Jobs and polling

Generation, revision, and async publishing take longer than a request should. They return 202 with a job id; you poll the job until it lands.

The job model

Any endpoint that starts long work responds immediately with a job. The job carries status, and on success, the result the work produced (the generated post, the revision outcome, the publish receipt).

Polling

curl https://api.taka.ai/v1/public/jobs/JOB_ID \
  -H "Authorization: Bearer taka_live_..."

• Poll every 10-30 seconds.
• Static images usually land in 1-3 minutes.
• Reels can take up to 20 minutes.

Timeouts

Stuck jobs fail themselves with code timeout, so a polling loop always terminates. You never need a client-side give-up timer to avoid hanging forever.

Honesty flags

Revision results tell you what actually happened, not what was attempted:

• revisionApplied: false means the quality floor kept the old version.
• noVisibleChange means the model returned the same words.
• imageryReverted means the photo stayed as it was.