docrev 0.9.13 → 0.9.15

package/lib/build.ts

@@ -1,1251 +1,1300 @@
-/**
- * Build system - combines sections → paper.md → PDF/DOCX/TEX
- *
- * Features:
- * - Reads rev.yaml config
- * - Combines section files into paper.md (persisted)
- * - Strips annotations appropriately per output format
- * - Runs pandoc with crossref filter
- */
-
-import * as fs from 'fs';
-import * as path from 'path';
-import { execSync, spawn, ChildProcess } from 'child_process';
-import YAML from 'yaml';
-import { stripAnnotations } from './annotations.js';
-import { buildRegistry, labelToDisplay, detectDynamicRefs, resolveForwardRefs, resolveSupplementaryRefs } from './crossref.js';
-import { processVariables, hasVariables } from './variables.js';
-import { processSlideMarkdown, hasSlideSyntax } from './slides.js';
-import { generatePptxTemplate, templateNeedsRegeneration, injectMediaIntoPptx, injectSlideNumbers, applyThemeFonts, applyCentering, applyBuildupColors } from './pptx-template.js';
-import { getThemePath, getThemeNames, PPTX_THEMES } from './pptx-themes.js';
-import { runPostprocess } from './postprocess.js';
-import { hasPandoc, hasPandocCrossref, hasLatex } from './dependencies.js';
-import { buildImageRegistry, writeImageRegistry } from './image-registry.js';
-import type { Author, JournalFormatting } from './types.js';
-import { getJournalProfile } from './journals.js';
-import { resolveCSL } from './csl.js';
-
-// =============================================================================
-// Constants
-// =============================================================================
-
-/** Supported output formats */
-const SUPPORTED_FORMATS = ['pdf', 'docx', 'tex', 'beamer', 'pptx'] as const;
-
-/** Maximum title length for output filename */
-const MAX_TITLE_FILENAME_LENGTH = 50;
-
-// =============================================================================
-// Interfaces
-// =============================================================================
-
-export interface CrossrefConfig {
-  figureTitle?: string;
-  tableTitle?: string;
-  figPrefix?: string | string[];
-  tblPrefix?: string | string[];
-  secPrefix?: string | string[];
-  linkReferences?: boolean;
-}
-
-export interface PdfConfig {
-  template?: string | null;
-  headerIncludes?: string | null;
-  documentclass?: string;
-  fontsize?: string;
-  geometry?: string;
-  linestretch?: number;
-  numbersections?: boolean;
-  toc?: boolean;
-  /**
-   * LaTeX engine: pdflatex (default), xelatex, lualatex, tectonic, etc.
-   * xelatex/lualatex are required for native UTF-8 rendering of Latin-Extended
-   * diacritics (Czech/Polish/Croatian/Spanish author names, species epithets).
-   */
-  engine?: string;
-  /** Roman/serif main font (xelatex/lualatex only — uses fontspec). */
-  mainfont?: string;
-  /** Sans-serif font (xelatex/lualatex only). */
-  sansfont?: string;
-  /** Monospace font (xelatex/lualatex only). */
-  monofont?: string;
-}
-
-export interface DocxConfig {
-  reference?: string | null;
-  keepComments?: boolean;
-  toc?: boolean;
-}
-
-export interface TexConfig {
-  standalone?: boolean;
-}
-
-export interface BeamerConfig {
-  theme?: string;
-  colortheme?: string | null;
-  fonttheme?: string | null;
-  aspectratio?: string | null;
-  navigation?: string | null;
-  section?: boolean;
-  notes?: string | false;
-  fit_images?: boolean;
-}
-
-export interface PptxConfig {
-  theme?: string;
-  reference?: string | null;
-  media?: string | null;
-  colors?: {
-    default?: string;
-    title?: string;
-  };
-  buildup?: {
-    grey?: string;
-    accent?: string;
-    enabled?: boolean;
-  };
-}
-
-export interface TablesConfig {
-  nowrap?: string[];
-}
-
-export interface PostprocessConfig {
-  pdf?: string | null;
-  docx?: string | null;
-  tex?: string | null;
-  pptx?: string | null;
-  beamer?: string | null;
-  all?: string | null;
-  [key: string]: string | null | undefined;
-}
-
-export interface BuildConfig {
-  title: string;
-  authors: (string | Author)[];
-  affiliations: Record<string, string>;
-  sections: string[];
-  bibliography: string | null;
-  csl: string | null;
-  crossref: CrossrefConfig;
-  pdf: PdfConfig;
-  docx: DocxConfig;
-  tex: TexConfig;
-  beamer: BeamerConfig;
-  pptx: PptxConfig;
-  tables: TablesConfig;
-  postprocess: PostprocessConfig;
-  _configPath?: string | null;
-}
-
-export interface BuildResult {
-  format: string;
-  success: boolean;
-  outputPath?: string;
-  error?: string;
-}
-
-interface BuildOptions {
-  verbose?: boolean;
-  config?: BuildConfig;
-  outputPath?: string;
-  crossref?: boolean;
-  _refsAutoInjected?: boolean;
-  _forwardRefsResolved?: number;
-}
-
-interface CombineOptions extends BuildOptions {
-  _refsAutoInjected?: boolean;
-}
-
-interface VariablesContext {
-  sectionContents: string[];
-}
-
-interface PandocResult {
-  outputPath: string;
-  success: boolean;
-  error?: string;
-}
-
-interface FullBuildResult {
-  results: BuildResult[];
-  paperPath: string;
-  warnings: string[];
-  forwardRefsResolved: number;
-  refsAutoInjected?: boolean;
-}
-
-interface DynamicRef {
-  type: string;
-  label: string;
-  match: string;
-  position: number;
-}
-
-interface Registry {
-  figures: Map<string, unknown>;
-  tables: Map<string, unknown>;
-  equations: Map<string, unknown>;
-  byNumber: {
-    fig?: Map<number, string>;
-    figS?: Map<number, string>;
-    tbl?: Map<number, string>;
-    tblS?: Map<number, string>;
-    eq?: Map<number, string>;
-  };
-}
-
-/**
- * Default rev.yaml configuration
- */
-export const DEFAULT_CONFIG: BuildConfig = {
-  title: 'Untitled Document',
-  authors: [],
-  affiliations: {},
-  sections: [],
-  bibliography: null,
-  csl: null,
-  crossref: {
-    figureTitle: 'Figure',
-    tableTitle: 'Table',
-    figPrefix: ['Fig.', 'Figs.'],
-    tblPrefix: ['Table', 'Tables'],
-    secPrefix: ['Section', 'Sections'],
-    linkReferences: true,
-  },
-  pdf: {
-    template: null,
-    documentclass: 'article',
-    fontsize: '12pt',
-    geometry: 'margin=1in',
-    linestretch: 1.5,
-    numbersections: false,
-    toc: false,
-  },
-  docx: {
-    reference: null,
-    keepComments: true,
-    toc: false,
-  },
-  tex: {
-    standalone: true,
-  },
-  // Slide formats
-  beamer: {
-    theme: 'default',
-    colortheme: null,
-    fonttheme: null,
-    aspectratio: null, // '169' for 16:9, '43' for 4:3
-    navigation: null, // 'horizontal', 'vertical', 'frame', 'empty'
-    section: true, // section divider slides
-    notes: 'show', // 'show' (presenter view), 'only' (notes only), 'hide', or false
-    fit_images: true, // scale images to fit within slide bounds
-  },
-  pptx: {
-    theme: 'default', // Built-in theme: default, dark, academic, minimal, corporate
-    reference: null, // Custom reference-doc (overrides theme)
-    media: null, // directory with logo images (e.g., logo-left.png, logo-right.png)
-  },
-  // Table formatting
-  tables: {
-    nowrap: [], // Column headers to apply nowrap formatting (converts Normal() → $\mathcal{N}()$ etc.)
-  },
-  // Postprocess scripts
-  postprocess: {
-    pdf: null,
-    docx: null,
-    tex: null,
-    pptx: null,
-    beamer: null,
-    all: null, // Runs after any format
-  },
-};
-
-// =============================================================================
-// Public API
-// =============================================================================
-
-/**
- * Merge journal formatting defaults into a config.
- * Priority: DEFAULT_CONFIG < journal formatting < rev.yaml explicit settings
- */
-export function mergeJournalFormatting(config: BuildConfig, formatting: JournalFormatting, directory: string): BuildConfig {
-  const merged = { ...config };
-
-  // CSL: only apply if user hasn't set one
-  if (formatting.csl && !config.csl) {
-    const resolved = resolveCSL(formatting.csl, directory);
-    if (resolved) {
-      merged.csl = resolved;
-    }
-    // If not resolved locally, store the name — pandoc --citeproc
-    // can sometimes resolve it, and the user can fetch with rev profiles --fetch-csl
-    if (!resolved) {
-      merged.csl = formatting.csl;
-    }
-  }
-
-  // PDF settings: merge only unset fields
-  if (formatting.pdf) {
-    const userPdf = config.pdf || {};
-    const defaults = DEFAULT_CONFIG.pdf;
-    merged.pdf = { ...config.pdf };
-    for (const [key, value] of Object.entries(formatting.pdf)) {
-      const k = key as keyof PdfConfig;
-      // Apply journal value only if user config matches the default (i.e., wasn't explicitly set)
-      if (value !== undefined && JSON.stringify(userPdf[k]) === JSON.stringify(defaults[k])) {
-        (merged.pdf as Record<string, unknown>)[k] = value;
-      }
-    }
-  }
-
-  // DOCX settings: merge only unset fields
-  if (formatting.docx) {
-    const userDocx = config.docx || {};
-    const defaults = DEFAULT_CONFIG.docx;
-    merged.docx = { ...config.docx };
-    for (const [key, value] of Object.entries(formatting.docx)) {
-      const k = key as keyof DocxConfig;
-      if (value !== undefined && JSON.stringify(userDocx[k]) === JSON.stringify(defaults[k])) {
-        (merged.docx as Record<string, unknown>)[k] = value;
-      }
-    }
-  }
-
-  // Crossref settings: merge only unset fields
-  if (formatting.crossref) {
-    const userCrossref = config.crossref || {};
-    const defaults = DEFAULT_CONFIG.crossref;
-    merged.crossref = { ...config.crossref };
-    for (const [key, value] of Object.entries(formatting.crossref)) {
-      const k = key as keyof CrossrefConfig;
-      if (value !== undefined && JSON.stringify(userCrossref[k]) === JSON.stringify(defaults[k])) {
-        (merged.crossref as Record<string, unknown>)[k] = value;
-      }
-    }
-  }
-
-  return merged;
-}
-
-/**
- * Load rev.yaml config from directory
- * @param directory - Project directory path
- * @returns Merged config with defaults
- * @throws {TypeError} If directory is not a string
- * @throws {Error} If rev.yaml exists but cannot be parsed
- */
-export function loadConfig(directory: string): BuildConfig {
-  if (typeof directory !== 'string') {
-    throw new TypeError(`directory must be a string, got ${typeof directory}`);
-  }
-
-  const configPath = path.join(directory, 'rev.yaml');
-
-  if (!fs.existsSync(configPath)) {
-    return { ...DEFAULT_CONFIG, _configPath: null };
-  }
-
-  try {
-    const content = fs.readFileSync(configPath, 'utf-8');
-    const userConfig = YAML.parse(content) || {};
-
-    // Deep merge with defaults
-    let config: BuildConfig = {
-      ...DEFAULT_CONFIG,
-      ...userConfig,
-      crossref: { ...DEFAULT_CONFIG.crossref, ...userConfig.crossref },
-      pdf: { ...DEFAULT_CONFIG.pdf, ...userConfig.pdf },
-      docx: { ...DEFAULT_CONFIG.docx, ...userConfig.docx },
-      tex: { ...DEFAULT_CONFIG.tex, ...userConfig.tex },
-      beamer: { ...DEFAULT_CONFIG.beamer, ...userConfig.beamer },
-      pptx: { ...DEFAULT_CONFIG.pptx, ...userConfig.pptx },
-      tables: { ...DEFAULT_CONFIG.tables, ...userConfig.tables },
-      postprocess: { ...DEFAULT_CONFIG.postprocess, ...userConfig.postprocess },
-      _configPath: configPath,
-    };
-
-    // Apply journal formatting defaults (between DEFAULT_CONFIG and user settings)
-    if (userConfig.journal) {
-      const profile = getJournalProfile(userConfig.journal);
-      if (profile?.formatting) {
-        config = mergeJournalFormatting(config, profile.formatting, directory);
-      }
-    }
-
-    return config;
-  } catch (err) {
-    const error = err as Error;
-    throw new Error(`Failed to parse rev.yaml: ${error.message}`);
-  }
-}
-
-/**
- * Find section files in directory
- * @param directory - Project directory path
- * @param configSections - Sections from rev.yaml (optional)
- * @returns Ordered list of section file names
- * @throws {TypeError} If directory is not a string
- */
-export function findSections(directory: string, configSections: string[] = []): string[] {
-  if (typeof directory !== 'string') {
-    throw new TypeError(`directory must be a string, got ${typeof directory}`);
-  }
-
-  // If sections specified in config, use that order
-  if (configSections.length > 0) {
-    const sections: string[] = [];
-    for (const section of configSections) {
-      const filePath = path.join(directory, section);
-      if (fs.existsSync(filePath)) {
-        sections.push(section);
-      } else {
-        console.warn(`Warning: Section file not found: ${section}`);
-      }
-    }
-    return sections;
-  }
-
-  // Try sections.yaml
-  const sectionsYamlPath = path.join(directory, 'sections.yaml');
-  if (fs.existsSync(sectionsYamlPath)) {
-    try {
-      const sectionsConfig = YAML.parse(fs.readFileSync(sectionsYamlPath, 'utf-8'));
-      if (sectionsConfig.sections) {
-        return Object.entries(sectionsConfig.sections)
-          .sort((a: [string, any], b: [string, any]) => (a[1].order ?? 999) - (b[1].order ?? 999))
-          .map(([file]) => file)
-          .filter((f) => fs.existsSync(path.join(directory, f)));
-      }
-    } catch (e) {
-      if (process.env.DEBUG) {
-        const error = e as Error;
-        console.warn('build: YAML parse error in sections.yaml:', error.message);
-      }
-    }
-  }
-
-  // Default: find all .md files except special ones
-  const exclude = ['paper.md', 'readme.md', 'claude.md'];
-  const files = fs.readdirSync(directory).filter((f) => {
-    if (!f.endsWith('.md')) return false;
-    if (exclude.includes(f.toLowerCase())) return false;
-    return true;
-  });
-
-  // Sort alphabetically as fallback
-  return files.sort();
-}
-
-/**
- * Combine section files into paper.md
- */
-export function combineSections(directory: string, config: BuildConfig, options: CombineOptions = {}): string {
-  const sections = findSections(directory, config.sections);
-
-  if (sections.length === 0) {
-    throw new Error('No section files found. Create .md files or specify sections in rev.yaml');
-  }
-
-  const parts: string[] = [];
-
-  // Add YAML frontmatter
-  const frontmatter = buildFrontmatter(config);
-  parts.push('---');
-  parts.push(YAML.stringify(frontmatter).trim());
-  parts.push('---');
-  parts.push('');
-
-  // Read all section contents for variable processing
-  const sectionContents: string[] = [];
-
-  // Check if we need to auto-inject references before supplementary
-  // Pandoc places refs at the end by default, which breaks when supplementary follows
-  const hasRefsSection = sections.some(s =>
-    s.toLowerCase().includes('reference') || s.toLowerCase().includes('refs')
-  );
-  const suppIndex = sections.findIndex(s =>
-    s.toLowerCase().includes('supp') || s.toLowerCase().includes('appendix')
-  );
-  const hasBibliography = config.bibliography && fs.existsSync(path.join(directory, config.bibliography));
-
-  // Track if we find an explicit refs div in any section
-  let hasExplicitRefsDiv = false;
-
-  // Combine sections
-  for (let i = 0; i < sections.length; i++) {
-    const section = sections[i];
-    if (!section) continue;
-    const filePath = path.join(directory, section);
-    let content = fs.readFileSync(filePath, 'utf-8');
-
-    // Remove any existing frontmatter from section files
-    content = stripFrontmatter(content);
-    sectionContents.push(content);
-
-    // Check if this section has an explicit refs div
-    if (content.includes('::: {#refs}') || content.includes(':::  {#refs}')) {
-      hasExplicitRefsDiv = true;
-    }
-
-    // Auto-inject references before supplementary if needed
-    if (i === suppIndex && hasBibliography && !hasRefsSection && !hasExplicitRefsDiv) {
-      parts.push('# References\n');
-      parts.push('::: {#refs}');
-      parts.push(':::');
-      parts.push('');
-      parts.push('');
-      options._refsAutoInjected = true;
-    }
-
-    parts.push(content.trim());
-    parts.push('');
-    parts.push(''); // Double newline between sections
-  }
-
-  let paperContent = parts.join('\n');
-
-  // Process template variables if any exist
-  if (hasVariables(paperContent)) {
-    paperContent = processVariables(paperContent, config as any, { sectionContents });
-  }
-
-  // Resolve forward references (refs that appear before their anchor definition)
-  // This fixes pandoc-crossref limitation with multi-file documents
-  if (hasPandocCrossref()) {
-    const registry = buildRegistry(directory, sections);
-    const { text, resolved } = resolveForwardRefs(paperContent, registry);
-    if (resolved.length > 0) {
-      paperContent = text;
-      // Store resolved count for optional reporting
-      options._forwardRefsResolved = resolved.length;
-    }
-
-    // Resolve supplementary references and strip their anchors.
-    // pandoc-crossref cannot produce "Figure S1" numbering — it numbers all
-    // figures sequentially. We resolve supplementary refs to plain text and
-    // remove the {#fig:...} attributes so crossref ignores them.
-    const supp = resolveSupplementaryRefs(paperContent, registry);
-    if (supp.resolved.length > 0) {
-      paperContent = supp.text;
-    }
-  }
-
-  const paperPath = path.join(directory, 'paper.md');
-
-  fs.writeFileSync(paperPath, paperContent, 'utf-8');
-
-  return paperPath;
-}
-
-/**
- * Build YAML frontmatter from config
- */
-function buildFrontmatter(config: BuildConfig): Record<string, unknown> {
-  const fm: Record<string, unknown> = {};
-
-  if (config.title) fm.title = config.title;
-
-  // Skip author in frontmatter when using numbered affiliations —
-  // the author block is injected separately per format
-  if (config.authors && config.authors.length > 0 && !hasNumberedAffiliations(config)) {
-    fm.author = config.authors;
-  }
-
-  if (config.bibliography) {
-    fm.bibliography = config.bibliography;
-  }
-
-  if (config.csl) {
-    fm.csl = config.csl;
-  }
-
-  return fm;
-}
-
-/**
- * Strip YAML frontmatter from content
- */
-function stripFrontmatter(content: string): string {
-  const match = content.match(/^---\r?\n([\s\S]*?)\r?\n---\r?\n/);
-  if (match) {
-    return content.slice(match[0].length);
-  }
-  return content;
-}
-
-/**
- * Check if config uses numbered affiliation mode
- * (authors have `affiliations` arrays and an affiliations map is defined)
- */
-function hasNumberedAffiliations(config: BuildConfig): boolean {
-  if (!config.affiliations || Object.keys(config.affiliations).length === 0) return false;
-  return config.authors.some(a => typeof a !== 'string' && a.affiliations && a.affiliations.length > 0);
-}
-
-/**
- * Generate LaTeX author block using authblk package for numbered superscript affiliations.
- * Returns LaTeX code to be injected via header-includes.
- */
-function generateLatexAuthorBlock(config: BuildConfig): string {
-  const lines: string[] = [];
-  lines.push('\\usepackage{authblk}');
-  lines.push('\\renewcommand\\Authfont{\\normalsize}');
-  lines.push('\\renewcommand\\Affilfont{\\small}');
-  lines.push('');
-
-  // Map affiliation keys to numbers
-  const affiliationKeys = Object.keys(config.affiliations);
-  const keyToNum = new Map<string, number>();
-  affiliationKeys.forEach((key, i) => keyToNum.set(key, i + 1));
-
-  // Authors
-  for (const author of config.authors) {
-    if (typeof author === 'string') {
-      lines.push(`\\author{${author}}`);
-      continue;
-    }
-    const marks = (author.affiliations || [])
-      .map(k => keyToNum.get(k))
-      .filter((n): n is number => n !== undefined);
-
-    const markStr = marks.length > 0 ? `[${marks.join(',')}]` : '';
-    let nameStr = author.name;
-    if (author.corresponding && author.email) {
-      nameStr += `\\thanks{Corresponding author: ${author.email}}`;
-    } else if (author.corresponding) {
-      nameStr += '\\thanks{Corresponding author}';
-    }
-    lines.push(`\\author${markStr}{${nameStr}}`);
-  }
-
-  // Affiliations
-  for (const [key, text] of Object.entries(config.affiliations)) {
-    const num = keyToNum.get(key);
-    if (num !== undefined) {
-      lines.push(`\\affil[${num}]{${text}}`);
-    }
-  }
-
-  return lines.join('\n');
-}
-
-/**
- * Generate markdown author block for DOCX output with superscript affiliations.
- * Returns markdown text to insert after the YAML frontmatter.
- */
-function generateMarkdownAuthorBlock(config: BuildConfig): string {
-  const lines: string[] = [];
-
-  // Map affiliation keys to numbers
-  const affiliationKeys = Object.keys(config.affiliations);
-  const keyToNum = new Map<string, number>();
-  affiliationKeys.forEach((key, i) => keyToNum.set(key, i + 1));
-
-  // Author line: Name^1,2^, Name^3^, ...
-  const authorParts: string[] = [];
-  for (const author of config.authors) {
-    if (typeof author === 'string') {
-      authorParts.push(author);
-      continue;
-    }
-    const marks = (author.affiliations || [])
-      .map(k => keyToNum.get(k))
-      .filter((n): n is number => n !== undefined);
-    let entry = author.name;
-    const superParts = marks.map(String);
-    if (author.corresponding) superParts.push('\\*');
-    if (superParts.length > 0) {
-      entry += `^${superParts.join(',')}^`;
-    }
-    authorParts.push(entry);
-  }
-  lines.push(authorParts.join(', '));
-  lines.push('');
-
-  // Affiliation lines: ^1^ Department of ...
-  for (const [key, text] of Object.entries(config.affiliations)) {
-    const num = keyToNum.get(key);
-    if (num !== undefined) {
-      lines.push(`^${num}^ ${text}`);
-    }
-  }
-
-  // Corresponding author footnote
-  const corresponding = config.authors.find(a => typeof a !== 'string' && a.corresponding) as Author | undefined;
-  if (corresponding?.email) {
-    lines.push('');
-    lines.push(`^\\*^ Corresponding author: ${corresponding.email}`);
-  }
-
-  lines.push('');
-  return lines.join('\n');
-}
-
-/**
- * Process markdown tables to apply nowrap formatting to specified columns.
- * Converts distribution notation (Normal, Student-t, Gamma) to LaTeX math.
- * @param content - Markdown content
- * @param tablesConfig - tables config from rev.yaml
- * @param format - output format (pdf, docx, etc.)
- * @returns processed content
- */
-export function processTablesForFormat(content: string, tablesConfig: TablesConfig, format: string): string {
-  // Only process for PDF/TeX output
-  if (format !== 'pdf' && format !== 'tex') {
-    return content;
-  }
-
-  // Check if we have nowrap columns configured
-  if (!tablesConfig?.nowrap?.length) {
-    return content;
-  }
-
-  const nowrapPatterns = tablesConfig.nowrap.map((p) => p.toLowerCase());
-
-  // Match pipe tables: header row, separator row, body rows
-  // Header: | Col1 | Col2 | Col3 |
-  // Separator: |:-----|:-----|:-----|
-  // Body: | val1 | val2 | val3 |
-  const tableRegex = /^(\|[^\n]+\|\r?\n\|[-:| ]+\|\r?\n)((?:\|[^\n]+\|\r?\n?)+)/gm;
-
-  return content.replace(tableRegex, (match, headerAndSep, body) => {
-    // Split header from separator
-    const lines = headerAndSep.split(/\r?\n/);
-    const headerLine = lines[0] ?? '';
-
-    // Parse header cells to find nowrap column indices
-    const headerCells = headerLine
-      .split('|')
-      .slice(1, -1)
-      .map((c: string) => c.trim().toLowerCase());
-
-    const nowrapCols: number[] = [];
-    headerCells.forEach((cell: string, i: number) => {
-      if (nowrapPatterns.some((p) => cell.includes(p))) {
-        nowrapCols.push(i);
-      }
-    });
-
-    // If no nowrap columns found in this table, return unchanged
-    if (nowrapCols.length === 0) {
-      return match;
-    }
-
-    // Process body rows
-    const bodyLines = body.split(/\r?\n/).filter((l: string) => l.trim());
-    const processedBody = bodyLines
-      .map((row: string) => {
-        // Split row into cells, keeping the pipe structure
-        const cells = row.split('|');
-        // cells[0] is empty (before first |), cells[last] is empty (after last |)
-
-        nowrapCols.forEach((colIdx) => {
-          const cellIdx = colIdx + 1; // Account for empty first element
-          if (cells[cellIdx] !== undefined) {
-            const cellContent = cells[cellIdx].trim();
-
-            // Skip if empty, already math, or already has LaTeX commands
-            if (!cellContent || cellContent.startsWith('$') || cellContent.startsWith('\\')) {
-              return;
-            }
-
-            // Convert distribution notation to LaTeX math
-            // Order matters: compound names (Half-Normal) must come before simple names (Normal)
-            let processed = cellContent;
-
-            // Half-Normal(x) → $\text{Half-Normal}(x)$ (must come before Normal)
-            processed = processed.replace(/Half-Normal\(([^)]+)\)/g, '$\\text{Half-Normal}($1)$');
-
-            // Normal(x, y) → $\mathcal{N}(x, y)$
-            processed = processed.replace(/Normal\(([^)]+)\)/g, '$\\mathcal{N}($1)$');
-
-            // Student-t(df, loc, scale) → $t_{df}(loc, scale)$
-            processed = processed.replace(/Student-t\((\d+),\s*([^)]+)\)/g, '$t_{$1}($2)$');
-
-            // Gamma(a, b) → $\text{Gamma}(a, b)$
-            processed = processed.replace(/Gamma\(([^)]+)\)/g, '$\\text{Gamma}($1)$');
-
-            // Exponential(x) → $\text{Exp}(x)$
-            processed = processed.replace(/Exponential\(([^)]+)\)/g, '$\\text{Exp}($1)$');
-
-            // Update cell with padding
-            cells[cellIdx] = ` ${processed} `;
-          }
-        });
-
-        return cells.join('|');
-      })
-      .join('\n');
-
-    return headerAndSep + processedBody + '\n';
-  });
-}
-
-/**
- * Prepare paper.md for specific output format
- */
-export function prepareForFormat(
-  paperPath: string,
-  format: string,
-  config: BuildConfig,
-  options: BuildOptions = {}
-): string {
-  const directory = path.dirname(paperPath);
-  let content = fs.readFileSync(paperPath, 'utf-8');
-
-  // Build crossref registry for reference conversion
-  // Pass sections from config to ensure correct file ordering
-  const registry = buildRegistry(directory, config.sections);
-
-  if (format === 'pdf' || format === 'tex') {
-    // Strip all annotations for clean output
-    content = stripAnnotations(content);
-
-    // Process tables for nowrap columns (convert Normal() → $\mathcal{N}()$ etc.)
-    content = processTablesForFormat(content, config.tables, format);
-
-    // Inject LaTeX author block with numbered affiliations
-    if (hasNumberedAffiliations(config)) {
-      const latexBlock = generateLatexAuthorBlock(config);
-      // Inject as header-includes in the YAML frontmatter
-      content = content.replace(/^(---\r?\n[\s\S]*?)(---\r?\n)/, (match, yamlContent, closing) => {
-        return `${yamlContent}header-includes: |\n${latexBlock.split('\n').map(l => '  ' + l).join('\n')}\n${closing}`;
-      });
-    }
-  } else if (format === 'docx') {
-    // Strip track changes, optionally keep comments
-    content = stripAnnotations(content, { keepComments: config.docx.keepComments });
-
-    // Convert @fig:label to "Figure 1" for Word readers
-    content = convertDynamicRefsToDisplay(content, registry);
-
-    // Inject markdown author block with superscript affiliations
-    if (hasNumberedAffiliations(config)) {
-      const mdBlock = generateMarkdownAuthorBlock(config);
-      // Insert after YAML frontmatter, before body content
-      content = content.replace(/^(---\r?\n[\s\S]*?---\r?\n)/, `$1\n${mdBlock}\n`);
-    }
-  } else if (format === 'beamer' || format === 'pptx') {
-    // Strip annotations for slide output
-    content = stripAnnotations(content);
-
-    // Process slide syntax (::: step, ::: notes)
-    if (hasSlideSyntax(content)) {
-      content = processSlideMarkdown(content, format);
-    }
-  }
-
-  // Write to temporary file
-  const preparedPath = path.join(directory, `.paper-${format}.md`);
-  fs.writeFileSync(preparedPath, content, 'utf-8');
-
-  return preparedPath;
-}
-
-/**
- * Convert @fig:label references to display format (Figure 1)
- */
-function convertDynamicRefsToDisplay(text: string, registry: Registry): string {
-  const refs = detectDynamicRefs(text);
-
-  // Process in reverse order to preserve positions
-  let result = text;
-  for (let i = refs.length - 1; i >= 0; i--) {
-    const ref = refs[i];
-    if (!ref) continue;
-    const display = labelToDisplay(ref.type, ref.label, registry as any);
-
-    if (display) {
-      result = result.slice(0, ref.position) + display + result.slice(ref.position + ref.match.length);
-    }
-  }
-
-  return result;
-}
-
-/**
- * Build pandoc arguments for format
- */
-export function buildPandocArgs(format: string, config: BuildConfig, outputPath: string): string[] {
-  const args: string[] = [];
-
-  // Output format
-  if (format === 'tex') {
-    args.push('-t', 'latex');
-    if (config.tex.standalone) {
-      args.push('-s');
-    }
-  } else if (format === 'pdf') {
-    args.push('-t', 'pdf');
-  } else if (format === 'docx') {
-    args.push('-t', 'docx');
-  } else if (format === 'beamer') {
-    args.push('-t', 'beamer');
-  } else if (format === 'pptx') {
-    args.push('-t', 'pptx');
-  }
-
-  // Output file (use basename since we set cwd to directory in runPandoc)
-  args.push('-o', path.basename(outputPath));
-
-  // Crossref filter (if available) - skip for slides
-  if (hasPandocCrossref() && format !== 'beamer' && format !== 'pptx') {
-    args.push('--filter', 'pandoc-crossref');
-  }
-
-  // Bibliography
-  if (config.bibliography) {
-    args.push('--citeproc');
-  }
-
-  // Format-specific options
-  if (format === 'pdf') {
-    if (config.pdf.template) {
-      args.push('--template', config.pdf.template);
-    }
-    if (config.pdf.engine) {
-      args.push(`--pdf-engine=${config.pdf.engine}`);
-    }
-    if (config.pdf.mainfont) {
-      args.push('-V', `mainfont=${config.pdf.mainfont}`);
-    }
-    if (config.pdf.sansfont) {
-      args.push('-V', `sansfont=${config.pdf.sansfont}`);
-    }
-    if (config.pdf.monofont) {
-      args.push('-V', `monofont=${config.pdf.monofont}`);
-    }
-    args.push('-V', `documentclass=${config.pdf.documentclass}`);
-    args.push('-V', `fontsize=${config.pdf.fontsize}`);
-    args.push('-V', `geometry:${config.pdf.geometry}`);
-    if (config.pdf.headerIncludes) {
-      args.push('-H', config.pdf.headerIncludes);
-    }
-    if (config.pdf.linestretch !== 1) {
-      args.push('-V', `linestretch=${config.pdf.linestretch}`);
-    }
-    if (config.pdf.numbersections) {
-      args.push('--number-sections');
-    }
-    if (config.pdf.toc) {
-      args.push('--toc');
-    }
-  } else if (format === 'docx') {
-    if (config.docx.reference) {
-      args.push('--reference-doc', config.docx.reference);
-    }
-    if (config.docx.toc) {
-      args.push('--toc');
-    }
-  } else if (format === 'beamer') {
-    // Beamer slide options
-    const beamer = config.beamer || {};
-    if (beamer.theme) {
-      args.push('-V', `theme=${beamer.theme}`);
-    }
-    if (beamer.colortheme) {
-      args.push('-V', `colortheme=${beamer.colortheme}`);
-    }
-    if (beamer.fonttheme) {
-      args.push('-V', `fonttheme=${beamer.fonttheme}`);
-    }
-    if (beamer.aspectratio) {
-      args.push('-V', `aspectratio=${beamer.aspectratio}`);
-    }
-    if (beamer.navigation) {
-      args.push('-V', `navigation=${beamer.navigation}`);
-    }
-    // Speaker notes - default to 'show' which creates presenter view PDF
-    // Options: 'show' (dual screen), 'only' (notes only), 'hide' (no notes), false (disabled)
-    const notesMode = beamer.notes !== undefined ? beamer.notes : 'show';
-    if (notesMode && notesMode !== 'hide') {
-      args.push('-V', `classoption=notes=${notesMode}`);
-    }
-    // Fit images within slide bounds (default: true)
-    if (beamer.fit_images !== false) {
-      const fitImagesHeader = `\\makeatletter
-\\def\\maxwidth{\\ifdim\\Gin@nat@width>\\linewidth\\linewidth\\else\\Gin@nat@width\\fi}
-\\def\\maxheight{\\ifdim\\Gin@nat@height>0.75\\textheight 0.75\\textheight\\else\\Gin@nat@height\\fi}
-\\makeatother
-\\setkeys{Gin}{width=\\maxwidth,height=\\maxheight,keepaspectratio}`;
-      args.push('-V', `header-includes=${fitImagesHeader}`);
-    }
-    // Slides need standalone
-    args.push('-s');
-  } else if (format === 'pptx') {
-    // PowerPoint options - handled separately in preparePptxTemplate
-    // Reference doc is set by caller after template generation
-  }
-
-  return args;
-}
-
-/**
- * Write crossref.yaml if needed
- */
-function ensureCrossrefConfig(directory: string, config: BuildConfig): void {
-  const crossrefPath = path.join(directory, 'crossref.yaml');
-
-  if (!fs.existsSync(crossrefPath) && hasPandocCrossref()) {
-    fs.writeFileSync(crossrefPath, YAML.stringify(config.crossref), 'utf-8');
-  }
-}
-
-/**
- * Get install instructions for missing dependency
- */
-function getInstallInstructions(tool: string): string {
-  const instructions: Record<string, string> = {
-    pandoc: 'https://pandoc.org/installing.html',
-    latex: 'https://www.latex-project.org/get/',
-  };
-  return instructions[tool] || 'Check documentation';
-}
-
-/**
- * Run pandoc build
- */
-export async function runPandoc(
-  inputPath: string,
-  format: string,
-  config: BuildConfig,
-  options: BuildOptions = {}
-): Promise<PandocResult> {
-  const directory = path.dirname(inputPath);
-  const baseName = config.title
-    ? config.title.toLowerCase().replace(/[^a-z0-9]+/g, '-').slice(0, 50)
-    : 'paper';
-
-  // Map format to file extension
-  const extMap: Record<string, string> = {
-    tex: '.tex',
-    pdf: '.pdf',
-    docx: '.docx',
-    beamer: '.pdf', // beamer outputs PDF
-    pptx: '.pptx',
-  };
-  const ext = extMap[format] || '.pdf';
-
-  // For beamer, use -slides suffix to distinguish from regular PDF
-  const suffix = format === 'beamer' ? '-slides' : '';
-  // Allow custom output path via options
-  const outputPath = options.outputPath || path.join(directory, `${baseName}${suffix}${ext}`);
-
-  // Ensure crossref.yaml exists
-  ensureCrossrefConfig(directory, config);
-
-  const args = buildPandocArgs(format, config, outputPath);
-
-  // Handle PPTX reference template and themes
-  let pptxMediaDir: string | null = null;
-  if (format === 'pptx') {
-    const pptx = config.pptx || {};
-
-    // Determine media directory (default: pptx/media or slides/media)
-    let mediaDir = pptx.media;
-    if (!mediaDir) {
-      if (fs.existsSync(path.join(directory, 'pptx', 'media'))) {
-        mediaDir = path.join(directory, 'pptx', 'media');
-      } else if (fs.existsSync(path.join(directory, 'slides', 'media'))) {
-        mediaDir = path.join(directory, 'slides', 'media');
-      }
-    } else if (!path.isAbsolute(mediaDir)) {
-      mediaDir = path.join(directory, mediaDir);
-    }
-    pptxMediaDir = mediaDir || null;
-
-    // Determine reference doc: custom reference overrides theme
-    let referenceDoc: string | null = null;
-    if (pptx.reference && fs.existsSync(path.join(directory, pptx.reference))) {
-      // Custom reference doc takes precedence
-      referenceDoc = path.join(directory, pptx.reference);
-    } else {
-      // Use built-in theme (default: 'default')
-      const themeName = pptx.theme || 'default';
-      const themePath = getThemePath(themeName);
-      if (themePath && fs.existsSync(themePath)) {
-        referenceDoc = themePath;
-      }
-    }
-
-    if (referenceDoc) {
-      args.push('--reference-doc', referenceDoc);
-    }
-
-    // Add color filter for PPTX (handles [text]{color=#RRGGBB} syntax)
-    const colorFilterPath = path.join(path.dirname(new URL(import.meta.url).pathname.replace(/^\/([A-Z]:)/, '$1')), 'pptx-color-filter.lua');
-    if (fs.existsSync(colorFilterPath)) {
-      args.push('--lua-filter', colorFilterPath);
-    }
-  }
-
-  // Add crossref metadata file if exists (skip for slides - they don't use crossref)
-  if (format !== 'beamer' && format !== 'pptx') {
-    const crossrefPath = path.join(directory, 'crossref.yaml');
-    if (fs.existsSync(crossrefPath) && hasPandocCrossref()) {
-      // Use basename since we set cwd to directory
-      args.push('--metadata-file', 'crossref.yaml');
-    }
-  }
-
-  // Input file (use basename since we set cwd to directory)
-  args.push(path.basename(inputPath));
-
-  return new Promise((resolve) => {
-    const pandoc: ChildProcess = spawn('pandoc', args, {
-      cwd: directory,
-      stdio: ['ignore', 'pipe', 'pipe'],
-    });
-
-    let stderr = '';
-    pandoc.stderr?.on('data', (data) => {
-      stderr += data.toString();
-    });
-
-    pandoc.on('close', async (code) => {
-      if (code === 0) {
-        // For PPTX, post-process to add slide numbers, buildup colors, and logos
-        if (format === 'pptx') {
-          try {
-            // Inject slide numbers into content slides only
-            await injectSlideNumbers(outputPath);
-          } catch (e) {
-            // Slide number injection failed but output was created
-          }
-          try {
-            // Apply colors (default text color, title color, buildup greying)
-            const pptxConfig = config.pptx || {};
-            const colorsConfig = pptxConfig.colors || {};
-            const buildupConfig = pptxConfig.buildup || {};
-            // Merge colors and buildup config for applyBuildupColors
-            const colorConfig = {
-              default: colorsConfig.default,
-              title: colorsConfig.title,
-              grey: buildupConfig.grey,
-              accent: buildupConfig.accent,
-              enabled: buildupConfig.enabled
-            };
-            await applyBuildupColors(outputPath, colorConfig);
-          } catch (e) {
-            // Color application failed but output was created
-          }
-          // Inject logos into cover slide (if media dir configured)
-          if (pptxMediaDir) {
-            try {
-              await injectMediaIntoPptx(outputPath, pptxMediaDir);
-            } catch (e) {
-              // Logo injection failed but output was created
-            }
-          }
-        }
-
-        // Run user postprocess scripts
-        const postResult = await runPostprocess(outputPath, format, config as unknown as Parameters<typeof runPostprocess>[2], options);
-        if (!postResult.success && options.verbose) {
-          console.error(`Postprocess warning: ${postResult.error}`);
-        }
-
-        resolve({ outputPath, success: true });
-      } else {
-        resolve({ outputPath, success: false, error: stderr || `Exit code ${code}` });
-      }
-    });
-
-    pandoc.on('error', (err) => {
-      resolve({ outputPath, success: false, error: err.message });
-    });
-  });
-}
-
-/**
- * Full build pipeline
- */
-export async function build(
-  directory: string,
-  formats: string[] = ['pdf', 'docx'],
-  options: BuildOptions = {}
-): Promise<FullBuildResult> {
-  const warnings: string[] = [];
-  let forwardRefsResolved = 0;
-
-  // Check pandoc
-  if (!hasPandoc()) {
-    const instruction = getInstallInstructions('pandoc');
-    throw new Error(`Pandoc not found. Install with: ${instruction}\nOr run: rev doctor`);
-  }
-
-  // Check LaTeX if PDF is requested
-  if ((formats.includes('pdf') || formats.includes('all')) && !hasLatex()) {
-    warnings.push(`LaTeX not found - PDF generation may fail. Install with: ${getInstallInstructions('latex')}`);
-  }
-
-  // Check pandoc-crossref
-  if (!hasPandocCrossref()) {
-    warnings.push('pandoc-crossref not found - figure/table numbering will not work');
-  }
-
-  // Load config (use passed config if provided, otherwise load from file)
-  const config = options.config || loadConfig(directory);
-
-  // Combine sections → paper.md
-  const buildOptions: CombineOptions = { ...options };
-  const paperPath = combineSections(directory, config, buildOptions);
-  forwardRefsResolved = buildOptions._forwardRefsResolved || 0;
-  const refsAutoInjected = buildOptions._refsAutoInjected || false;
-
-  // Expand 'all' to all formats
-  if (formats.includes('all')) {
-    formats = ['pdf', 'docx', 'tex'];
-  }
-
-  // Build and save image registry when DOCX is being built
-  // This allows import to restore proper image syntax from Word documents
-  if (formats.includes('docx')) {
-    const paperContent = fs.readFileSync(paperPath, 'utf-8');
-    const crossrefReg = buildRegistry(directory, config.sections);
-    const imageReg = buildImageRegistry(paperContent, crossrefReg as any);
-    if ((imageReg as any).figures?.length > 0) {
-      writeImageRegistry(directory, imageReg);
-    }
-  }
-
-  const results: BuildResult[] = [];
-
-  for (const format of formats) {
-    // Prepare format-specific version
-    const preparedPath = prepareForFormat(paperPath, format, config, options);
-
-    // Run pandoc
-    const result = await runPandoc(preparedPath, format, config, options);
-    results.push({ format, ...result });
-
-    // Clean up temp file
-    try {
-      fs.unlinkSync(preparedPath);
-    } catch {
-      // Ignore cleanup errors
-    }
-  }
-
-  return { results, paperPath, warnings, forwardRefsResolved, refsAutoInjected };
-}
-
-/**
- * Get build status summary
- */
-export function formatBuildResults(results: BuildResult[]): string {
-  const lines: string[] = [];
-
-  for (const r of results) {
-    if (r.success) {
-      lines.push(`  ${r.format.toUpperCase()}: ${path.basename(r.outputPath!)}`);
-    } else {
-      lines.push(`  ${r.format.toUpperCase()}: FAILED - ${r.error}`);
-    }
-  }
-
-  return lines.join('\n');
-}
+/**
+ * Build system - combines sections → paper.md → PDF/DOCX/TEX
+ *
+ * Features:
+ * - Reads rev.yaml config
+ * - Combines section files into paper.md (persisted)
+ * - Strips annotations appropriately per output format
+ * - Runs pandoc with crossref filter
+ */
+
+import * as fs from 'fs';
+import * as path from 'path';
+import { execSync, spawn, ChildProcess } from 'child_process';
+import YAML from 'yaml';
+import { stripAnnotations } from './annotations.js';
+import { buildRegistry, labelToDisplay, detectDynamicRefs, resolveForwardRefs, resolveSupplementaryRefs } from './crossref.js';
+import { processVariables, hasVariables } from './variables.js';
+import { processSlideMarkdown, hasSlideSyntax } from './slides.js';
+import { generatePptxTemplate, templateNeedsRegeneration, injectMediaIntoPptx, injectSlideNumbers, applyThemeFonts, applyCentering, applyBuildupColors } from './pptx-template.js';
+import { getThemePath, getThemeNames, PPTX_THEMES } from './pptx-themes.js';
+import { runPostprocess } from './postprocess.js';
+import { hasPandoc, hasPandocCrossref, hasLatex } from './dependencies.js';
+import { buildImageRegistry, writeImageRegistry } from './image-registry.js';
+import type { Author, JournalFormatting } from './types.js';
+import { getJournalProfile } from './journals.js';
+import { resolveCSL } from './csl.js';
+
+// =============================================================================
+// Constants
+// =============================================================================
+
+/** Supported output formats */
+const SUPPORTED_FORMATS = ['pdf', 'docx', 'tex', 'beamer', 'pptx'] as const;
+
+/** Maximum title length for output filename */
+const MAX_TITLE_FILENAME_LENGTH = 50;
+
+// =============================================================================
+// Interfaces
+// =============================================================================
+
+export interface CrossrefConfig {
+  figureTitle?: string;
+  tableTitle?: string;
+  figPrefix?: string | string[];
+  tblPrefix?: string | string[];
+  secPrefix?: string | string[];
+  linkReferences?: boolean;
+}
+
+export interface PdfConfig {
+  template?: string | null;
+  headerIncludes?: string | null;
+  documentclass?: string;
+  fontsize?: string;
+  geometry?: string;
+  linestretch?: number;
+  numbersections?: boolean;
+  toc?: boolean;
+  /**
+   * LaTeX engine: pdflatex (default), xelatex, lualatex, tectonic, etc.
+   * xelatex/lualatex are required for native UTF-8 rendering of Latin-Extended
+   * diacritics (Czech/Polish/Croatian/Spanish author names, species epithets).
+   */
+  engine?: string;
+  /** Roman/serif main font (xelatex/lualatex only — uses fontspec). */
+  mainfont?: string;
+  /** Sans-serif font (xelatex/lualatex only). */
+  sansfont?: string;
+  /** Monospace font (xelatex/lualatex only). */
+  monofont?: string;
+}
+
+export interface DocxConfig {
+  reference?: string | null;
+  keepComments?: boolean;
+  toc?: boolean;
+}
+
+export interface TexConfig {
+  standalone?: boolean;
+}
+
+export interface BeamerConfig {
+  theme?: string;
+  colortheme?: string | null;
+  fonttheme?: string | null;
+  aspectratio?: string | null;
+  navigation?: string | null;
+  section?: boolean;
+  notes?: string | false;
+  fit_images?: boolean;
+}
+
+export interface PptxConfig {
+  theme?: string;
+  reference?: string | null;
+  media?: string | null;
+  colors?: {
+    default?: string;
+    title?: string;
+  };
+  buildup?: {
+    grey?: string;
+    accent?: string;
+    enabled?: boolean;
+  };
+}
+
+export interface TablesConfig {
+  nowrap?: string[];
+}
+
+export interface PostprocessConfig {
+  pdf?: string | null;
+  docx?: string | null;
+  tex?: string | null;
+  pptx?: string | null;
+  beamer?: string | null;
+  all?: string | null;
+  [key: string]: string | null | undefined;
+}
+
+export interface BuildConfig {
+  title: string;
+  authors: (string | Author)[];
+  affiliations: Record<string, string>;
+  sections: string[];
+  bibliography: string | null;
+  csl: string | null;
+  crossref: CrossrefConfig;
+  pdf: PdfConfig;
+  docx: DocxConfig;
+  tex: TexConfig;
+  beamer: BeamerConfig;
+  pptx: PptxConfig;
+  tables: TablesConfig;
+  postprocess: PostprocessConfig;
+  /**
+   * Directory (relative to the project) where final outputs land. Created on
+   * demand. Set to null/empty to keep outputs alongside paper.md (legacy
+   * behavior).
+   */
+  outputDir?: string | null;
+  _configPath?: string | null;
+}
+
+export interface BuildResult {
+  format: string;
+  success: boolean;
+  outputPath?: string;
+  error?: string;
+}
+
+interface BuildOptions {
+  verbose?: boolean;
+  config?: BuildConfig;
+  outputPath?: string;
+  crossref?: boolean;
+  _refsAutoInjected?: boolean;
+  _forwardRefsResolved?: number;
+}
+
+interface CombineOptions extends BuildOptions {
+  _refsAutoInjected?: boolean;
+}
+
+interface VariablesContext {
+  sectionContents: string[];
+}
+
+interface PandocResult {
+  outputPath: string;
+  success: boolean;
+  error?: string;
+}
+
+interface FullBuildResult {
+  results: BuildResult[];
+  paperPath: string;
+  warnings: string[];
+  forwardRefsResolved: number;
+  refsAutoInjected?: boolean;
+}
+
+interface DynamicRef {
+  type: string;
+  label: string;
+  match: string;
+  position: number;
+}
+
+interface Registry {
+  figures: Map<string, unknown>;
+  tables: Map<string, unknown>;
+  equations: Map<string, unknown>;
+  byNumber: {
+    fig?: Map<number, string>;
+    figS?: Map<number, string>;
+    tbl?: Map<number, string>;
+    tblS?: Map<number, string>;
+    eq?: Map<number, string>;
+  };
+}
+
+/**
+ * Default rev.yaml configuration
+ */
+export const DEFAULT_CONFIG: BuildConfig = {
+  title: 'Untitled Document',
+  authors: [],
+  affiliations: {},
+  sections: [],
+  bibliography: null,
+  csl: null,
+  crossref: {
+    figureTitle: 'Figure',
+    tableTitle: 'Table',
+    figPrefix: ['Fig.', 'Figs.'],
+    tblPrefix: ['Table', 'Tables'],
+    secPrefix: ['Section', 'Sections'],
+    linkReferences: true,
+  },
+  pdf: {
+    template: null,
+    documentclass: 'article',
+    fontsize: '12pt',
+    geometry: 'margin=1in',
+    linestretch: 1.5,
+    numbersections: false,
+    toc: false,
+  },
+  docx: {
+    reference: null,
+    keepComments: true,
+    toc: false,
+  },
+  tex: {
+    standalone: true,
+  },
+  // Slide formats
+  beamer: {
+    theme: 'default',
+    colortheme: null,
+    fonttheme: null,
+    aspectratio: null, // '169' for 16:9, '43' for 4:3
+    navigation: null, // 'horizontal', 'vertical', 'frame', 'empty'
+    section: true, // section divider slides
+    notes: 'show', // 'show' (presenter view), 'only' (notes only), 'hide', or false
+    fit_images: true, // scale images to fit within slide bounds
+  },
+  pptx: {
+    theme: 'default', // Built-in theme: default, dark, academic, minimal, corporate
+    reference: null, // Custom reference-doc (overrides theme)
+    media: null, // directory with logo images (e.g., logo-left.png, logo-right.png)
+  },
+  // Table formatting
+  tables: {
+    nowrap: [], // Column headers to apply nowrap formatting (converts Normal() → $\mathcal{N}()$ etc.)
+  },
+  // Postprocess scripts
+  postprocess: {
+    pdf: null,
+    docx: null,
+    tex: null,
+    pptx: null,
+    beamer: null,
+    all: null, // Runs after any format
+  },
+  // Final outputs land here (created on demand). Set to null or '' to keep
+  // outputs in the project root.
+  outputDir: 'output',
+};
+
+// =============================================================================
+// Public API
+// =============================================================================
+
+/**
+ * Merge journal formatting defaults into a config.
+ * Priority: DEFAULT_CONFIG < journal formatting < rev.yaml explicit settings
+ */
+export function mergeJournalFormatting(config: BuildConfig, formatting: JournalFormatting, directory: string): BuildConfig {
+  const merged = { ...config };
+
+  // CSL: only apply if user hasn't set one
+  if (formatting.csl && !config.csl) {
+    const resolved = resolveCSL(formatting.csl, directory);
+    if (resolved) {
+      merged.csl = resolved;
+    }
+    // If not resolved locally, store the name — pandoc --citeproc
+    // can sometimes resolve it, and the user can fetch with rev profiles --fetch-csl
+    if (!resolved) {
+      merged.csl = formatting.csl;
+    }
+  }
+
+  // PDF settings: merge only unset fields
+  if (formatting.pdf) {
+    const userPdf = config.pdf || {};
+    const defaults = DEFAULT_CONFIG.pdf;
+    merged.pdf = { ...config.pdf };
+    for (const [key, value] of Object.entries(formatting.pdf)) {
+      const k = key as keyof PdfConfig;
+      // Apply journal value only if user config matches the default (i.e., wasn't explicitly set)
+      if (value !== undefined && JSON.stringify(userPdf[k]) === JSON.stringify(defaults[k])) {
+        (merged.pdf as Record<string, unknown>)[k] = value;
+      }
+    }
+  }
+
+  // DOCX settings: merge only unset fields
+  if (formatting.docx) {
+    const userDocx = config.docx || {};
+    const defaults = DEFAULT_CONFIG.docx;
+    merged.docx = { ...config.docx };
+    for (const [key, value] of Object.entries(formatting.docx)) {
+      const k = key as keyof DocxConfig;
+      if (value !== undefined && JSON.stringify(userDocx[k]) === JSON.stringify(defaults[k])) {
+        (merged.docx as Record<string, unknown>)[k] = value;
+      }
+    }
+  }
+
+  // Crossref settings: merge only unset fields
+  if (formatting.crossref) {
+    const userCrossref = config.crossref || {};
+    const defaults = DEFAULT_CONFIG.crossref;
+    merged.crossref = { ...config.crossref };
+    for (const [key, value] of Object.entries(formatting.crossref)) {
+      const k = key as keyof CrossrefConfig;
+      if (value !== undefined && JSON.stringify(userCrossref[k]) === JSON.stringify(defaults[k])) {
+        (merged.crossref as Record<string, unknown>)[k] = value;
+      }
+    }
+  }
+
+  return merged;
+}
+
+/**
+ * Load rev.yaml config from directory
+ * @param directory - Project directory path
+ * @returns Merged config with defaults
+ * @throws {TypeError} If directory is not a string
+ * @throws {Error} If rev.yaml exists but cannot be parsed
+ */
+export function loadConfig(directory: string): BuildConfig {
+  if (typeof directory !== 'string') {
+    throw new TypeError(`directory must be a string, got ${typeof directory}`);
+  }
+
+  const configPath = path.join(directory, 'rev.yaml');
+
+  if (!fs.existsSync(configPath)) {
+    return { ...DEFAULT_CONFIG, _configPath: null };
+  }
+
+  try {
+    const content = fs.readFileSync(configPath, 'utf-8');
+    const userConfig = YAML.parse(content) || {};
+
+    // Deep merge with defaults
+    let config: BuildConfig = {
+      ...DEFAULT_CONFIG,
+      ...userConfig,
+      crossref: { ...DEFAULT_CONFIG.crossref, ...userConfig.crossref },
+      pdf: { ...DEFAULT_CONFIG.pdf, ...userConfig.pdf },
+      docx: { ...DEFAULT_CONFIG.docx, ...userConfig.docx },
+      tex: { ...DEFAULT_CONFIG.tex, ...userConfig.tex },
+      beamer: { ...DEFAULT_CONFIG.beamer, ...userConfig.beamer },
+      pptx: { ...DEFAULT_CONFIG.pptx, ...userConfig.pptx },
+      tables: { ...DEFAULT_CONFIG.tables, ...userConfig.tables },
+      postprocess: { ...DEFAULT_CONFIG.postprocess, ...userConfig.postprocess },
+      _configPath: configPath,
+    };
+
+    // Apply journal formatting defaults (between DEFAULT_CONFIG and user settings)
+    if (userConfig.journal) {
+      const profile = getJournalProfile(userConfig.journal);
+      if (profile?.formatting) {
+        config = mergeJournalFormatting(config, profile.formatting, directory);
+      }
+    }
+
+    return config;
+  } catch (err) {
+    const error = err as Error;
+    throw new Error(`Failed to parse rev.yaml: ${error.message}`);
+  }
+}
+
+/**
+ * Find section files in directory
+ * @param directory - Project directory path
+ * @param configSections - Sections from rev.yaml (optional)
+ * @returns Ordered list of section file names
+ * @throws {TypeError} If directory is not a string
+ */
+export function findSections(directory: string, configSections: string[] = []): string[] {
+  if (typeof directory !== 'string') {
+    throw new TypeError(`directory must be a string, got ${typeof directory}`);
+  }
+
+  // If sections specified in config, use that order
+  if (configSections.length > 0) {
+    const sections: string[] = [];
+    for (const section of configSections) {
+      const filePath = path.join(directory, section);
+      if (fs.existsSync(filePath)) {
+        sections.push(section);
+      } else {
+        console.warn(`Warning: Section file not found: ${section}`);
+      }
+    }
+    return sections;
+  }
+
+  // Try sections.yaml
+  const sectionsYamlPath = path.join(directory, 'sections.yaml');
+  if (fs.existsSync(sectionsYamlPath)) {
+    try {
+      const sectionsConfig = YAML.parse(fs.readFileSync(sectionsYamlPath, 'utf-8'));
+      if (sectionsConfig.sections) {
+        return Object.entries(sectionsConfig.sections)
+          .sort((a: [string, any], b: [string, any]) => (a[1].order ?? 999) - (b[1].order ?? 999))
+          .map(([file]) => file)
+          .filter((f) => fs.existsSync(path.join(directory, f)));
+      }
+    } catch (e) {
+      if (process.env.DEBUG) {
+        const error = e as Error;
+        console.warn('build: YAML parse error in sections.yaml:', error.message);
+      }
+    }
+  }
+
+  // Default: find all .md files except special ones
+  const exclude = ['paper.md', 'readme.md', 'claude.md'];
+  const files = fs.readdirSync(directory).filter((f) => {
+    if (!f.endsWith('.md')) return false;
+    if (exclude.includes(f.toLowerCase())) return false;
+    return true;
+  });
+
+  // Sort alphabetically as fallback
+  return files.sort();
+}
+
+/**
+ * Combine section files into paper.md
+ */
+export function combineSections(directory: string, config: BuildConfig, options: CombineOptions = {}): string {
+  const sections = findSections(directory, config.sections);
+
+  if (sections.length === 0) {
+    throw new Error('No section files found. Create .md files or specify sections in rev.yaml');
+  }
+
+  const parts: string[] = [];
+
+  // Add YAML frontmatter
+  const frontmatter = buildFrontmatter(config);
+  parts.push('---');
+  parts.push(YAML.stringify(frontmatter).trim());
+  parts.push('---');
+  parts.push('');
+
+  // Read all section contents for variable processing
+  const sectionContents: string[] = [];
+
+  // Check if we need to auto-inject references before supplementary
+  // Pandoc places refs at the end by default, which breaks when supplementary follows
+  const hasRefsSection = sections.some(s =>
+    s.toLowerCase().includes('reference') || s.toLowerCase().includes('refs')
+  );
+  const suppIndex = sections.findIndex(s =>
+    s.toLowerCase().includes('supp') || s.toLowerCase().includes('appendix')
+  );
+  const hasBibliography = config.bibliography && fs.existsSync(path.join(directory, config.bibliography));
+
+  // Track if we find an explicit refs div in any section
+  let hasExplicitRefsDiv = false;
+
+  // Combine sections
+  for (let i = 0; i < sections.length; i++) {
+    const section = sections[i];
+    if (!section) continue;
+    const filePath = path.join(directory, section);
+    let content = fs.readFileSync(filePath, 'utf-8');
+
+    // Remove any existing frontmatter from section files
+    content = stripFrontmatter(content);
+    sectionContents.push(content);
+
+    // Check if this section has an explicit refs div
+    if (content.includes('::: {#refs}') || content.includes(':::  {#refs}')) {
+      hasExplicitRefsDiv = true;
+    }
+
+    // Auto-inject references before supplementary if needed
+    if (i === suppIndex && hasBibliography && !hasRefsSection && !hasExplicitRefsDiv) {
+      parts.push('# References\n');
+      parts.push('::: {#refs}');
+      parts.push(':::');
+      parts.push('');
+      parts.push('');
+      options._refsAutoInjected = true;
+    }
+
+    parts.push(content.trim());
+    parts.push('');
+    parts.push(''); // Double newline between sections
+  }
+
+  let paperContent = parts.join('\n');
+
+  // Process template variables if any exist
+  if (hasVariables(paperContent)) {
+    paperContent = processVariables(paperContent, config as any, { sectionContents });
+  }
+
+  // Resolve forward references (refs that appear before their anchor definition)
+  // This fixes pandoc-crossref limitation with multi-file documents
+  if (hasPandocCrossref()) {
+    const registry = buildRegistry(directory, sections);
+    const { text, resolved } = resolveForwardRefs(paperContent, registry);
+    if (resolved.length > 0) {
+      paperContent = text;
+      // Store resolved count for optional reporting
+      options._forwardRefsResolved = resolved.length;
+    }
+
+    // Resolve supplementary references and strip their anchors.
+    // pandoc-crossref cannot produce "Figure S1" numbering — it numbers all
+    // figures sequentially. We resolve supplementary refs to plain text and
+    // remove the {#fig:...} attributes so crossref ignores them.
+    const supp = resolveSupplementaryRefs(paperContent, registry);
+    if (supp.resolved.length > 0) {
+      paperContent = supp.text;
+    }
+  }
+
+  const paperPath = path.join(directory, 'paper.md');
+
+  fs.writeFileSync(paperPath, paperContent, 'utf-8');
+
+  return paperPath;
+}
+
+/**
+ * Build YAML frontmatter from config
+ */
+function buildFrontmatter(config: BuildConfig): Record<string, unknown> {
+  const fm: Record<string, unknown> = {};
+
+  if (config.title) fm.title = config.title;
+
+  // Skip author in frontmatter when using numbered affiliations —
+  // the author block is injected separately per format
+  if (config.authors && config.authors.length > 0 && !hasNumberedAffiliations(config)) {
+    fm.author = config.authors;
+  }
+
+  if (config.bibliography) {
+    fm.bibliography = config.bibliography;
+  }
+
+  if (config.csl) {
+    fm.csl = config.csl;
+  }
+
+  return fm;
+}
+
+/**
+ * Strip YAML frontmatter from content
+ */
+function stripFrontmatter(content: string): string {
+  const match = content.match(/^---\r?\n([\s\S]*?)\r?\n---\r?\n/);
+  if (match) {
+    return content.slice(match[0].length);
+  }
+  return content;
+}
+
+/**
+ * Check if config uses numbered affiliation mode
+ * (authors have `affiliations` arrays and an affiliations map is defined)
+ */
+function hasNumberedAffiliations(config: BuildConfig): boolean {
+  if (!config.affiliations || Object.keys(config.affiliations).length === 0) return false;
+  return config.authors.some(a => typeof a !== 'string' && a.affiliations && a.affiliations.length > 0);
+}
+
+/**
+ * Generate LaTeX author block using authblk package for numbered superscript affiliations.
+ * Returns LaTeX code to be injected via header-includes.
+ */
+function generateLatexAuthorBlock(config: BuildConfig): string {
+  const lines: string[] = [];
+  lines.push('\\usepackage{authblk}');
+  lines.push('\\renewcommand\\Authfont{\\normalsize}');
+  lines.push('\\renewcommand\\Affilfont{\\small}');
+  lines.push('');
+
+  // Map affiliation keys to numbers
+  const affiliationKeys = Object.keys(config.affiliations);
+  const keyToNum = new Map<string, number>();
+  affiliationKeys.forEach((key, i) => keyToNum.set(key, i + 1));
+
+  // Authors
+  for (const author of config.authors) {
+    if (typeof author === 'string') {
+      lines.push(`\\author{${author}}`);
+      continue;
+    }
+    const marks = (author.affiliations || [])
+      .map(k => keyToNum.get(k))
+      .filter((n): n is number => n !== undefined);
+
+    const markStr = marks.length > 0 ? `[${marks.join(',')}]` : '';
+    let nameStr = author.name;
+    if (author.corresponding && author.email) {
+      nameStr += `\\thanks{Corresponding author: ${author.email}}`;
+    } else if (author.corresponding) {
+      nameStr += '\\thanks{Corresponding author}';
+    }
+    lines.push(`\\author${markStr}{${nameStr}}`);
+  }
+
+  // Affiliations
+  for (const [key, text] of Object.entries(config.affiliations)) {
+    const num = keyToNum.get(key);
+    if (num !== undefined) {
+      lines.push(`\\affil[${num}]{${text}}`);
+    }
+  }
+
+  return lines.join('\n');
+}
+
+/**
+ * Generate markdown author block for DOCX output with superscript affiliations.
+ * Returns markdown text to insert after the YAML frontmatter.
+ */
+function generateMarkdownAuthorBlock(config: BuildConfig): string {
+  const lines: string[] = [];
+
+  // Map affiliation keys to numbers
+  const affiliationKeys = Object.keys(config.affiliations);
+  const keyToNum = new Map<string, number>();
+  affiliationKeys.forEach((key, i) => keyToNum.set(key, i + 1));
+
+  // Author line: Name^1,2^, Name^3^, ...
+  const authorParts: string[] = [];
+  for (const author of config.authors) {
+    if (typeof author === 'string') {
+      authorParts.push(author);
+      continue;
+    }
+    const marks = (author.affiliations || [])
+      .map(k => keyToNum.get(k))
+      .filter((n): n is number => n !== undefined);
+    let entry = author.name;
+    const superParts = marks.map(String);
+    if (author.corresponding) superParts.push('\\*');
+    if (superParts.length > 0) {
+      entry += `^${superParts.join(',')}^`;
+    }
+    authorParts.push(entry);
+  }
+  lines.push(authorParts.join(', '));
+  lines.push('');
+
+  // Affiliation lines: ^1^ Department of ...
+  for (const [key, text] of Object.entries(config.affiliations)) {
+    const num = keyToNum.get(key);
+    if (num !== undefined) {
+      lines.push(`^${num}^ ${text}`);
+    }
+  }
+
+  // Corresponding author footnote
+  const corresponding = config.authors.find(a => typeof a !== 'string' && a.corresponding) as Author | undefined;
+  if (corresponding?.email) {
+    lines.push('');
+    lines.push(`^\\*^ Corresponding author: ${corresponding.email}`);
+  }
+
+  lines.push('');
+  return lines.join('\n');
+}
+
+/**
+ * Process markdown tables to apply nowrap formatting to specified columns.
+ * Converts distribution notation (Normal, Student-t, Gamma) to LaTeX math.
+ * @param content - Markdown content
+ * @param tablesConfig - tables config from rev.yaml
+ * @param format - output format (pdf, docx, etc.)
+ * @returns processed content
+ */
+export function processTablesForFormat(content: string, tablesConfig: TablesConfig, format: string): string {
+  // Only process for PDF/TeX output
+  if (format !== 'pdf' && format !== 'tex') {
+    return content;
+  }
+
+  // Check if we have nowrap columns configured
+  if (!tablesConfig?.nowrap?.length) {
+    return content;
+  }
+
+  const nowrapPatterns = tablesConfig.nowrap.map((p) => p.toLowerCase());
+
+  // Match pipe tables: header row, separator row, body rows
+  // Header: | Col1 | Col2 | Col3 |
+  // Separator: |:-----|:-----|:-----|
+  // Body: | val1 | val2 | val3 |
+  const tableRegex = /^(\|[^\n]+\|\r?\n\|[-:| ]+\|\r?\n)((?:\|[^\n]+\|\r?\n?)+)/gm;
+
+  return content.replace(tableRegex, (match, headerAndSep, body) => {
+    // Split header from separator
+    const lines = headerAndSep.split(/\r?\n/);
+    const headerLine = lines[0] ?? '';
+
+    // Parse header cells to find nowrap column indices
+    const headerCells = headerLine
+      .split('|')
+      .slice(1, -1)
+      .map((c: string) => c.trim().toLowerCase());
+
+    const nowrapCols: number[] = [];
+    headerCells.forEach((cell: string, i: number) => {
+      if (nowrapPatterns.some((p) => cell.includes(p))) {
+        nowrapCols.push(i);
+      }
+    });
+
+    // If no nowrap columns found in this table, return unchanged
+    if (nowrapCols.length === 0) {
+      return match;
+    }
+
+    // Process body rows
+    const bodyLines = body.split(/\r?\n/).filter((l: string) => l.trim());
+    const processedBody = bodyLines
+      .map((row: string) => {
+        // Split row into cells, keeping the pipe structure
+        const cells = row.split('|');
+        // cells[0] is empty (before first |), cells[last] is empty (after last |)
+
+        nowrapCols.forEach((colIdx) => {
+          const cellIdx = colIdx + 1; // Account for empty first element
+          if (cells[cellIdx] !== undefined) {
+            const cellContent = cells[cellIdx].trim();
+
+            // Skip if empty, already math, or already has LaTeX commands
+            if (!cellContent || cellContent.startsWith('$') || cellContent.startsWith('\\')) {
+              return;
+            }
+
+            // Convert distribution notation to LaTeX math
+            // Order matters: compound names (Half-Normal) must come before simple names (Normal)
+            let processed = cellContent;
+
+            // Half-Normal(x) → $\text{Half-Normal}(x)$ (must come before Normal)
+            processed = processed.replace(/Half-Normal\(([^)]+)\)/g, '$\\text{Half-Normal}($1)$');
+
+            // Normal(x, y) → $\mathcal{N}(x, y)$
+            processed = processed.replace(/Normal\(([^)]+)\)/g, '$\\mathcal{N}($1)$');
+
+            // Student-t(df, loc, scale) → $t_{df}(loc, scale)$
+            processed = processed.replace(/Student-t\((\d+),\s*([^)]+)\)/g, '$t_{$1}($2)$');
+
+            // Gamma(a, b) → $\text{Gamma}(a, b)$
+            processed = processed.replace(/Gamma\(([^)]+)\)/g, '$\\text{Gamma}($1)$');
+
+            // Exponential(x) → $\text{Exp}(x)$
+            processed = processed.replace(/Exponential\(([^)]+)\)/g, '$\\text{Exp}($1)$');
+
+            // Update cell with padding
+            cells[cellIdx] = ` ${processed} `;
+          }
+        });
+
+        return cells.join('|');
+      })
+      .join('\n');
+
+    return headerAndSep + processedBody + '\n';
+  });
+}
+
+/**
+ * Apply format-specific transforms (table normalization, author blocks,
+ * crossref display conversion, slide syntax). Caller is responsible for
+ * stripping annotations beforehand — the dual-output paths keep comments
+ * in the markdown stream and need to apply these transforms separately
+ * from annotation handling.
+ *
+ * @param content - Markdown content (annotations already stripped as needed)
+ * @param format - Output format
+ * @param config - Build config
+ * @param registry - Crossref registry for the project
+ * @returns Transformed markdown
+ */
+export function applyFormatTransforms(
+  content: string,
+  format: string,
+  config: BuildConfig,
+  registry: Registry
+): string {
+  if (format === 'pdf' || format === 'tex') {
+    content = processTablesForFormat(content, config.tables, format);
+
+    if (hasNumberedAffiliations(config)) {
+      const latexBlock = generateLatexAuthorBlock(config);
+      content = content.replace(/^(---\r?\n[\s\S]*?)(---\r?\n)/, (_match, yamlContent, closing) => {
+        return `${yamlContent}header-includes: |\n${latexBlock.split('\n').map(l => '  ' + l).join('\n')}\n${closing}`;
+      });
+    }
+  } else if (format === 'docx') {
+    content = convertDynamicRefsToDisplay(content, registry);
+
+    if (hasNumberedAffiliations(config)) {
+      const mdBlock = generateMarkdownAuthorBlock(config);
+      content = content.replace(/^(---\r?\n[\s\S]*?---\r?\n)/, `$1\n${mdBlock}\n`);
+    }
+  } else if (format === 'beamer' || format === 'pptx') {
+    if (hasSlideSyntax(content)) {
+      content = processSlideMarkdown(content, format);
+    }
+  }
+
+  return content;
+}
+
+/**
+ * Prepare paper.md for specific output format
+ */
+export function prepareForFormat(
+  paperPath: string,
+  format: string,
+  config: BuildConfig,
+  _options: BuildOptions = {}
+): string {
+  const directory = path.dirname(paperPath);
+  let content = fs.readFileSync(paperPath, 'utf-8');
+
+  // Build crossref registry for reference conversion
+  // Pass sections from config to ensure correct file ordering
+  const registry = buildRegistry(directory, config.sections);
+
+  // Strip annotations per format
+  if (format === 'docx') {
+    content = stripAnnotations(content, { keepComments: config.docx.keepComments });
+  } else {
+    content = stripAnnotations(content);
+  }
+
+  // Apply shared format transforms
+  content = applyFormatTransforms(content, format, config, registry);
+
+  // Write to temporary file
+  const preparedPath = path.join(directory, `.paper-${format}.md`);
+  fs.writeFileSync(preparedPath, content, 'utf-8');
+
+  return preparedPath;
+}
+
+/**
+ * Convert @fig:label references to display format (Figure 1)
+ */
+function convertDynamicRefsToDisplay(text: string, registry: Registry): string {
+  const refs = detectDynamicRefs(text);
+
+  // Process in reverse order to preserve positions
+  let result = text;
+  for (let i = refs.length - 1; i >= 0; i--) {
+    const ref = refs[i];
+    if (!ref) continue;
+    const display = labelToDisplay(ref.type, ref.label, registry as any);
+
+    if (display) {
+      result = result.slice(0, ref.position) + display + result.slice(ref.position + ref.match.length);
+    }
+  }
+
+  return result;
+}
+
+/**
+ * Build pandoc arguments for format
+ */
+export function buildPandocArgs(format: string, config: BuildConfig, outputPath: string): string[] {
+  const args: string[] = [];
+
+  // Output format
+  if (format === 'tex') {
+    args.push('-t', 'latex');
+    if (config.tex.standalone) {
+      args.push('-s');
+    }
+  } else if (format === 'pdf') {
+    args.push('-t', 'pdf');
+  } else if (format === 'docx') {
+    args.push('-t', 'docx');
+  } else if (format === 'beamer') {
+    args.push('-t', 'beamer');
+  } else if (format === 'pptx') {
+    args.push('-t', 'pptx');
+  }
+
+  // Output file. runPandoc sets cwd to the project directory and passes a
+  // path relative to that cwd; passing it through here unchanged lets pandoc
+  // write to subdirectories like output/<title-slug>.<ext>.
+  args.push('-o', outputPath);
+
+  // Crossref filter (if available) - skip for slides
+  if (hasPandocCrossref() && format !== 'beamer' && format !== 'pptx') {
+    args.push('--filter', 'pandoc-crossref');
+  }
+
+  // Bibliography
+  if (config.bibliography) {
+    args.push('--citeproc');
+  }
+
+  // Format-specific options
+  if (format === 'pdf') {
+    if (config.pdf.template) {
+      args.push('--template', config.pdf.template);
+    }
+    if (config.pdf.engine) {
+      args.push(`--pdf-engine=${config.pdf.engine}`);
+    }
+    if (config.pdf.mainfont) {
+      args.push('-V', `mainfont=${config.pdf.mainfont}`);
+    }
+    if (config.pdf.sansfont) {
+      args.push('-V', `sansfont=${config.pdf.sansfont}`);
+    }
+    if (config.pdf.monofont) {
+      args.push('-V', `monofont=${config.pdf.monofont}`);
+    }
+    args.push('-V', `documentclass=${config.pdf.documentclass}`);
+    args.push('-V', `fontsize=${config.pdf.fontsize}`);
+    args.push('-V', `geometry:${config.pdf.geometry}`);
+    if (config.pdf.headerIncludes) {
+      args.push('-H', config.pdf.headerIncludes);
+    }
+    if (config.pdf.linestretch !== 1) {
+      args.push('-V', `linestretch=${config.pdf.linestretch}`);
+    }
+    if (config.pdf.numbersections) {
+      args.push('--number-sections');
+    }
+    if (config.pdf.toc) {
+      args.push('--toc');
+    }
+  } else if (format === 'docx') {
+    if (config.docx.reference) {
+      args.push('--reference-doc', config.docx.reference);
+    }
+    if (config.docx.toc) {
+      args.push('--toc');
+    }
+  } else if (format === 'beamer') {
+    // Beamer slide options
+    const beamer = config.beamer || {};
+    if (beamer.theme) {
+      args.push('-V', `theme=${beamer.theme}`);
+    }
+    if (beamer.colortheme) {
+      args.push('-V', `colortheme=${beamer.colortheme}`);
+    }
+    if (beamer.fonttheme) {
+      args.push('-V', `fonttheme=${beamer.fonttheme}`);
+    }
+    if (beamer.aspectratio) {
+      args.push('-V', `aspectratio=${beamer.aspectratio}`);
+    }
+    if (beamer.navigation) {
+      args.push('-V', `navigation=${beamer.navigation}`);
+    }
+    // Speaker notes - default to 'show' which creates presenter view PDF
+    // Options: 'show' (dual screen), 'only' (notes only), 'hide' (no notes), false (disabled)
+    const notesMode = beamer.notes !== undefined ? beamer.notes : 'show';
+    if (notesMode && notesMode !== 'hide') {
+      args.push('-V', `classoption=notes=${notesMode}`);
+    }
+    // Fit images within slide bounds (default: true)
+    if (beamer.fit_images !== false) {
+      const fitImagesHeader = `\\makeatletter
+\\def\\maxwidth{\\ifdim\\Gin@nat@width>\\linewidth\\linewidth\\else\\Gin@nat@width\\fi}
+\\def\\maxheight{\\ifdim\\Gin@nat@height>0.75\\textheight 0.75\\textheight\\else\\Gin@nat@height\\fi}
+\\makeatother
+\\setkeys{Gin}{width=\\maxwidth,height=\\maxheight,keepaspectratio}`;
+      args.push('-V', `header-includes=${fitImagesHeader}`);
+    }
+    // Slides need standalone
+    args.push('-s');
+  } else if (format === 'pptx') {
+    // PowerPoint options - handled separately in preparePptxTemplate
+    // Reference doc is set by caller after template generation
+  }
+
+  return args;
+}
+
+/**
+ * Write crossref.yaml if needed
+ */
+function ensureCrossrefConfig(directory: string, config: BuildConfig): void {
+  const crossrefPath = path.join(directory, 'crossref.yaml');
+
+  if (!fs.existsSync(crossrefPath) && hasPandocCrossref()) {
+    fs.writeFileSync(crossrefPath, YAML.stringify(config.crossref), 'utf-8');
+  }
+}
+
+/**
+ * Get install instructions for missing dependency
+ */
+function getInstallInstructions(tool: string): string {
+  const instructions: Record<string, string> = {
+    pandoc: 'https://pandoc.org/installing.html',
+    latex: 'https://www.latex-project.org/get/',
+  };
+  return instructions[tool] || 'Check documentation';
+}
+
+/**
+ * Resolve the absolute directory where final outputs should land.
+ * Honors config.outputDir; falls back to the project directory when null/empty.
+ */
+export function resolveOutputDir(directory: string, config: BuildConfig): string {
+  const out = config.outputDir;
+  if (!out) return directory;
+  return path.isAbsolute(out) ? out : path.join(directory, out);
+}
+
+/**
+ * Run pandoc build
+ */
+export async function runPandoc(
+  inputPath: string,
+  format: string,
+  config: BuildConfig,
+  options: BuildOptions = {}
+): Promise<PandocResult> {
+  const directory = path.dirname(inputPath);
+  const baseName = config.title
+    ? config.title.toLowerCase().replace(/[^a-z0-9]+/g, '-').slice(0, 50)
+    : 'paper';
+
+  // Map format to file extension
+  const extMap: Record<string, string> = {
+    tex: '.tex',
+    pdf: '.pdf',
+    docx: '.docx',
+    beamer: '.pdf', // beamer outputs PDF
+    pptx: '.pptx',
+  };
+  const ext = extMap[format] || '.pdf';
+
+  // For beamer, use -slides suffix to distinguish from regular PDF
+  const suffix = format === 'beamer' ? '-slides' : '';
+  // Allow custom output path via options. Auto-named outputs go through the
+  // configured outputDir (default 'output/'); explicit paths are honored as-is
+  // so callers can route temp/intermediate artefacts where they want.
+  const outputPath = options.outputPath
+    ? options.outputPath
+    : path.join(resolveOutputDir(directory, config), `${baseName}${suffix}${ext}`);
+
+  if (!options.outputPath) {
+    const outDir = path.dirname(outputPath);
+    if (!fs.existsSync(outDir)) {
+      fs.mkdirSync(outDir, { recursive: true });
+    }
+  }
+
+  // Ensure crossref.yaml exists
+  ensureCrossrefConfig(directory, config);
+
+  // Pandoc runs with cwd = directory, so pass the output path relative to it.
+  const args = buildPandocArgs(format, config, path.relative(directory, outputPath) || path.basename(outputPath));
+
+  // Handle PPTX reference template and themes
+  let pptxMediaDir: string | null = null;
+  if (format === 'pptx') {
+    const pptx = config.pptx || {};
+
+    // Determine media directory (default: pptx/media or slides/media)
+    let mediaDir = pptx.media;
+    if (!mediaDir) {
+      if (fs.existsSync(path.join(directory, 'pptx', 'media'))) {
+        mediaDir = path.join(directory, 'pptx', 'media');
+      } else if (fs.existsSync(path.join(directory, 'slides', 'media'))) {
+        mediaDir = path.join(directory, 'slides', 'media');
+      }
+    } else if (!path.isAbsolute(mediaDir)) {
+      mediaDir = path.join(directory, mediaDir);
+    }
+    pptxMediaDir = mediaDir || null;
+
+    // Determine reference doc: custom reference overrides theme
+    let referenceDoc: string | null = null;
+    if (pptx.reference && fs.existsSync(path.join(directory, pptx.reference))) {
+      // Custom reference doc takes precedence
+      referenceDoc = path.join(directory, pptx.reference);
+    } else {
+      // Use built-in theme (default: 'default')
+      const themeName = pptx.theme || 'default';
+      const themePath = getThemePath(themeName);
+      if (themePath && fs.existsSync(themePath)) {
+        referenceDoc = themePath;
+      }
+    }
+
+    if (referenceDoc) {
+      args.push('--reference-doc', referenceDoc);
+    }
+
+    // Add color filter for PPTX (handles [text]{color=#RRGGBB} syntax)
+    const colorFilterPath = path.join(path.dirname(new URL(import.meta.url).pathname.replace(/^\/([A-Z]:)/, '$1')), 'pptx-color-filter.lua');
+    if (fs.existsSync(colorFilterPath)) {
+      args.push('--lua-filter', colorFilterPath);
+    }
+  }
+
+  // Add crossref metadata file if exists (skip for slides - they don't use crossref)
+  if (format !== 'beamer' && format !== 'pptx') {
+    const crossrefPath = path.join(directory, 'crossref.yaml');
+    if (fs.existsSync(crossrefPath) && hasPandocCrossref()) {
+      // Use basename since we set cwd to directory
+      args.push('--metadata-file', 'crossref.yaml');
+    }
+  }
+
+  // Input file (use basename since we set cwd to directory)
+  args.push(path.basename(inputPath));
+
+  return new Promise((resolve) => {
+    const pandoc: ChildProcess = spawn('pandoc', args, {
+      cwd: directory,
+      stdio: ['ignore', 'pipe', 'pipe'],
+    });
+
+    let stderr = '';
+    pandoc.stderr?.on('data', (data) => {
+      stderr += data.toString();
+    });
+
+    pandoc.on('close', async (code) => {
+      if (code === 0) {
+        // For PPTX, post-process to add slide numbers, buildup colors, and logos
+        if (format === 'pptx') {
+          try {
+            // Inject slide numbers into content slides only
+            await injectSlideNumbers(outputPath);
+          } catch (e) {
+            // Slide number injection failed but output was created
+          }
+          try {
+            // Apply colors (default text color, title color, buildup greying)
+            const pptxConfig = config.pptx || {};
+            const colorsConfig = pptxConfig.colors || {};
+            const buildupConfig = pptxConfig.buildup || {};
+            // Merge colors and buildup config for applyBuildupColors
+            const colorConfig = {
+              default: colorsConfig.default,
+              title: colorsConfig.title,
+              grey: buildupConfig.grey,
+              accent: buildupConfig.accent,
+              enabled: buildupConfig.enabled
+            };
+            await applyBuildupColors(outputPath, colorConfig);
+          } catch (e) {
+            // Color application failed but output was created
+          }
+          // Inject logos into cover slide (if media dir configured)
+          if (pptxMediaDir) {
+            try {
+              await injectMediaIntoPptx(outputPath, pptxMediaDir);
+            } catch (e) {
+              // Logo injection failed but output was created
+            }
+          }
+        }
+
+        // Run user postprocess scripts
+        const postResult = await runPostprocess(outputPath, format, config as unknown as Parameters<typeof runPostprocess>[2], options);
+        if (!postResult.success && options.verbose) {
+          console.error(`Postprocess warning: ${postResult.error}`);
+        }
+
+        resolve({ outputPath, success: true });
+      } else {
+        resolve({ outputPath, success: false, error: stderr || `Exit code ${code}` });
+      }
+    });
+
+    pandoc.on('error', (err) => {
+      resolve({ outputPath, success: false, error: err.message });
+    });
+  });
+}
+
+/**
+ * Full build pipeline
+ */
+export async function build(
+  directory: string,
+  formats: string[] = ['pdf', 'docx'],
+  options: BuildOptions = {}
+): Promise<FullBuildResult> {
+  const warnings: string[] = [];
+  let forwardRefsResolved = 0;
+
+  // Check pandoc
+  if (!hasPandoc()) {
+    const instruction = getInstallInstructions('pandoc');
+    throw new Error(`Pandoc not found. Install with: ${instruction}\nOr run: rev doctor`);
+  }
+
+  // Check LaTeX if PDF is requested
+  if ((formats.includes('pdf') || formats.includes('all')) && !hasLatex()) {
+    warnings.push(`LaTeX not found - PDF generation may fail. Install with: ${getInstallInstructions('latex')}`);
+  }
+
+  // Check pandoc-crossref
+  if (!hasPandocCrossref()) {
+    warnings.push('pandoc-crossref not found - figure/table numbering will not work');
+  }
+
+  // Load config (use passed config if provided, otherwise load from file)
+  const config = options.config || loadConfig(directory);
+
+  // Combine sections → paper.md
+  const buildOptions: CombineOptions = { ...options };
+  const paperPath = combineSections(directory, config, buildOptions);
+  forwardRefsResolved = buildOptions._forwardRefsResolved || 0;
+  const refsAutoInjected = buildOptions._refsAutoInjected || false;
+
+  // Expand 'all' to all formats
+  if (formats.includes('all')) {
+    formats = ['pdf', 'docx', 'tex'];
+  }
+
+  // Build and save image registry when DOCX is being built
+  // This allows import to restore proper image syntax from Word documents
+  if (formats.includes('docx')) {
+    const paperContent = fs.readFileSync(paperPath, 'utf-8');
+    const crossrefReg = buildRegistry(directory, config.sections);
+    const imageReg = buildImageRegistry(paperContent, crossrefReg as any);
+    if ((imageReg as any).figures?.length > 0) {
+      writeImageRegistry(directory, imageReg);
+    }
+  }
+
+  const results: BuildResult[] = [];
+
+  for (const format of formats) {
+    // Prepare format-specific version
+    const preparedPath = prepareForFormat(paperPath, format, config, options);
+
+    // Run pandoc
+    const result = await runPandoc(preparedPath, format, config, options);
+    results.push({ format, ...result });
+
+    // Clean up temp file
+    try {
+      fs.unlinkSync(preparedPath);
+    } catch {
+      // Ignore cleanup errors
+    }
+  }
+
+  return { results, paperPath, warnings, forwardRefsResolved, refsAutoInjected };
+}
+
+/**
+ * Get build status summary
+ */
+export function formatBuildResults(results: BuildResult[]): string {
+  const lines: string[] = [];
+
+  for (const r of results) {
+    if (r.success) {
+      lines.push(`  ${r.format.toUpperCase()}: ${path.basename(r.outputPath!)}`);
+    } else {
+      lines.push(`  ${r.format.toUpperCase()}: FAILED - ${r.error}`);
+    }
+  }
+
+  return lines.join('\n');
+}