Skip to content
Arivie

Cost guardrails

When to use this

Section titled “When to use this”

Apply cost guardrails when you need hard ceilings on agent spend: fewer rows returned per execute, shorter SQL timeouts, and optional per-request token budgets before the model provider bills you. These limits are enforced inside @arivie/agent tools — not advisory prompts.

Architecture

Section titled “Architecture”

Context modes — pick preload for small catalogs to reduce browse/RAG cost

Base example: examples/with-nextjs/.

Config patch

Section titled “Config patch”

Add limits to defineArivie in arivie.config.ts:

defineArivie({
  // ... owner, model, db, semantic, resolveUser
  limits: {
    rowsPerQuery: 50,
    queryTimeoutMs: 30_000,
    tokensPerRequest: 8_000,
    tokensPerUserPerMonth: null,
  },
});

rowsPerQuery and queryTimeoutMs apply to execute and compile_metric. tokensPerRequest is reserved for handler-level enforcement as the budget wiring lands in Sprint 5 CI gates (REQ-32). Pair limits with semantic.mode: "preload" when your catalog is small to avoid browse/RAG token overhead.

Run it

Section titled “Run it”
Terminal window
cd arivie && pnpm install
pnpm --filter with-nextjs dev
curl -X POST http://localhost:3000/api/arivie -H 'Content-Type: application/json' -d '{"prompt":"How many customers?"}'

Stress-test with a wide SELECT prompt and confirm the agent respects rowsPerQuery in the returned result set.