Protocol-level simulation of the flOMNI hardware injected via socket_cls:
real device/controller classes run unchanged against state machines
implementing fgalil.dmc, galil_micos_upr.dmc, the Smaract MCS protocol
and the Orchestra CommunicationServer. Cameras reuse the real device
classes with a synthetic frame source. Includes simulated_flomni device
config (boots referenced at 'in' positions) and an offline harness
covering moves, referencing, rt feedback/tracker, flyer scan readout
and the gripper transfer routine.
Replace 11 identical global-var-backed getter/setter property pairs with a
single _GlobalVarParam descriptor. Behavior is unchanged: same keys, defaults,
and None-handling; stitch_x/stitch_y retain int enforcement via the descriptor's
type_ guard (typeguard.check_type). Net -57 lines with no functional change.
xrayeye_update_frame() now checks all configured beamline states and, if any
are not valid, warns the user (the frame may be empty when the beam is down)
and asks before taking a frame. The automated alignment path is unaffected.
ftransfer_sample_change(): clarify the remove-only prompt to "sample currently
in the sample stage", and include tray slot 20 in the empty-slot scan
(range(1, 20) -> range(1, 21)) in both the remove and mount branches, matching
check_position_is_valid's 0-20 range.
When the eye is already out, optics in, and feedback running, skip the
eye-out/optics-in transition to avoid an interferometer reset. Enables
repeated alignment scans and interleaving alignment into a tomogram
without leaving the current measurement position.
feedback_disable/enable_with_reset cycle re-zeros the interferometers
and moves off the current sample position. Only run it when the FZP
actually needs to move; add force_feedback_reset kwarg and _ffzp_is_in
helper.
X-Ray Eye alignment GUI + script now share a single, device-configured
pixel calibration and the GUI shows live ROI dimensions.
GUI (x_ray_eye.py):
- Add read-only "ROI size" field below the message box, showing the active
ROI's x/y extent, updated live via each ROI's sigRegionChanged. Rectangles
report width/height; circular ROIs read raw state["size"] so an ellipse
state would be reflected (currently aspect-locked upstream).
- Add "Calibration" field showing microns/pixel, or "uncalibrated (pixels)"
when the camera parameter is absent.
- Resolve mm/pixel from dev.cam_xeye.user_parameter["pixel_calibration"];
fall back to raw pixels (scale 1.0, unit "px") if unavailable.
- Enlarge the message field from 60 to 90 px height.
Alignment script (x_ray_eye_align.py):
- Replace the hard-coded PIXEL_CALIBRATION constant with a pixel_calibration
property reading the same cam_xeye user parameter; fall back to
PIXEL_CALIBRATION_DEFAULT if unavailable.
- Fold the recurring factor of 2 into the calibration: xval_x_* come straight
from raw ROI pixels on submit, so the true scale is 0.05/113 mm/pixel.
Default is now 0.05/113 and the three /2 factors (FZP center x, height
centering, FOV) are removed. Computed physical values are unchanged.
Device config (cam_xeye_pixel_calibration.yaml):
- Add pixel_calibration userParameter (0.05/113 mm per raw ROI pixel) to
cam_xeye. Merge the userParameter block into the existing device entry.
scilog_last_scans now puts the user comment at the top of the entry and
drops the flOMNI logo attachment. number_of_scans accepts 0 to send a
comment-only entry. When fewer projection records exist than requested
(e.g. non-ptycho scans, which write no timing record), the available ones
are used and a note reports how many of the requested count were found.
At BEC session start, compare bec.active_account against the new global
var defaults_applied_for_account. On a genuine account change, interactively
offer to reset all tomo scan parameters to their defaults for the new
experiment; a client restart within the same account is a silent no-op.
Add _set_default_tomo_params() writing all 20 scan parameters via the
property setters.
process looks like a BEC client), plus a delayed (500 ms) backup
stop_devices request, defaulting to stop-all so the widget doubles as
a generic emergency stop when opened standalone from the launcher menu
The gripper routines (#GRGET/#GRPUT) run autonomously on the Galil
(thread 3); Ctrl+C previously only killed the BEC-side polling loop
while the controller kept executing, including ongoing motion.
- add FlomniSampleTransferMixin.ftransfer_abort(): dev.ftransy.stop()
(stop_devices endpoint -> device server -> XQ#STOP), wait for thread 3
to halt, then put the controller back in posmode; deliberately no
ensure_gripper_up() - recovery after an abort is manual
- wrap the mntprgs polling loops in ftransfer_get_sample/put_sample in
try/except KeyboardInterrupt -> ftransfer_abort()
- answering "No" at a confirmation point now also hard-aborts the
controller routine instead of leaving it parked in #CONFIRL
- ConsoleButtonsWidget ABORT: keep SIGINT to the client, plus a delayed
(500 ms) backup stop_devices request so the hardware stops even if
the client is hung; delay ensures the controlled client-side abort
wins the race against #STOP clearing mntprgs
Add a small ConsoleButtonsWidget (Yes/No/Abort + message label) as a
GUI alternative to console yesno() prompts during sample transfer,
docked below the live camera view.
- bec_widgets: new ConsoleButtonsWidget (csaxs_bec/bec_widgets/widgets/
console_buttons/). Yes/No write to a response property polled by the
caller; Abort sends a real SIGINT to the BEC client process (captured
via os.getppid() at construction), stopping any blocking call
(including an in-progress motor move) exactly like a console Ctrl+C.
- omny_general_tools: add OMNYTools.gui_yesno(), a GUI-backed
alternative to yesno() with the same message/default/autoconfirm
semantics, polling the widget's response instead of blocking on
input(). Not yet used as a general yesno() replacement.
- gui_tools: flomnigui_show_cameras() now also opens the console widget
docked below the two camera views (set_layout_ratios for relative
sizing, since dock_area.new() has no pixel width/height kwarg).
flomnigui_show_gui() now passes an explicit window geometry, right-
aligned on a 2560x1440 screen.
- flomni: add ftransfer_confirm_dialog(), which uses the GUI console
when available (falling back to the plain CLI yesno() otherwise) and
wire it into ftransfer_confirm()'s per-step "All OK? Continue?"
prompt during sample transfer. Other yesno() call sites (stage-in
flight-tube check, gripper-move stage-out check) intentionally left
on the CLI for now.
Known issue: requested window height in flomnigui_show_gui() is not
taking effect as expected at startup; needs follow-up.
Passing -1 stows the sample currently in the sample stage into a free
tray slot without mounting a replacement, matching the old control
system's sample_change(-1) behavior. Existing swap-path logic is
unchanged.
Add collect_empty_frames(), which acquires 10 flat-field frames at
angle 0 with fsamx shifted out of the beam by fovx/2 (opposite the
normal alignment x-offset). Called once at the start of a new
tomo_scan() (not on resume). Frames are excluded from the ptycho
reconstruction queue and logged to tomography_scannumbers.txt with
subtomo_number=0.
Expose the ConstEmitter runtime constants through the high-level flomni
interface, alongside the existing laser_tracker_* methods, and add an
interactive tweak tool for the two tracking target positions.
New methods on FlomniSampleTransferMixin (thin wrappers delegating to
dev.rtx.controller, same pattern as laser_tracker_show_all):
laser_parameters_show_all()
laser_parameters_set_targety(val)
laser_parameters_set_targetz(val)
laser_parameters_set_intensity_threshold_laser(val)
laser_parameters_set_psd_intensity_threshold_tracking_low(val)
laser_tweak(): interactive 2D raw-terminal loop for the two tracking targets,
modeled on OMNYTools.tweak_cursor.
left/right : target y -/+ (fixed 0.01 step, absolute set of current +/- step)
up/down : target z +/-
s : return both targets to their launch positions
q : quit (terminal state restored in finally)
On launch and after every change it prints the current y/z targets and the
fzp x interferometer signal (read_ssi_interferometer channel 1), so the signal
response can be watched while tweaking.
termios/tty/fcntl/sys are imported locally in laser_tweak to avoid touching
module-level imports. Additive change; no existing behaviour modified. Requires
the ConstEmitter emitter_* methods on RtFlomniController.
Replace ~44 per-slot blocking device reads per poll with a single bulk
flomni_samples.read(), and add change-detection so cells only repaint
when their state actually changes. The old per-slot accessors stalled
the GUI event loop every 2s; steady-state polling now does no UI work.
Writes invalidate the changed slot's cache so local edits still repaint
immediately.
GUI for ftransfer_show / ftransfer_modify. Displays the
sample stage (0), gripper (100) and magazine slots 1-20 in a 4x5 grid
matching the CLI layout. Right-click per slot corrects the records
(new / rename / clear), each action confirmed by its own dialog and
written immediately. Bookkeeping only: writes flomni_samples directly
(mirroring flomni_modify_storage_non_interactive), never moves hardware.
Guarded on fsamroy; OMNY shows a not-implemented placeholder.
Expose the flOMNI ConstEmitter runtime constants (served over the control-loop
TCP server, command letter 'e') to BEC users. These were previously baked into
Orchestra module parameter files; they can now be read and set live.
New RtFlomniController methods (all added to USER_ACCESS):
emitter_get() -> dict of the four values
emitter_show_all() -> PrettyTable printout
emitter_set_tracking_target_y(val) -> "ey<val>"
emitter_set_tracking_target_z(val) -> "ez<val>"
emitter_set_intensity_threshold_laser(val) -> "et<val>"
emitter_set_psd_intensity_threshold_tracking_low(val) -> "el<val>"
emitter_get issues "es" and parses the CSV reply
(targetpos_tracking_y, targetpos_tracking_z, intensity_threshold_laser,
PSD_intensity_threshold_tracking_low), same pattern as the Ts/sr readouts.
Setters use socket_put with a fixed %.4f value, same pattern as the pa
setpoint command; they are fire-and-forget (the server does not ack sets).
The rotation-loop laser threshold (intensity_threshold_rot_laser) is a startup
constant on the server side and is intentionally neither read back nor settable
here.
Named methods were chosen over a single opaque command to match the existing
laser_tracker_* / feedback_* convention and to self-document each value.
Requires the corresponding flOMNI control-loop change (ConstEmitter module +
'e' command in CommunicationServer). Additive change; no existing behaviour
modified.
Expose the ConstEmitter runtime constants through the high-level flomni
interface, alongside the existing laser_tracker_* methods, and add an
interactive tweak tool for the two tracking target positions.
New methods on FlomniSampleTransferMixin (thin wrappers delegating to
dev.rtx.controller, same pattern as laser_tracker_show_all):
laser_parameters_show_all()
laser_parameters_set_targety(val)
laser_parameters_set_targetz(val)
laser_parameters_set_intensity_threshold_laser(val)
laser_parameters_set_psd_intensity_threshold_tracking_low(val)
laser_tweak(): interactive 2D raw-terminal loop for the two tracking targets,
modeled on OMNYTools.tweak_cursor.
left/right : target y -/+ (fixed 0.01 step, absolute set of current +/- step)
up/down : target z +/-
s : return both targets to their launch positions
q : quit (terminal state restored in finally)
On launch and after every change it prints the current y/z targets and the
fzp x interferometer signal (read_ssi_interferometer channel 1), so the signal
response can be watched while tweaking.
termios/tty/fcntl/sys are imported locally in laser_tweak to avoid touching
module-level imports. Additive change; no existing behaviour modified. Requires
the ConstEmitter emitter_* methods on RtFlomniController.
A block is external and self-resolving, so the immediate "measurement stopped,
confirm" warning chime was misleading there - the user cannot act on it and
should just wait. Suppress startWarning when the destination status is blocked;
the blocked track still handles it (silent until the 60-min blockedChime, then
hourly). scanning->idle (success tone) and scanning->other (warning tone) are
unchanged. Help text updated.
- Audio: a normal finish (scanning -> idle) now plays a rising success
arpeggio (C5-E5-G5) instead of the falling warning chime, so a completed
measurement is audibly distinct from a stop to any other state (which keeps
the falling 660->440 warning). startWarning(finished) selects the chime for
both the initial tone and the 30s repeats; the Watch LED text reads
"Measurement finished" vs "Measurement stopped" accordingly. Help text
updated.
- Layout: body uses min-height:100dvh (with 100vh fallback) so the page no
longer extends below the visible viewport on mobile, where iOS 100vh counts
the retracted-toolbar height and leaves empty scroll room at the bottom.
The instrument-details pane showed only a hand-built 6-row settings summary,
far less than the tomo queue shows per job, and nothing useful for a tomo
started outside the queue.
- Flomni _collect_setup_data now returns current_params: the tomo parameters
read live from global vars, keyed by the same names the queue stores in
job.params (_CURRENT_PARAM_KEYS, matching TQ_PARAM_DISPLAY). Values left raw;
the page formats them. Replaces the old settings dict.
- Extract the queue's param-table builder into a shared buildParamRows() JS
helper (carries the computed Projections row), used by both the queue and the
new instrument 'Current measurement' section, so they can't drift.
- renderInstrument renders current_params via buildParamRows (Sample name row
first); card hides only when there are neither params nor temperatures.
- Base _collect_setup_data contract updated: settings -> current_params.
umvr_fsamy_tracked/umv_fsamy_tracked require feedback OFF but did not
enforce it. Add an _internal flag; for user/CLI calls, warn that feedback
must be off (active alignment may be lost), prompt (default yes), and
disable feedback on confirm. Internal callers pass _internal=True since
they already manage the feedback lifecycle (ftransfer_flomni_stage_out,
x_ray_eye_align height-centering and mvy jog).
Match the deleted-widget detection used by all other flomniGuiTools
methods. The previous _flomnigui_check_attribute_not_exists path relied
on hasattr(self.gui.flomni, "Image"/"Image_0"), which returns stale/
mismatched results when docks are recreated, leaving an empty window
after a user closed the GUI mid sample-transfer. _is_deleted() checks
registry membership and reliably triggers a rebuild.
renderTomoQueue auto-opened the running job whenever the open-state snapshot
was empty, so collapsing the only open job (the running one) made the snapshot
empty and re-opened it on the next refresh. Open-state was also keyed by
positional index, which mis-assigns the flag when jobs are added/removed/
reordered.
- Snapshot open jobs by stable key (added_at|label) via a data-key attribute
instead of positional index.
- Auto-open a running job only once, on first appearance, tracked in
window.__tqAutoOpened; afterwards honor the DOM snapshot so a manual collapse
sticks. Prune the set when jobs disappear.
A block interrupts the running scan, clearing active_request_block, so
the old _derive_status() (which only returned 'blocked' inside the
queue_has_active_scan branch) fell through to 'idle' with the blocked
LED off, despite watched beamline states being out of spec.
- _derive_status: return 'blocked' on beamline_blocking OR queue_locks,
independent of active scan; only fresh-heartbeat 'scanning' outranks it.
- _cycle: beamline_blocking = enabled is True AND any(state mismatched).
Gated on enabled so a disabled/unknown interlock with an out-of-spec
state does not raise BLOCKED. Also counted as activity so the idle
clock restarts once cleared.
- beamline-states card summary updated in lock-step: mismatched states are
only called "currently blocking" when enabled, else "out of spec
(interlock disabled/unknown, not blocking)", so pill and card can't
disagree.
- status-detail JS: describe blocking beamline states when no explicit
queue lock is present. Update help text.
webpage_generator: report blocked on interlock even when scan is interrupted
- Add HAS_TOMO_QUEUE and TOMO_TYPES capability flags to
WebpageGeneratorBase, defaulting to flomni's current behavior
(type1 grid-8, type2/3 golden-capped) so this is a no-op for flomni.
- Gate the tomo_queue global-var read in _cycle() behind
HAS_TOMO_QUEUE; unaffected: queue_status/queue_locks/beamline_states,
which come from BEC's own primary scan queue and stay common to
all setups.
- _render_html() now conditionally emits the 'Tomo queue' card and
its DEFAULT_ORDER slot, and drives the projection-count formula
from a TOMO_TYPES JSON blob (generic calcProjections()) instead of
a hardcoded per-instrument JS branch.
- LamniWebpageGenerator: HAS_TOMO_QUEUE=False, TOMO_TYPES with
2-subtomo and 8-subtomo equally_spaced_grid entries (placeholder
keys pending real tomo_type values from the LamNI producer side).
A block interrupts the running scan, clearing active_request_block, so
the old _derive_status() (which only returned 'blocked' inside the
queue_has_active_scan branch) fell through to 'idle' with the blocked
LED off, despite watched beamline states being out of spec.
- _derive_status: return 'blocked' on beamline_blocking OR queue_locks,
independent of active scan; only fresh-heartbeat 'scanning' outranks it.
- _cycle: compute beamline_blocking = any(state mismatched), defined
identically to the beamline-states card so pill and card can't disagree;
also count it as activity so the idle clock restarts once cleared.
- status-detail JS: describe blocking beamline states when no explicit
queue lock is present. Update help text."
x01dcandholler
appel_c
wakonig_k