deadline_exceeded

deadline_exceeded means a run did not finish within its deadline. Read this page when runs time out, to choose a better deadline_ms and to retry safely.

HTTP	`code`	Retryable
`504`	`deadline_exceeded`	Yes, with a longer `deadline_ms` or a smaller input, and a new `Idempotency-Key`.

What happened

Route: POST /v1/runs. Every run has a deadline, set with deadline_ms:

`deadline_ms`	Value
Default, decision run	30,000 plus 120,000 per 262,144 input tokens: about 30 seconds for a small state, 90 seconds for 128k tokens
Range, decision run	1 to 150,000; values outside are clamped, so `0` becomes `1` and `300000` becomes `150000`
Default, verdict run	`30000` (30 seconds)
Range, verdict run	1 to 120,000

The deadline covers choosing a model and the model call. When it passes in either step, the run fails with this code. The message always reads “the request deadline passed before a route was chosen”, even when the model call is what timed out. Do not use the message to tell the two apart. The run was admitted, then failed before the model served it. Your organization is not charged: the run’s hold is released. The run itself is not moved out of pending, and it stays pending. It is listed by GET /v1/runs and counted under in_flight in GET /v1/usage. Because of that, a replay with the same Idempotency-Key returns 202 with "state": "pending", never the answer. Retry with a new key.

How to fix

Raise deadline_ms, up to 120000. Set your HTTP client’s own timeout above it, or your client gives up first and you never see the answer.
Shrink the work. Trim the state, lower think_tokens, lower draws, or split a large question set across runs.
Retry with a new Idempotency-Key. deadline_ms is not part of the key’s identity, so the old key replays the pending run (202) instead of running with the new deadline.
If runs time out at the maximum deadline with a small input, retry later with backoff and report the request_id.

Example

{"error":{"code":"deadline_exceeded","message":"the request deadline passed before a route was chosen","request_id":"req_7d3f0c1a9b2e4f6a8c0d1e2f3a4b5c6d"}}

A retry with a longer deadline and a new key per attempt. Keep the client timeout above deadline_ms:

BODY='{"kind":"decision","state":{"ticket":"I was charged twice this month and nobody answers my emails."},"questions":{"urgent":{"type":"noul","instructions":"reply within the hour?"}},"max_output_tokens":16,"deadline_ms":60000}'

for attempt in 1 2 3 4; do
  # A new Idempotency-Key per attempt: the failed run stays pending under the old key.
  RESP=$(curl -sS -w '\n%{http_code}' https://api.opentype.dev/v1/runs \
    -H "Authorization: Bearer $OPENTYPE_API_KEY" \
    -H "Content-Type: application/json" \
    -H "Idempotency-Key: $(uuidgen)" \
    -d "$BODY")
  STATUS=$(printf '%s' "$RESP" | tail -n1)
  printf '%s\n' "$RESP" | sed '$d'
  case "$STATUS" in
    5??) sleep $((2 ** attempt)) ;;  # 500, 503, 504: back off, then try again
    *) break ;;                        # success or 4xx: stop
  esac
done

const body = {"kind":"decision","state":{"ticket":"I was charged twice this month and nobody answers my emails."},"questions":{"urgent":{"type":"noul","instructions":"reply within the hour?"}},"max_output_tokens":16,"deadline_ms":60000};

async function createRunWithRetry(payload: unknown, maxAttempts = 4) {
  for (let attempt = 1; ; attempt++) {
    const res = await fetch("https://api.opentype.dev/v1/runs", {
      method: "POST",
      headers: {
        Authorization: `Bearer ${process.env.OPENTYPE_API_KEY}`,
        "Content-Type": "application/json",
        // A new key per attempt: the failed run stays pending under the old key.
        "Idempotency-Key": crypto.randomUUID(),
      },
      body: JSON.stringify(payload),
    });
    const data = await res.json();
    if (res.ok) return data;

    const { code, message, request_id } = data.error;
    console.error(`POST /v1/runs -> ${res.status} ${code}: ${message} (${request_id})`);
    if (res.status < 500 || attempt >= maxAttempts) throw new Error(`${code} (${request_id})`);
    await new Promise((r) => setTimeout(r, 2 ** attempt * 1000));
  }
}

const run = await createRunWithRetry(body);

import os
import time
import uuid

import requests

body = {"kind":"decision","state":{"ticket":"I was charged twice this month and nobody answers my emails."},"questions":{"urgent":{"type":"noul","instructions":"reply within the hour?"}},"max_output_tokens":16,"deadline_ms":60000}

def create_run_with_retry(payload: dict, max_attempts: int = 4) -> dict:
    for attempt in range(1, max_attempts + 1):
        resp = requests.post(
            "https://api.opentype.dev/v1/runs",
            headers={
                "Authorization": f"Bearer {os.environ['OPENTYPE_API_KEY']}",
                # A new key per attempt: the failed run stays pending under the old key.
                "Idempotency-Key": str(uuid.uuid4()),
            },
            json=payload,
            timeout=160,  # longer than the largest deadline (150,000 ms)
        )
        data = resp.json()
        if resp.ok:
            return data

        err = data["error"]
        print(f"POST /v1/runs -> {resp.status_code} {err['code']}: {err['message']} ({err['request_id']})")
        if resp.status_code < 500 or attempt == max_attempts:
            raise RuntimeError(f"{err['code']} ({err['request_id']})")
        time.sleep(2 ** attempt)

run = create_run_with_retry(body)

Limits - deadline_ms and every other bound.
Polling - read a run later with GET /v1/runs/{run_id}.
Idempotency - when to reuse an Idempotency-Key and when to send a new one.
Error handling - a status-to-action table and a retry helper for every error.
Problem codes - every code, its status, and whether a retry can help.

Overview

Errors

Limits

Release notes

What happened

How to fix

Example

​What happened

​How to fix

​Example

​Related

What happened

How to fix

Example

Related