v0.2.1 — Rate limiting and budget-aware guards

Package: n8n-nodes-uno-q@0.2.1 (bridge unchanged at 0.2.0)

Added

• UnoQTool — Rate Limit collection (Max Calls + Per of minute / hour / day). Caps how often the AI Agent may invoke the tool; excess calls short-circuit with a structured rejection { refused: true, error: "Refused: rate limit of N per <window> exceeded. Retry in ~Xs." } that the LLM reads and can react to. The check runs before the Method Guard, and the call is recorded only after both gates pass — so guard-rejected calls do not consume rate-limit budget. Counters are a sliding window kept in-memory per n8n process (reset on container restart, not shared across queue-mode workers, which remains unsupported for the same reason the bridge singleton is).
• budget variable in Method Guard scope — a read-only view of the call history for traffic-aware policies. budget.used(window) returns prior successful calls in the last 'minute' | 'hour' | 'day' and works whether or not a Rate Limit is configured, so guards can implement soft caps without committing to hard enforcement. budget.remaining and budget.resetsInMs expose cap-aware state when the Rate Limit field is set (number / number) and are null otherwise — enabling patterns like priority reservation ("refuse low-priority params when remaining < 3"). Exposed via new Function('method', 'params', 'budget', <guard body>).

See packages/n8n-nodes/README.md#rate-limit-and-budget for the full contract and three worked patterns.