static parse cannot. This needs the [gmat] extra and a discoverable GMAT install in the worker's environment; its findings join the Problems panel.

Both report progress and can be cancelled mid-run; a cancelled draft leaves your file exactly as it was.

Next