Skip to content

docs: simplify the model figure prose to two fixed paragraphs - #738

Merged
Jammy2211 merged 1 commit into
mainfrom
feature/model-figure-prose-simplify
Sep 14, 2026
Merged

Jammy2211 merged 1 commit into
mainfrom
feature/model-figure-prose-simplify

Conversation

@Jammy2211

Copy link
Copy Markdown
Collaborator

Docs Changed

  • docs/general/model_cookbook.md
  • docs/overview/overview_3_features.md

Documentation only — no source, no API change.

Section rewritten, not deleted: ## Solved Parameters becomes ## Solved and Unconfigured Parameters and keeps every fact — a linear profile's intensity solved by the inversion, a Basis owning no amplitude of its own, al.Pixelization.reconstruction solved with regularization coefficients its only sampled parameters, al.ps.PointSolved.centre solved analytically where al.ps.Point samples it, and areas_factor having no configured prior so the model cannot be fitted until one is supplied. The pill/dash/colour framing is gone.

Removed: the ## Reading the Figure section, and the cross-reference that pointed at it.

Sphinx: 131 warnings before and after, against a main baseline built in the same environment. Note that the recorded baseline of 134 is stale: main itself builds 131 in this environment.

What changed

Every af.ModelPlotter figure was accompanied by a prose block — an opener, a map/legend paragraph, and one or more paragraphs describing what that particular figure shows ("the greyed radius_break pills on the source are fixed values, which is the difference between the footer's count of sampled scalars and the number of pills drawn"). That block is replaced everywhere by two fixed paragraphs:

The same model can also be visualized as a figure, making its structure easier to understand at a glance.

The figure shows how the model is organized: which parameters belong to each component, and whether they are free, fixed, shared, linked by an expression, solved during the fit, or not configured. model.info provides the corresponding numerical details, including the prior assigned to each free parameter and the value of each fixed parameter.

The shorter notes at second and third figure sites keep their place but lose the same vocabulary — pills, plates, badges, greyed, footer counts — retaining only statements of fact about the model.

What was deliberately kept

  • Teaching content that happened to sit inside a figure block is preserved as its own paragraph, reworded to stand alone. Each one is listed below.
  • EP factor-graph blocks are exempt. Box / pill / edge is standard probabilistic-graphical-model notation, and the state overlay reports per-factor update counts, update age and status — information graph.info does not provide. This is not the redundancy being removed.
  • Image :alt: text and make_figures.py generator docstrings are exempt: describing how a figure looks is the whole job of alt text, and stripping it would degrade screen-reader accessibility.

Verification

  • End-to-end gate over scripts/*.py, notebooks/*.ipynb and docs/**/*.md: zero non-exempt occurrences of **map**, **legend**, pill, badge, plate, chip, greyed or footer-count language.
  • The canonical sentence appears exactly once per affected file.
  • Notebooks are regenerated from the scripts via PyAutoHands/autohands/generate.py, never hand-edited; check_navigator.py passes.
  • Every touched runnable script runs headless under the smoke profile, exit 0, no traceback.

No API changes — documentation and example prose only.

🤖 Generated with Claude Code

https://claude.ai/code/session_01LPCvGNHahZqynPyEfMijsb

Replace the map/legend prose block beside every `af.ModelPlotter` figure with
two fixed paragraphs, deleting the file-specific reading commentary that
followed it, and strip the same figure-rendering vocabulary (pills, plates,
badges, greyed pills, footer counts) from the shorter notes at later figure
sites. The human's verdict: the commentary is information overload and the
figure should be self-explanatory.

Teaching content that happened to sit inside those blocks is preserved,
reworded to stand on its own without referring to the figure's vocabulary.

The EP factor-graph blocks are exempt: box / pill / edge is standard
probabilistic-graphical-model notation, and the state overlay carries
per-factor update counts, ages and statuses that `graph.info` does not
provide. Image `:alt:` text is exempt too, since describing what a figure
looks like is its purpose.

Issue: PyAutoLabs/autofit_workspace#156

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LPCvGNHahZqynPyEfMijsb
@Jammy2211 Jammy2211 added the pending-release PR queued for the next release build label Sep 14, 2026
@Jammy2211
Jammy2211 merged commit a287408 into main Sep 14, 2026
4 checks passed
@Jammy2211
Jammy2211 deleted the feature/model-figure-prose-simplify branch September 14, 2026 22:27
@Jammy2211 Jammy2211 removed the pending-release PR queued for the next release build label Sep 26, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant