See how Polaris renders a recipe and picks a model tier — live, in your browser, no server, no tokens spent.
Rendered prompt
{{diff}}, {{status}} and {{files}} are shown as a sample here —
Polaris fills them from your repo's live git when the recipe runs (via polaris recipe run
or the MCP prompt). That project-context rendering is what a static prompt library can't do.
Estimated complexity
Provider preference
Polaris picks the first provider that's actually available in your environment (after the
--local / --max-cost filters). The scorer is deterministic and free — no model call.