- Embed IDC logo as base64 in two variants: white for the dark header
  band and color for the light footer (color variant derived from the
  white asset by region recoloring with brand palette).
- Add institutional footer: Jefatura de la Unidad de Investigacion 2026,
  Mg. Mario Quiroz Martinez, Instituto de Educacion Superior Diseno y
  Comunicacion.
- Exported dashboard.html now includes the same institutional footer
  with the white logo variant.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H2Xeoryp5wTqhJfaYt5NWN

- New 'Generar prompt del esquema' button (+ copy) in the configuration
  section: builds a focused LLM/deep-research prompt that produces the
  full classification schema with every indicated attribute — proyecto,
  descripcion, categorias, subcategorias sugeridas, enfoques, IPD label
  and levels, target context, category/approach inference signals,
  default approach per category, and countries.
- Add 'subcategorias_sugeridas' as a first-class config attribute:
  documented in the config template, populated in the DEMO, requested in
  the deep-research superprompt, and consumed in the editor as a
  per-category autocomplete datalist for the subcategoria field.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H2Xeoryp5wTqhJfaYt5NWN

Section 04 (coding editor):
- New 'Eliminar duplicados' button: detects duplicate references by DOI
  (or by normalized title + year) and keeps the most complete record of
  each group (coded > edited > LLM > inherited > imported > inferred >
  pending; tie-broken by filled metadata). Confirms count before
  removing and renumbers the corpus.
- Dominant-word rule for categories and enfoques: the highest-hierarchy
  word prevails when classifying (e.g. 'Innovacion pedagogica' ->
  'pedagogico'). Each category/enfoque cell shows its dominant word; a
  red asterisk flags an 'associated (not direct)' link when the term is
  compound. Dominant words are configurable via a new optional
  'palabras_dominantes' map (documented in the template, populated in the
  DEMO, and requested in the schema prompt); default falls back to the
  last significant word. Added an explanatory note in the section.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H2Xeoryp5wTqhJfaYt5NWN

- New 'Subcategoria' column between Categoria and Enfoque in the editor
  table, with a text input wired to the per-category subcategorias
  suggestions datalist (updates live when the category changes).
- Removed Subcategoria from the expandable metadata panel to avoid
  duplication; adjusted the metadata row colspan to 10 for the new
  11-column layout.
- Subcategoria edits persist through the 19-field CSV/JSON export.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H2Xeoryp5wTqhJfaYt5NWN

- New 'Autores y obras seminales' panel in the generated dashboard:
  card list of every reference flagged as seminal, showing authors (with
  a star), title, year, journal, category and DOI, sorted by year.
- Added an 'Obras seminales' KPI with the seminal count.
- Empty-state message prompts marking the 'Sem.' checkbox in the editor
  when no seminal references exist.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H2Xeoryp5wTqhJfaYt5NWN

- New floating 'Guia de botones' toggle: activates a help mode that
  outlines every actionable control and shows an explanatory popup
  (title + description) on hover, focus or tap.
- In help mode, clicks on any button are intercepted so the popup shows
  instead of executing the action, letting users safely explore what
  each control does without triggering downloads or dialogs.
- Added data-help text to all 23 controls across configuration, load,
  editor and export sections. Esc or toggling off restores normal use.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H2Xeoryp5wTqhJfaYt5NWN

Section 02 now documents the PDF/MD -> XML procedure and generates the
extraction prompt:
- Step-by-step guide: upload PDF/MD to a NotebookLM notebook, paste the
  extraction prompt, process in batches of 15 papers (controlling source
  count to avoid duplicates), move the table to Word/Excel, export to XML
  and load it here.
- 'Generar prompt para NotebookLM' button (+ copy) builds the full
  PRISMA/bibliometric extraction prompt, auto-filled from the active
  configuration: object of study, controlled vocabulary (categories,
  subcategories, approaches) and their dominant words as morphological
  stems with asterisk (e.g. creativ*, pedagog*). Falls back to
  [CORCHETES] placeholders when no config is applied.
- The prompt requests the 22-column table (19 fields + palabra_dominante
  cat/sub/enc) and 15-row batching. Importer now accepts the
  'relevancia_objeto' header as an alias of relevancia_tema; extra
  dominant-word columns are ignored on import.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H2Xeoryp5wTqhJfaYt5NWN