# The prompt sent to the agent (hall)

```text
I'm organising the Harbourline Expo at Harbourline Expo Hall. The hall is modelled in Laydyne and open in the Studio. Last week we compared three stage positions by hand; they are saved as layout options (and design options) "Stage at the far end", "Stage in the centre" and "Stage near the entrance". The live layout is the far-end one. Scenario expo-day is the visitors' day (3 runs from seed 7).

Now I want Laydyne to search the setup instead of us choosing among three:
- Where the stage goes: in any of the three zones (the far end, the centre, near the entrance, as in the saved arrangements, with the booths in the other two zones), and shifted sideways by −8, −4, 0, 4 or 8 m. The stage set is stage, lectern, screen-1, screen-2 and the audience zone; the booths are booth-1 to booth-40 with their back walls booth-1-wall to booth-40-wall.
- How wide the cross aisles between the booths are: the booths are 3 m wide and 0.9 m apart now; try 0.9, 1.2, 1.5 and 1.8 m.
- Sight lines must hold: everyone in the audience zone sees the whole stage front (leave out 0.5 m at each end; eyes at 1.6 m, the stage front at 2.4 m), and our streaming camera on the west wall (x 1.6, z 31.2, 2.9 m high, 30° wide, 20° high, 60 m range, aimed at the stage front) sees at least 99 % of it.
- No new layout-check problems; aisles at least 0.9 m.
- Goal: the least walking in total; among equals, the lowest average wait.
- The same 3 runs and seed 7 as before.

Use search_layouts (you can search the stage position and the aisle width in one search or one after the other). Then compare the best with each of our three hand-made arrangements using compare_process_options on the saved options, and tell me which differences are beyond the noise. Show me the best setup in 2D.

Finish with a short table for the three hand-made arrangements and the search's best: where the stage is, aisle width, the sight-line results, total walking and average wait, with 95 % intervals. Say it is the best of the candidates searched, not a proven optimum, and what the model does not calculate (crowding, people blocking the view…).
```