Live end-to-end observability: distributed traces correlate frontend requests, streaming events, and database queries. SLO burn-rates, per-user budgets, and circuit breakers protect the system under load — and page you when something goes wrong.
Loading dashboard…
A concurrency-focused load test suite lives in scripts/load-test.ts. Run it against the preview or published URL to validate that backpressure, circuit breakers, and SLOs hold under realistic spikes:
# Smoke run (50 concurrent users, 30s) bun scripts/load-test.ts --url https://vellum-io.lovable.app --users 50 --duration 30 # Spike scenario (ramp to 500 users) bun scripts/load-test.ts --scenario spike --url https://vellum-io.lovable.app
Upgrade the Lovable Cloud instance from Backend → Advanced settings → Upgrade instance for more Postgres CPU, memory, and connection headroom — the single biggest lever once code-level optimizations are in place.
