[adamjohnwright]

Spec 001, T008 (research R4, FR-016, FR-003).

Measured. Release 98's 1,170 human diagrams, each drawn by the render service from the release's own files and placed by the layout's measure() and place(). None failed.

• 1,168 have 12 px labels; the median diagram is 3,654 px wide, the widest 19,652.
• Even on its best page, only 80 (7%) print labels at 5 pt, 311 at 3 pt, 652 at 2 pt, 1,070 at 1 pt. Median 2.1 pt.
• Today's book has no floor: event-pdf fits each diagram to the portrait text width and half the page height. From the files' sizes, its labels print at a median of about 1.85 pt (inference; the book's pages were not measured).

Decided (with the user, for a reader on a screen):

• each figure prints as large as the page allows, in portrait or on a landscape page of its own where that is larger (1,060 diagrams and all 221 illustrations), never split;
• each figure links to its interactive form on the website (FR-003 amended): a pathway's diagram in the Pathway Browser, the event selected where an ancestor's diagram draws it; a reaction's figure to its page;
• a figure whose labels print below 1 pt is listed in the run report (100 diagrams).

The rule is FLOOR_PT and the comparison in place(), so it is easy to change once curators have looked.

Illustrations. Their lettering is outlines; src/layout/lettering.ts reads it from the PATHWAY TEXT paths at the render service's viewport scale, so they are no longer reported as unmeasured. Every label is about a 12 px font; the reading from the render's SVG and from the file agreed.

Tests. 11 for the lettering, 6 for the links, the placement tests rewritten. Each mutant listed in tasks.md made one fail. The link was printed and read back: one annotation over the figure on its landscape page (two without the link's inline-block). All 143 tests pass against the fixture graph.

Also from #105's review: serve's bind names the analysis gateway as its second caller.

🤖 Generated with Claude Code