Errors & limits

Status codes

Code	Meaning	Notes
401	Auth failed	Missing/invalid Bearer, or a session cookie sent to inference.
402	insufficient_balance	Confirmed USDC balance too low; top up.
404	Model not found	Use a live model id from GET /v1/models.
429	buyer_concurrency_limit	One in-flight completion per wallet across all keys.
500	Settlement failed	Fail-closed; no successful body is returned.
502	Upstream error	The provider returned an error.
503	No eligible supply	no_eligible_supply, insufficient_routable_at_price, or pool_empty.

503 body

{
  "error": {
    "code": "no_eligible_supply",
    "message": "...",
    "buyerMaxCents": 80,
    "cheapestPostedCents": 75,
    "routableDiemAboveMax": "12.500000"
  }
}

If the cheapest posted price is above your max, raise your max price to access that supply.

Limits

• Concurrency: one in-flight completion per wallet (429). Serialize or retry with backoff.
• Per-key forwarding: one in-flight forward per supplier key; the router falls through or waits.
• Streaming: not supported in v1 (400 on stream: true).
• Top-up: $1 minimum, $1,000 maximum.

Retry guidance

• 429: back off and retry; do not parallelize completions on one wallet.
• 402: top up, wait for confirmation, retry.
• 503: retry shortly (supply is perishable and refreshes), or raise your max price.
• 500 / 502: safe to retry; no charge is applied on a failed settlement.