Rate limits

Number of concurrent async jobs

There is a hard, client-specific limit of 500 concurrent async jobs at any given point in time. An async job consist of data upload, inventory classification, training and prediction. If this limit is reached, a 429 TOO_MANY_CONCURRENT_JOBS error will be thrown.

A way to avoid reaching this limit is to gather data from multiple datasets into one job instead of running one dataset per job. Additionally, it is important to ensure that a data upload job has finished before starting a training job for that dataset id. Likewise, the training job should finish before the prediction job begins.

Maximum data upload

The Upload Raw Data endpoint can receive up to 14 GB per request. If this limit is exceeded a 500 Internal Server Error error will be returned. We are working towards returning a more descriptive error in such cases.

Maximum payload in POST requests

In addition to this, there is a hard limit on a payload size for the POST requests of 6 MB. If this is exceeded you will receive a 413 Request too long or if the payload exceeds 10 MB 413 HTTP content length exceeded 10485760 bytes.