GUI

The GUI is a Vite and React application in gui/ that talks only to the HTTP API. Build it once, then simulacra serve serves it from gui/dist next to the API:

bun run build:gui
bun run simulacra serve --data .

--data . points the server at a directory whose runs/ and audits/ hold what the CLI produced; the default is ./simulacra-data. Until the GUI is built, / answers with a plain-text note and the API is available regardless. For development, bun run dev:gui starts Vite and bun run dev:mock a mock API.

The GUI has three pages.

Runs

Runs and audits with their progress, and a form that starts a run: pick a built-in example or paste YAML, set the seed (default 1), optionally the number of ticks (the scenario's total by default) and the provider, mock or llm. Submitting posts to /api/runs; the new run appears in the list and streams its events.

Run

One run: the tick timeline with failure counts, the force-directed network coloured by stance, an agent inspector that shows the causal chain from observation through prompt to decision and effects, the metric series, and the integrity and cost panels. The inspector reads the same chain simulacra inspect prints.

Audit

One audit: the plan and its evidence grade, the perturbation axes, the conditions, the pairwise tests with Holm-corrected p values, the sensitivity ranking, direction consistency, cross-model means, distribution tests, integrity and cost. The offline HTML report of the same audit is served at /api/audits/:id/report.html; the audit reference explains each section.