Skip to main content
Yomira jobs run asynchronously. Create a simulation, receive a simulation_id, then poll until the job is completed. This keeps agent workflows stable even when the simulation takes longer than a normal request.

​
Simulation size

Use smaller runs while drafting and larger runs when the artifact is close to real publication.

​
Timeout behavior

If the initial request returns 202 Accepted, the job is queued correctly. Poll:
If a client or AI agent times out, do not restart the simulation immediately. Poll the existing simulation_id first.

​
Cost control

When testing:
  • start with target_n: 40
  • limit max_agent_voices
  • export Markdown only after completion
  • simulate fewer options first, then expand
When making a real decision:
  • include the exact artifact
  • include richer context
  • simulate every option that could actually be chosen
  • use a larger target_n only after the options are concrete

​
When self-serve is not enough

Self-serve simulations are context-enriched synthetic simulations. For high-stakes enterprise work, use a grounded pilot where Yomira builds the audience dataset from customer notes, interviews, reviews, CRM context, social data, or market source material.