Interactive Pdf Binder

What it is

Interactive PDF Binder Navigation is not a DOM component. It is the design-system record for the navigation layer that tool-pdf-interactive.py (reportlab + pypdf) stamps onto an assembled binder of source PDFs. The recipe's html/css describe only the design-system preview surface; the shipped artifact is a static PDF whose interactivity lives entirely in PDF link annotations and the document outline.

Every value on this page — geometry, colour, type — is drawn from the canonical Python constants in the generator, extracted directly rather than from prose, and consolidated into the Paper pillar as the paper.*.pdf-nav / pdf-binder token groups. Corrected 2026-07-29: the TOC-entry position/pitch and INDEX-button size were re-verified against three shipped production binders and found to have never matched the canonical values — the originals came from an early draft, not measurement. See Coordinate space and geometry for the corrected figures.

This is the only member of the Paper document family that is genuinely interactive within its viewer. The other five families are static CSS-print layouts: legal subscription agreement, legal prospectus, legal agency suite, financial report layout, and financial statement (year-end).

When to use

• Assembling several source PDFs into one deliverable — a binder of agreements, schedules, or statements that a reader must navigate between. The slip-sheet TOC gives each source document a cover and a jump target.
• A binder that will be read on screen, in a PDF viewer. The GoTo links and document outline are the point; a binder intended only for print does not need this layer.
• When the reader needs to return to the index repeatedly. The INDEX button on every content page is the return path.
• When several documents share a parent — reach for the grouped-TOC variants (toc-group-header / toc-entry-child) rather than repeating the shared name in every child's title. A group of exactly one document stays a plain numbered row — a header over a single child costs a line and groups nothing.

When not to use

• Do not use it for a single, authored document. If there is nothing to bind, there is no TOC to generate — reach for the relevant CSS-print family instead.
• Do not use it to restyle the source PDFs. The generator overlays navigation; it does not reflow, re-tag, or re-typeset the documents it binds.
• Do not treat the point-space geometry as CSS @page geometry. This component is rendered in PDF point coordinates with a bottom-left origin — it deliberately keeps its own paper.primitive.pdf-nav.* group rather than sharing the CSS-print page primitives.
• Do not invent a numbered sub-level for grouped entries. House style is a single number spine (1. 2. 4., headers consume a number, children carry none) — see Rejected alternatives below for why a decimal scheme (4.1/4.2) was tried and dropped.

Four navigation surfaces

1 · Slip-sheet TOC cover

A generated cover page inserted before each source document. It carries a header title, an organisation subtitle, a 1.5pt rule, an optional right-aligned draft/version label (first sheet only), and an italic footer instruction. A single slip sheet holds up to 8 TOC entries; that ceiling is currently a hard limit in the generator, though the real constraint appears to be vertical space, not row count (see Open questions oq-2).

2 · TOC entry (top-level)

One navigable row per bound document, in two states:

• Inactive — a navy number and title for a document other than the current one, with an invisible full-rect GoTo link to that document's slip sheet.
• Active — the row for the current document: a grey-light highlight rectangle, black text, and a drawn 7.2×7.2pt filled-ink marker square (not a glyph — see below).

3 · Grouped TOC entries — group header and child

Added 2026-07-17, adopted as the house standard for TOC slip sheets. When several documents share a parent, a toc-group-header row takes the next number in the spine and labels the group; its toc-entry-child rows carry no number of their own, leading instead with an en-dash. One number spine, no exceptions — everything else in the recipe (an inactive/active top-level entry, a group header) shares column toc-num-x; a numbered child would open a second number column and a second title column, and the page would stop having a single left edge.

A clickable document can itself be the group's parent — a doc-parented group — when schedules or annexes belong to one specific agreement rather than to a generic label (e.g. Schedules A–D nested under the Long-Form Agreement they amend). In that case the parent keeps its own PDF and slip sheet; its bookmark nests the children under its own outline item (parent= its own add_outline_item return, never a synthetic group node), and its own link rect narrows (toc-docparent-link-dy/-height) so it cannot overlap the first child's link rect.

Numbering the children 4.1–4.4 (or as new siblings 4.–7.) would assert a false relationship — in a legal binder, that they are peers of the agreement rather than part of it. This is a domain-correctness question, not a taste call.

4 · INDEX return button

A 54×14pt navy rounded rectangle (3pt corner radius) with a white INDEX label, placed lower-right. Stamped on every content page — never on a slip sheet — and carries a transparent GoTo link back to page 0, top.

The label was renamed from HOME on 2026-07-17: "HOME" read as browser/app vocabulary, the one word on the page that broke the printed-instrument register, while "INDEX" names the button's actual destination and matches the binder's own footer sentence, "Interactive Index: Click a document title above…". The change is zero-geometry — at Helvetica-Bold 8pt, HOME measured 24.0pt wide and INDEX measures 24.4pt, well inside the 54pt button.

Variants

Variant	Surface	Description
slip-sheet	TOC cover	Generated cover preceding each source document: header title, org subtitle, 1.5pt rule, optional right-aligned draft label (first sheet only), italic footer instruction. Holds up to 8 TOC entries per sheet.
toc-entry-inactive	TOC entry	Navigable row for a non-current document: navy number + title, invisible full-rect GoTo link to that document's slip sheet.
toc-entry-active	TOC entry	Row for the current document: grey-light highlight rect, black text, drawn 7.2×7.2pt marker square; no link (self).
toc-group-header	Grouped TOC	Labels a group of following children: 11pt Helvetica-Bold, ink (not navy — it isn't navigable), title case, no highlight/marker/link/rule, takes the next number in sequence. A clickable document may itself be a doc-parented group header.
toc-entry-child	Grouped TOC	Nested under a group header: no number, an en-dash at toc-child-dash-x, title one 18pt step past toc-title-x. Same colour/type/marker/link rules as a top-level entry.
home-button	Return button	54×14pt navy rounded-rect (3pt radius) with white INDEX label, lower-right placement. Stamped on every content page (never slip sheets); transparent GoTo link to page 0 top.

Coordinate space and geometry

The binder is generated in PDF point-space on US Letter, 612×792pt, origin bottom-left — not a CSS @page box model. All geometry comes from paper.primitive.pdf-nav.*, corrected 2026-07-29 against three shipped production binders (MOU, Agency Agreements, MX Prospectus) plus the Client A reference generator:

Primitive	Token	Value	Was (stale)
Page width	{paper.primitive.pdf-nav.page-width}	612pt	—
Page height	{paper.primitive.pdf-nav.page-height}	792pt	—
Content zone	margin-left / margin-right	72pt–540pt	—
First TOC entry baseline	toc-entry-first-y	595pt	565pt
TOC entry step (centre-to-centre)	{paper.primitive.pdf-nav.toc-entry-step}	48pt	65pt
TOC entry height	toc-entry-height	46pt	(unchanged)
TOC entry width	toc-entry-width	468pt	530pt
TOC number x-position	toc-num-x	96pt	64pt
TOC title x-position	toc-title-x	114pt	82pt
INDEX button width	{paper.primitive.pdf-nav.home-width}	54pt	64pt
INDEX button height	{paper.primitive.pdf-nav.home-height}	14pt	20pt
INDEX corner radius	{paper.primitive.pdf-nav.home-corner-radius}	3pt	4pt
Slip-sheet rule stroke	rule-stroke	1.5pt	—

Grouped-entry geometry (new 2026-07-17):

Primitive	Token	Value
Child en-dash x-position	toc-child-dash-x	120pt
Child rect x / width / height	toc-child-rect-x/width/height	98pt / 442pt / 22pt
Child rect y-offset	toc-child-rect-dy	-7pt (not -25pt — see below)
Child row pitch	toc-child-step	26pt
Doc-parented pitch to first child	toc-docparent-to-child-step	36pt
Doc-parent's own link rect y-offset / height	toc-docparent-link-dy / -height	-15pt / 36pt
Active-row marker size	active-marker-size	7.2×7.2pt

A bug in the original reference generator, not copied here: its child rows use (98, y−25, 442, 24) — a band that sits entirely below the text baseline at y, so the highlight box and click target land under the wrong strip. It went unnoticed because those children are rarely rendered active. toc-child-rect-dy: -7pt brackets the baseline correctly; all binders shipped 2026-07-17 use the corrected value.

The 1.5pt slip-sheet rule is the emphasis step of the shared Paper rule-weight ladder — the same 1.5pt used for cover rules and summary-page borders elsewhere in the pillar.

Active marker is a drawn rectangle, not a glyph

An earlier revision used ► (U+25BA) as the active-row marker. That character is absent from Helvetica's WinAnsi encoding (the encoding reportlab's core-14 fonts use), so readers substituted .notdef — which happened to render as a filled black box in some viewers, but .notdef rendering is undefined and reader-dependent (a hollow box, a different shape, or nothing at all in another viewer). Resolution: draw a 7.2×7.2pt filled-ink rectangle instead — same appearance, deterministic in every reader, measured directly off the originally-rendered .notdef box. Do not pin a marker glyph in the recipe; pin the rectangle.

Known trade, flagged not resolved: the earlier glyph lived in the text layer, so the active row was machine-extractable. A drawn rectangle is graphics — slip-sheet text is now byte-identical across every sheet, and the active row's only remaining signal for a screen-reader user is the highlight shading plus ink-vs-navy text colour, not an extractable character.

Colour and type tokens

Colour resolves through paper.semantic.pdf-binder.* to the paper.primitive.color tier:

Semantic token	Resolves to	Value	Role
pdf-binder.toc-entry-inactive	pdf-nav-navy	#002e63	Inactive TOC number + title
pdf-binder.toc-entry-active	ink	black	Active-entry text
pdf-binder.toc-entry-highlight	pdf-nav-grey-light	#f5f5f5	Active-entry highlight fill
pdf-binder.toc-group-header	ink	black	Group-header text — not navy; a header isn't navigable, so navy (which means "this row links") would be a lie.
pdf-binder.active-marker-fill	ink	black	The drawn active-row marker square
pdf-binder.home-button-fill	pdf-nav-navy	#002e63	INDEX button rectangle
pdf-binder.home-button-label	pdf-nav-on-navy	#ffffff	INDEX label
pdf-binder.header-ink	ink	black	Slip-sheet header title
pdf-binder.supporting-ink	pdf-nav-grey-dark	#4d4d4d	Subtitle + italic footer instruction
pdf-binder.version-label-ink	pdf-nav-grey-label	#737373	Draft / version label

Type is set in the PDF core-14 Helvetica stack (Helvetica, Arial) — no font embedding required. A group header reuses the same pdf-nav.toc-entry token as any other row (11pt bold) — differentiation is colour and grouping proximity, never a second type voice:

Token	Size / weight	Applied to
pdf-binder.binder-title-type	16pt bold	Slip-sheet header title
pdf-binder.home-label-type	8pt bold	INDEX label
pdf-nav.subtitle	10pt regular	Org subtitle
pdf-nav.toc-entry	11pt bold	TOC rows, including group headers and children
pdf-nav.footer	9pt regular, rendered italic (Helvetica-Oblique)	Footer instruction
pdf-nav.draft-label	9pt regular	Draft / version label

Accessibility

This produces a static PDF page, so web-UI accessibility mechanisms — focus rings, ARIA roles, keyboard tab order — do not apply. What applies is PDF-native accessibility, and it is partial by design:

• Navigation is /GoTo link annotations plus a PDF document outline (/PageMode /UseOutlines), with /DisplayDocTitle true so viewers announce the document title rather than the filename.
• Grouped entries mirror their visual nesting structurally. A group header's children nest under its outline item via add_outline_item(..., parent=...). For a doc-parented group this is the only machine-readable statement that the children belong to that document rather than being its siblings — indent and dash alone don't carry that.
• A group header is never its own link target. It labels a group that begins at its first child; a GoTo there would land where clicking the child already lands. Its bookmark resolves to the first child instead, since an outline entry must resolve somewhere.
• Contrast is strong. Navy #002e63 on white measures ~13:1, and white on navy ~13:1 — both AAA. This is the print/PDF contrast target the recipe declares against WCAG 2.2 AA.
• Tagged-PDF reading order is not handled. Source PDFs pass through untagged — the generator does not add Tagged-PDF structure, so reading-order conformance for assistive technology depends entirely on the tagging of the source documents. This is a known, open accessibility gap, not a solved property (see Open questions oq-3). Do not claim screen-reader conformance for a binder whose sources are untagged.

The artifact is static: no motion, no transitions.

Rejected alternatives — do not reopen

ISO 2145 decimal numbering (4 / 4.1 / 4.2, no trailing stop) for grouped entries: verified correct as a standard, rejected as a house decision. Nothing in the corpus used it, it solved no problem anyone had, and decimal children would open a second number column, destroying the single-number-spine layout every binder in production actually uses. House style is 1. 2. 4. — Arabic, trailing stop, headers consume a number, children carry none.

A full closing-binder / offering-memorandum register (Times, centred masthead, thick-thin double rule, an INDEX cover label, dot-leader folios, a document-control foot zone, navy demoted to folios only): mocked up in full against real reference points and rejected. Folios are wrong for this product — every source document already carries its own page numbers, so a second binder-level number confuses readers, and no folios means no dot leaders (a leader exists only to carry the eye to a number). The grey highlight is load-bearing wayfinding, not a "generated-UI tell" to remove. A centred index block reads wrong under this masthead. The token cost (~10 changed + ~7 new, all requiring cosign) was disproportionate to move one component off the system's own font.pdf-nav.

Worth keeping from the exercise: paper.primitive.font.serif-legal already exists for exactly the legal-body-text role a Times-based binder would have wanted — the Paper pillar already splits Helvetica-for-navigation from Times-for-legal-bodies, which is the correct split for a generated artifact that must rebuild identically anywhere.

Open questions

Carried verbatim from the recipe so downstream consumers do not treat unresolved items as settled:

• oq-1 (resolved 2026-07-17) — phantom colour. An earlier draft listed a 7th colour GREY_MID ("inactive entry subtitles") that does not exist in the canonical Python source. Confirmed absent from every production instance; stays dropped.
• oq-2 — 8-entry ceiling. The max-8-TOC-entries-per-slip-sheet limit is currently hard-coded. Evidence from a real 8-row slip sheet shows the actual constraint is vertical space, not row count (a child row costs 26pt against a top-level row's 48pt) — recommend re-expressing the cap as a computed space budget; not yet done.
• oq-3 — Tagged-PDF conformance. Reading-order conformance for assistive technology is unhandled; source PDFs pass through untagged.
• oq-4 (resolved 2026-07-17) — navy hex. Pinned to #002e63, confirmed against the live Python source's exact RGB constant (0.0, 0.18, 0.39).
• oq-5 — no title wrapping. Nothing wraps or ellipsizes; a long title silently runs past its box's right edge. Should the recipe specify wrapping, or a documented max length (~400pt ≈ 66 characters at the child level)?
• oq-6 — doc-parented group depth. Should a doc-parented group ever nest more than one level? No production instance does; recommend capping at one until a real case exists.
• oq-7 — landscape slip sheets. All current instances are portrait. Right-margin geometry is already expressed via margin-right rather than a literal value, so a landscape variant is lower-risk than it would otherwise be — not built, not requested.