Skip to main content

Performance and Costs

Use this guide to balance output quality, run time, and provider spend.

What Drives Cost​

Primary drivers:

  • Number of requested outputs (contentTargets counts)
  • Model choice for text and image generation
  • Retries caused by transient provider failures
  • Higher token budgets (maxTokens) for long outputs

You can inspect per-run totals in generation.analytics.json.

What Drives Runtime​

Primary drivers:

  • Total output count and mix of content types
  • Primary content type (article primary uses structured flow; non-article primary uses generic flow)
  • Image model speed and retry behavior
  • Network latency and provider backoff windows

Article-primary runs execute structured article stages; non-article-primary runs execute generic primary stages and still render one primary cover image.

Cost-Control Patterns​

  1. Validate with dry-run first:
ideon write --dry-run "Your idea" --primary article=1 --secondary x-thread=1 --secondary x-post=1
  1. Start with fewer variants, then scale:
ideon write "Your idea" --primary article=1 --secondary x-post=1

  1. Use job files for repeatable experiments and adjust one variable at a time.

  2. Prefer faster models during exploration, then switch to quality-focused models for final runs.

Runtime-Control Patterns​

  1. Keep contentTargets tight in early iterations.
  2. Use x-thread only when thread structure is truly needed.
  3. Resume interrupted runs instead of restarting from scratch:
ideon write resume
  1. If a run repeatedly fails at one stage, diagnose with Troubleshooting before increasing target counts.

Practical Workflow​

  1. Run dry-run and confirm orchestration and outputs.
  2. Run a low-count live generation.
  3. Inspect markdown quality and generation.analytics.json.
  4. Increase counts or switch models.
  5. Save the resolved job.json for reproducible future runs.