mcp, cli, ui: camera target, fit and fov in screenshots, snapshot --camera and ui.camera; screenshot section (#467)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo

docs/mcp.md

@@ -122,7 +122,7 @@ software on the computer.
 | `show_results` | `step?`, `field?`, `bands?`, `range?`, `deformed?`, `scale?`, `mode?`, `frame?`, `time?` | The `results.*` commands its arguments ask for (`deformed: false` draws the undeformed mesh; `scale` is a factor, 0 for the undeformed shape so a sequence from 0 swings the deformation out, or `"auto"`; `range: {min?, max?}` clamps the legend, values above `max` drawn in the over colour); the reply has the contour range (smallest and largest value and their nodes: `field: "displacement"` gives the largest displacement), modes or frames. |
 | `probe` | `step?`, `nodes?`, `at?`, `elements?`, `points?`, `forces?`, `mode?`, `frame?` | `solve.probe`: stored results at nodes, at the nodes nearest to points `at` (structural steps), elements (explicit) or points (fluid). |
 | `reactions` | `step?`, `sets?`, `about?` | `results.reactions`: reaction force and moment per node set and in total. |
-| `screenshot` | `view?`, `camera?`, `source?` (window, model), `width?`, `height?`, `path?` | A PNG image content block and what it shows. Headless: an offscreen render with the viewport's GPU pipeline (`FIELDCRAFT_GPU=auto|hardware|software|cpu`; the reply's `renderer` names the tier and adapter, or why the CPU drew it) framing the model, the geometry, the analysis setup's symbols (boundary conditions, loads, reference points, couplings; `sampled_symbols` lists definitions drawn with a sample) and the results display (deformed shape, contour bands). Bridged: the whole app window drawn offscreen (`ui.capture`, the default) or a model render with the app's camera (`ui.render`). `view` turns the camera to a standard view first; `camera` `{yaw, pitch, distance?}` to any orientation, in degrees (yaw 0 looks along +Y, −45 is the iso view's; pitch is the elevation, clamped to ±90) and model units (`ui.camera`; headless, the picture is framed unless `distance` is given). Give one of the two; the orientation is kept for the next picture, so a sequence of `camera` angles orbits smoothly. |
+| `screenshot` | `view?`, `camera?`, `section?`, `source?` (window, model), `width?`, `height?`, `path?` | A PNG image content block and what it shows. Headless: an offscreen render with the viewport's GPU pipeline (`FIELDCRAFT_GPU=auto|hardware|software|cpu`; the reply's `renderer` names the tier and adapter, or why the CPU drew it) framing the model, the geometry, the analysis setup's symbols (boundary conditions, loads, reference points, couplings; `sampled_symbols` lists definitions drawn with a sample) and the results display (deformed shape, contour bands). Bridged: the whole app window drawn offscreen (`ui.capture`, the default) or a model render with the app's camera (`ui.render`). `view` turns the camera to a standard view first (framing everything drawn); `camera` aims it: `{yaw, pitch, distance?, target?, fit?, fov?}`, in degrees (yaw 0 looks along +Y, −45 is the iso view's; pitch is the elevation, clamped to ±90) and model units. `target` is `[x, y, z]` or `{"center_of": "mass" \| "bbox", "bodies"?: [names]}` (design bodies drawn; an unknown name is an error), by default the centre of mass of the model drawn (element faces, else the bodies), so a sequence of yaws orbits the object and keeps it at the picture's centre. The distance is `distance`; with `fit: true` the one at which the targeted bodies (or the whole model) fill the frame from this direction; without either, the one at which the whole model stays in view from every direction, so a sweep keeps its size. `fov` is the vertical field of view in degrees (0, the default, orthographic; up to 150). Give `view` or `camera`; the camera is kept for the next picture, and the reply's `camera` is the one drawn with (target, distance, angles in radians). Bridged, `camera` moves the app's camera the same way (`ui.camera`; the window is otherwise untouched). `section` cuts the mesh as `fieldcraft-cli snapshot --section` does: `x`, `y`, `z` (at the middle) or `AXIS=VALUE`, hiding the elements on the camera's side (headless only). `fieldcraft-cli snapshot --camera JSON` takes the same camera. |
 | `select` | `filter?`, `box_mode?`, `mode?`, `clear?`, `geometry?`, `nodes?`, `elements?`, `faces?`, `at?`, `box?` | Bridged only: changes the app's viewport selection (`ui.select`; control-protocol.md, Selection) and returns it. Headless it is an error: selection belongs to the window. |
 | `selection` | — | Bridged only: the viewport selection (`ui.selection`), with `params`, the selection as a set's members for `model.set.create`. |
 | `undo`, `redo` | — | `edit.undo`, `edit.redo`. |