# V3 portfolio graph-label preservation · 2026-09-08

This receipt closes the focused TODO to preserve the readable labels on
`portfolio-v3-embodiment`. The graph remains unchanged: five compact title and
detail pairs stay inside their nodes, while the full V3 meaning remains in the
graph description, authored point descriptions and speaker notes. The
[authoring review](../../authoring/v3-graph-label-review.md) records the exact
contract and its provenance boundary.

## What is now enforced

- `v3-graph-labels.test.mjs` locks the five label pairs and checks the detailed
  V3 meaning across the graph description, slide descriptions and notes. Its
  two mutation cases prove that a long label or removed note detail fails.
- `v3-graph-labels.mjs` measures actual SVG text and rectangle geometry in
  Chromium. Every line must remain visible and inside a 210×76 node with the
  intended 16-unit horizontal insets. The tightest current line, **Artists and
  performance**, retains 7.77 design units before the right inset in the
  compact and embedded runs.
- The same browser check requires the complete ordered reading alternative and
  graph description at 390px and 320px. It checks standalone and embedded
  layouts, accessibility, page width, and source/artifact immutability.
- Five full authored explanations are injected into the SVG one at a time and
  must fail the node-bound audit. Reduced 12-unit type and hidden text must also
  fail. Each mutation is restored and the production graph is rechecked.

The graph and TODO were introduced together in
`fc4cd2de4039327b34e99ee27eef4076ab5a59d2`; its parent contains no V3 graph
source. Git therefore cannot supply the exact intermediate pre-repair strings.
The five long controls are explicitly the complete descriptions retained on
the same slide, not claimed reconstructions of unavailable history.

## Visual review

The [visual-review ledger](visual-review.json) binds eight inspected images:
standalone desktop, compact and 901px SVG states; 390px and 320px reading
states; embedded desktop and mobile; and the actual delivered PDF page 4. The
SVG labels remain contained and visually separated from connectors. Mobile
states preserve all five label pairs and the full graph description without
page-width overflow. The PDF retains the complete graph, label hierarchy,
takeaway and footer.

The embedded mobile image records its initial viewport. The complete lower
description, takeaway and controls are reviewed in the full-page standalone
mobile captures and validated by the browser report; the receipt does not
describe the entire tall slide as fitting in one phone viewport.

## Verification

All final checks run under the repository memory supervisor where applicable:

| Gate | Result |
| --- | --- |
| Focused source/meaning tests | 3 tests pass, including 2 mutations |
| Focused Playwright geometry | 7 captures and 7 rejected defects; pass |
| Complete portfolio guide | 12 slides; desktop, compact, mobile, accessibility, sources, PDF sequence and actual audio pass |
| Focused presentation center | 26 selected guides, 34 baseline-bound unselected guides, 444 embedded slide captures and 26 PDF sequences pass |
| Presentation Node suite | 281 tests pass |
| Presentation Python suite | 250 tests pass |
| Builder and narration freshness | 578 slides / 347 sources and 578 tracks / 14.36 hours pass |
| Teaching inventory freshness | 40 reviewed, 10,018 structural and 103,286 unassigned documentation units; 3,718 unassigned workflow units; current-scope checks pass |
| Repository inventory | 140,123 entries with no manifest parse or shape issues; pass |
| Formatting, lint, links and diff checks | Pass |

A portfolio-only center invocation is retained as an expected precondition
failure. Twenty-five other guide HTML files are new or changed after the last
canonical full-center receipt, so the harness rejected one of them before
launching the browser. The passing run includes those 25 guides plus
`portfolio-map`; this exercises the current files while preserving exact
baseline hashes for the other 34 guides. The expected failure is not counted as
presentation behavior evidence.

No graph source, slide copy, generated HTML, PDF, narration, catalog entry or
teaching assignment changed in this package. The integration baseline is
`a2baf7058d141a7e8ce7f610fabcb5d017d10fc6`; work began at
`e40e356d23e3ec05dc8994d5858c693ab9a83c1d` before the concurrent Eve guides
were integrated.

## Reproduce

Use the prepared presentation dependencies and run browser or broad test gates
serially:

```sh
node --test tools/presentations/tests/v3-graph-labels.test.mjs
OSHUN_V3_GRAPH_QA_DIR=/tmp/v3-graph CHROME_CHANNEL=chromium \
  python3 tools/presentations/run-supervised.py \
  node tools/presentations/tests/v3-graph-labels.mjs
OSHUN_DECK_QA_DIR=/tmp/portfolio-guide CHROME_CHANNEL=chromium \
  python3 tools/presentations/run-supervised.py \
  node tools/presentations/tests/eve-oshun-deck.mjs --guide portfolio-map
OSHUN_CENTER_QA_DIR=/tmp/portfolio-center CHROME_CHANNEL=chromium \
  python3 tools/presentations/run-supervised.py \
  node tools/presentations/tests/presentation-center.mjs --guides \
  eve-capabilities,eve-product-graph,eve-coverage-analysis,eve-work-tracking,eve-discussions-decisions,eve-delivery-tracking,eve-operator-analysis,eve-studio-analysis,eve-studio-actions,eve-feature-audits,eve-member-capabilities,eve-workflow-practice,eve-foundations,eve-invocation-detail,eve-context,eve-memory-detail,eve-trust,eve-safety-detail,eve-execution,eve-quality,eve-model-detail,eve-operations,eve-operations-detail,eve-current-runtime,v2-rollback-session,portfolio-map
```

`verify-receipt.py` checks current inputs, retained hashes and result structure,
reruns the fast regression, and proves that the graph, slide, generated guide,
PDF, narration and assignment tree remain unchanged from the baseline.

```sh
python3 docs/presentations/presentation-center/verification/v3-graph-labels-2026-09-08/verify-receipt.py
```

This receipt covers one graph-label regression. It does not establish detailed
V3 teaching coverage, normal-speed listening, reader comprehension or final
typography acceptance for other diagrams.
