Skip to main content

LLM Routing Quickstart

Prerequisites​

  • BFF with /chat/completions enabled
  • PDP reachable with model/egress constraints
  • Pricing map and (optional) receipts/budget holds configured

Steps​

  1. Set pricing and budgets
export LLM_PRICING_JSON='{"gpt-4o-mini":{"in":0.00015,"out":0.0006},"gpt-4.1":{"in":0.0005,"out":0.0015}}'
export REDIS_URL=redis://redis:6379/0
  1. Call the endpoint
curl -sS -X POST "$BFF/api/chat/completions" \
  -H "Content-Type: application/json" \
  --data '{"model":"gpt-4.1","messages":[{"role":"user","content":"Summarize"}],"max_tokens":128}' -i

Validate​

  • x-aria-model-selected present
  • x-aria-model-rerouted: true|false
  • 402 = budget denial; 403 = policy denial