Configuration

All four settings are optional — the defaults are sensible.

• Max output tokens (default 16000) — upper bound on the layout JSON the model may return. Larger pages need more.
• Temperature (default 0.3) — lower values give steadier, more valid JSON.
• Request timeout (default 150 seconds) — how long to wait for the model before failing.
• Debug (default off) — include the raw model output in the response for troubleshooting.