Usage and rate limits - Namu API

Two limits

	Limit	Resets	When reached
Allowance	1,800 audio seconds (30 minutes) per account	The 1st of each month, UTC	`429 quota_exceeded`
Rate limit	60 requests per minute per key	Every minute	`429 rate_limited` with `Retry-After`

The allowance is shared by all the keys of an account; the rate limit applies to each key. What each request counts.

Check your usage

In the console, or from code:

curl https://api.namuai-lab.com/v1/usage \
  -H "Authorization: Bearer $NAMU_API_KEY"

usage = requests.get(
    "https://api.namuai-lab.com/v1/usage",
    headers={"Authorization": f"Bearer {os.environ['NAMU_API_KEY']}"},
).json()
left = usage["month_audio_seconds_limit"] - usage["month_audio_seconds_used"]
print(f"{left / 60:.1f} minutes left this month")

Response

{ "month_audio_seconds_used": 312.4, "month_audio_seconds_limit": 1800.0 }

/v1/usage needs a key (any scope), does not count against the rate limit, and may lag by up to 30 seconds.

Stay inside the limits

Cache what you repeat. The same sentence spoken twice costs twice; stored once, it costs once.
Send only speech. Trim long silences at the start and end of recordings before you send them: transcription counts the whole file.
Spread out batch work. Processing an archive? Send requests one after another rather than all at once, and honour Retry-After.
Watch the month. Check /v1/usage daily in production and warn yourself before the allowance runs out.

Headers

Header	On	Meaning
`X-Request-Id`	every response	A unique id for the request
`X-Audio-Seconds`	`/v1/speech` audio responses	Length of the audio returned
`Retry-After`	`429 rate_limited`	Seconds to wait before trying again

Need more for a pilot or a launch? Write to us with what you are building and the volume you expect.