[Jammy2211]

Summary

For point-source modeling (e.g. a lensed quasar), passing an image of the dataset to AnalysisPoint now gives a figure with the point positions clearly plotted over it. Before this change, AnalysisPoint(image=...) accepted the image and silently dropped it.

The new image_with_points.png shows:

• observed positions as white markers with a black edge, which stand out on any colormap
• after a fit, model positions as red crosses and the critical curves

Positions are placed with PyAutoArray's own overlay reflection (_overlay_yx_for_origin), so they stay registered to the image.

Closes part of #779; workspace and autolens_visualization follow-ups come next.

API Changes

• AnalysisPoint now stores self.image. It is for visualization only and does not enter the likelihood.
• New aplt.plot_image_with_points(image, dataset, fit=None, ...).
• New PlotterPoint.image_with_points(...), gated by plots.yaml point_dataset.image_with_points and fit_point_dataset.image_with_points (both default true).
• VisualizerPoint emits the figure before and after the fit, including on quick_update, whenever an image is supplied.
• Purely additive; no migration needed.
  See full details below.

Test Plan

• pytest test_autolens/ — 838 passed, 1 xfailed (local, worktree).
• New tests check that the analysis stores the image, that the plot writes image_with_points.png, and that the visualizer writes it before and after the fit (with and without quick_update) only when an image is given.
• Visual check: a render of PSF blobs at known (y,x) positions has its markers exactly on the peaks, with the critical curve overlaid.

Readiness

At ship time, Heart readiness was YELLOW (score 70):

• workspace validation not passing (1 failed, cloud#39224774724: euclid scripts/tools/diagnose_latent.py)
• manifest drift: workspace checkouts (manifest ↔ disk) — 1 mismatch(es) vs PyAutoMind/repos.yaml
• release validation incomplete: no rehearsal for current source

The Heart door read RED at task start. The human authorized development to proceed on 2026-10-21 ("I override, go"). None of these reasons involve point-source code.

Full API Changes (for automation & release notes)

Added

• autolens.plot.plot_image_with_points(image, dataset, fit=None, output_path=None, output_format=None, image_plane_lines=None, image_plane_line_colors=None, title_prefix=None) — writes image_with_points.png
• autolens.point.model.plotter.PlotterPoint.image_with_points(image, dataset, fit=None, image_plane_lines=None, image_plane_line_colors=None)
• plots.yaml: point_dataset.image_with_points: true, fit_point_dataset.image_with_points: true

Changed

• AnalysisPoint.__init__ now stores image as self.image (previously accepted and discarded).

Migration

• None.

Generated by the PyAutoLabs agent workflow.

🤖 Generated with Claude Code

[Jammy2211]

Follow-up commit b0ce0a3, found while rendering the autolens_visualization figures:

• Model crosses are now drawn only for image-plane fits (AbstractFitPositionsImagePair). For FitPositionsSource, model_data is the source-plane centre, which had been drawn over the image near (0,0).
• Observed positions are now hollow white rings, so the PSF peaks and faint images are no longer hidden under the markers.
• New parametrised test covering image-plane vs source-plane fits.

pytest test_autolens/: 840 passed, 1 xfailed.

[Jammy2211]

Workspace PR: PyAutoLabs/autolens_workspace#591
Visualization PR: PyAutoLabs/autolens_visualization#6