[adamjohnwright]

Spec 001, T007 (with T005's (c)). Replaces #96, with its review's findings fixed.

Measured each event of the event-pdf baseline's sample at depth 1 on this host. Each ran in a render service of its own:

• at the website settings, two at once;
• with a fresh cache key and no cache hits;
• with its browser warmed on one unrelated reaction;
• reading the release's diagram files locally through RENDER_FILES (WebsiteAngular #777).

Event	Figures (s)	Total (s)	event-pdf (s)
Five reactions	2.6–2.7	4.0–4.9	0.5–1.7
Pathways, 1–34 reactions beneath	4.0–19.8	6.5–22.6	1.3–3.8
Pathways, 143–1,150 beneath	11.0–18.9	14.5–23.2	4.4–9.5
Signal Transduction	26.3	32.9	46

• Figures alone exceed event-pdf's whole time for every reaction and every pathway but Signal Transduction.

SC-005 decided (the user left it to us): without an analysis, a single-event PDF reads its figures from the release's figures' folder (the book's, SC-010) and keeps event-pdf's times; with an analysis, its figures are drawn on request within 6 s plus 2 s per figure. serve takes --figures; T022 and T023 follow.

• One figure with a token: 4.5 s alone, 9.1 s for the last of four at once, 19.6 s for the last of eight.
• Script: ci/measure/single.ts; outputs in ci/measure/results/2026-11-02/single/.

Also in this PR:

• figures requires a render service reading the release's files locally, and refuses one that does not (contracts/render.md, T046, T042, R10).
• R7 records that T005's pathway timings included fetches from the public download site, before RENDER_FILES existed.
• T005 and T006 carry their PR number, Spec 001 T005 and T006: a no-go, and the figures drawn in a step of their own #95.

From #96's review: a pathway's figures take about 1.8 s each (1.4–2.6), not 1.5; the warm-up figure was a pathway no sampled event draws, not a reaction; how the outputs were assembled is recorded; the script counts pages with pdf.js and keeps figure times in request order; cli.md and the folder contract name the new refusal and its limit.

🤖 Generated with Claude Code

[adamjohnwright]

Replaced by the follow-up PR: re-measured with a checked warm-up, SC-005 widened, review findings fixed.