chem-pdf2ppt 2.0.0

package/SKILL_EN.md

@@ -0,0 +1,473 @@
+---
+name: pdf-to-ppt
+description: Convert chemistry (experimental or theoretical/computational) academic paper PDFs into professional academic PowerPoint presentations. Automatically identifies paper type (experimental/theoretical/hybrid), extracts key information, and generates structured PPTX files ready for group meetings, defenses, and academic reports. Trigger when the user mentions "paper to PPT", "PDF to PPT", "group meeting slides", "journal club", "academic presentation", "chemistry paper PPT", or provides a chemistry-related PDF/arXiv paper asking for presentation generation.
+---
+
+# PDF → Chemistry Academic PPT Generator
+
+Convert chemistry academic paper PDFs into professional, ready-to-use academic PowerPoint presentations.
+
+## Core Principles
+
+**Let the paper's scientific argument drive the PPT structure, not a fixed template.**
+
+Chemistry papers fall into two primary paradigms with fundamentally different narrative logic:
+
+- **Experimental Chemistry**: Problem → Design Strategy → Synthesis/Preparation → Characterization → Performance → Mechanism → Conclusions
+- **Theoretical/Computational Chemistry**: Problem → Methods → Model Validation → Electronic Structure/Energy Analysis → Mechanism → Experimental Comparison → Conclusions
+- **Experimental + Theoretical Hybrid**: Problem → Experimental Part → Computational Part → Cross-Validation → Unified Mechanism → Conclusions
+
+The generated PPT should guide the audience through:
+1. Why is this chemistry problem important?
+2. What has been achieved, what is the bottleneck?
+3. What is the design strategy / computational approach?
+4. What is the key experimental evidence / computational result?
+5. Is the mechanistic explanation self-consistent?
+6. What new understanding has been gained? Is it generalizable?
+7. What are the limitations and open questions?
+
+## Input Formats
+
+Any of the following are accepted:
+- Full paper PDF
+- arXiv / chemRxiv preprint PDF
+- Abstract + figures + results text
+- Structured reading notes
+- User-pasted paper content
+
+Default output language is Chinese; preserve key chemical terminology, abbreviations, compound names, and method names in English.
+
+## Workflow
+
+### Step 0: Install Dependencies (first use)
+
+```bash
+pip install -r <SKILL_ROOT>/requirements.txt
+```
+
+### Step 1: Read and Understand the Paper
+
+Use PyMuPDF (fitz) to extract full text:
+
+```python
+import fitz
+doc = fitz.open("paper.pdf")
+full_text = ""
+for page in doc:
+    full_text += page.get_text()
+```
+
+Identify the following key information from the full text:
+- **Title, authors, journal, year, DOI**
+- **Paper chemistry type**: Experimental / Computational / Hybrid
+- **Core chemistry problem**: catalysis, synthesis, materials, energy, environment, pharmaceuticals, mechanism, etc.
+- **Research gap**: limitations of prior work
+- **Core claim/hypothesis**
+- **Key methods**:
+  - Experimental: synthesis routes, catalyst preparation, characterization techniques (XRD, TEM, SEM, XPS, BET, NMR, IR, Raman, EPR, XAS, etc.), performance testing conditions
+  - Computational: level of theory (DFT functional, basis set), software (VASP, Gaussian, CP2K, QE, etc.), model systems, free energy methods
+- **Key results**: performance data, characterization conclusions, energy data, selectivity/conversion/yield, TOF, stability
+- **Key figures**: Figure numbers and content summaries
+- **Mechanism / structure-property relationships**
+- **Innovations and limitations**
+
+**Important**: Do not fabricate data, mechanisms, or figure details not present in the paper. Mark uncertain information as "[TBD]".
+
+### Step 2: Classify Paper Type and Select Narrative Arc
+
+#### Type A: Experimental Chemistry
+**Signals**: synthesis steps, wet chemistry methods, materials preparation, catalyst testing, characterization data (XRD/TEM/SEM/XPS), performance metrics (conversion/selectivity/yield/stability)
+
+**Narrative arc** (question-to-evidence):
+1. Chemical problem and significance
+2. Prior strategy limitations → This work's design rationale
+3. Synthesis/preparation route
+4. Structure and composition characterization
+5. Catalytic/performance evaluation
+6. Mechanistic investigation (in-situ characterization, control experiments, kinetics)
+7. Structure-activity relationships
+8. Conclusions and outlook
+
+**PPT structure** (12–16 slides):
+```
+Slide 1:  Title
+Slide 2:  Background & chemical problem
+Slide 3:  Prior work & bottlenecks
+Slide 4:  Design strategy
+Slide 5:  Synthesis/preparation route
+Slide 6:  Structure & composition (XRD/TEM/XPS...)
+Slide 7:  Morphology & microstructure (SEM/TEM/HRTEM...)
+Slide 8:  Performance evaluation
+Slide 9:  Key control experiments or comparison table
+Slide 10: Mechanism (in-situ / kinetics / poisoning...)
+Slide 11: Structure-activity / active site discussion
+Slide 12: Benchmark comparison
+Slide 13: Summary & innovations
+Slide 14: Limitations & outlook
+```
+
+#### Type B: Theoretical/Computational Chemistry
+**Signals**: DFT calculations, molecular dynamics, Monte Carlo, electronic structure, reaction path search, free energy calculations, software names (VASP/Gaussian/CP2K/QE/GROMACS), k-point/cutoff energy
+
+**Narrative arc** (method-to-mechanism):
+1. Chemical problem and need for computation
+2. Limitations of prior computational studies
+3. Computational methods and model systems
+4. Method validation / benchmarking
+5. Electronic structure / adsorption / reaction intermediates
+6. Energy profiles and reaction pathways
+7. Selectivity origins / rate-determining step analysis
+8. Comparison with experiment (if available)
+9. Conclusions and outlook
+
+**PPT structure** (12–16 slides):
+```
+Slide 1:  Title
+Slide 2:  Background & chemical problem
+Slide 3:  Prior computational studies
+Slide 4:  Computational methods & models
+Slide 5:  Method validation / benchmarking
+Slide 6:  Key intermediate / transition state structures
+Slide 7:  Energy profiles / free energy diagrams
+Slide 8:  Electronic structure analysis (PDOS/Bader/COHP...)
+Slide 9:  Selectivity analysis
+Slide 10: Microkinetics / volcano plot (if applicable)
+Slide 11: Comparison with experiment (if applicable)
+Slide 12: Summary & innovations
+Slide 13: Limitations & outlook
+```
+
+#### Type C: Experimental + Theoretical Hybrid
+**Narrative arc** (experiment-theory-unified):
+1. Chemical problem and joint strategy
+2. Experimental: synthesis + characterization + performance
+3. Computational: model + method + results
+4. Experiment-theory cross-validation
+5. Unified mechanistic model
+6. Conclusions
+
+**PPT structure** (14–18 slides):
+```
+Slide 1:  Title
+Slide 2:  Background & chemical problem
+Slide 3:  Research strategy (experimental + computational)
+Slide 4:  Experimental: synthesis & characterization
+Slide 5:  Experimental: performance / catalytic results
+Slide 6:  Experimental: key characterization evidence
+Slide 7:  Computational: methods & models
+Slide 8:  Computational: energetics & electronic structure
+Slide 9:  Computational: reaction pathways / mechanism
+Slide 10: Experiment-theory cross-validation
+Slide 11: Unified mechanism model
+Slide 12: Summary & outlook
+```
+
+### Step 3: Extract Figures
+
+Use `scripts/extract_charts.py` for multi-strategy figure extraction:
+
+```bash
+python <SKILL_ROOT>/scripts/extract_charts.py paper.pdf output/figures 300 --report
+```
+
+This automatically:
+1. Extracts vector graphics using `cluster_drawings()` (PyMuPDF ≥1.23) or `get_drawings()` with manual clustering (PyMuPDF 1.19+)
+2. Extracts embedded bitmap images
+3. Falls back to full-page rendering if insufficient figures found
+4. Tries multiple tolerance parameters when default clustering fails
+5. Generates `extraction_report.json` summarizing all extraction results
+
+Then map extracted figures to paper Figure numbers by checking page numbers and order.
+
+**Figure selection principles**:
+- Select only figures that support the paper's argument (typically 4–8)
+- Priority: strategy/scheme diagram → core results → mechanism → validation/controls
+- Fewer, clearer figures are better than many crowded ones
+
+### Step 4: Write Slide-by-Slide Content
+
+For each slide, prepare:
+- **Title** (conclusion-style: "Ru SAs achieve 98% CO conversion at 300°C" not just "Catalytic Performance")
+- **3–4 bullet points** (concise)
+- **Associated figure** (Figure number and file path)
+- **Figure caption** (short interpretation)
+- **One core takeaway**
+- **Speaker notes** (optional oral commentary)
+
+Each slide should convey ONE core message. Result slides should prioritize figures — let the data speak.
+
+### Step 5: Build PPTX
+
+Use `scripts/create_ppt.py`:
+
+```python
+import sys
+sys.path.insert(0, '<SKILL_ROOT>/scripts')
+from create_ppt import ChemistryPPT
+
+ppt = ChemistryPPT(theme="academic")  # "academic" | "dark" | "nature"
+
+ppt.add_title_slide(
+    title_cn="中文标题",
+    title_en="English Title",
+    authors="Authors et al.",
+    journal="J. Am. Chem. Soc., 2024, 146, xxx",
+    doi="10.xxxx/xxxx"
+)
+
+ppt.add_section_slide("Part 1: Research Background")
+
+ppt.add_content_slide(
+    title="Core Challenge in Electrocatalytic CO₂ Reduction",
+    bullets=[
+        "CO₂RR produces a wide range of products (CO, HCOOH, CH₄, C₂H₄, EtOH...) — selectivity control is difficult",
+        "Cu-based catalysts are the only metals capable of C₂₊ products, but FE is typically < 50%",
+        "Key bottleneck: simultaneous optimization of *CO binding energy and C-C coupling kinetics"
+    ],
+    notes="Emphasize the uniqueness of Cu and the selectivity challenge"
+)
+
+ppt.add_figure_slide(
+    title="HAADF-STEM Confirms Single-Atom Ru Dispersion on Cu Surface",
+    figure_path="output/figures/p3_fig1.png",
+    figure_label="Figure 1",
+    bullets=[
+        "HAADF-STEM shows Ru atoms (bright dots) uniformly dispersed — no NPs or clusters",
+        "EDS elemental mapping confirms uniform Ru distribution on Cu matrix",
+        "XANES indicates Ruᵟ⁺ oxidation state (0 < δ < 3)"
+    ],
+    caption="Source: Fig. 1a-c, adapted from original paper",
+    layout="figure_right"  # "figure_right" | "figure_top" | "figure_full"
+)
+
+ppt.add_table_slide(
+    title="Catalytic Performance Comparison",
+    headers=["Catalyst", "FE(C₂₊)%", "j (mA/cm²)", "Stability (h)", "Ref"],
+    rows=[
+        ["Ru₁/Cu", "82%", "300", "100", "This work"],
+        ["Cu NPs", "45%", "150", "20", "Nat. Catal. 2020"],
+        ["Ag/Cu", "60%", "200", "50", "JACS 2022"],
+    ]
+)
+
+ppt.add_summary_slide(
+    title="Summary & Outlook",
+    bullets=[
+        "First demonstration of Ru SAC alloy for highly selective CO₂-to-C₂₊ conversion (FE 82%)",
+        "Operando XAS + DFT reveal Ru sites promote *CO enrichment and C-C coupling",
+        "Design strategy extensible to other SAC systems (Pt/Cu, Pd/Cu)",
+        "Future: MEA testing under practical conditions"
+    ]
+)
+
+ppt.save("output/presentation.pptx")
+ppt.save_report("output/presentation.pptx")  # JSON build report
+```
+
+**Slide type reference**:
+
+| Method | Purpose |
+|--------|---------|
+| `add_title_slide()` | Cover: bilingual title, authors, journal info |
+| `add_section_slide()` | Section divider |
+| `add_content_slide()` | Text content with bullets and optional subtitle |
+| `add_figure_slide()` | Figure + explanation (4 layout options) |
+| `add_table_slide()` | Data comparison table |
+| `add_summary_slide()` | Summary/conclusions |
+| `add_thankyou_slide()` | Thank you / Q&A |
+
+### Step 6: (Optional) Generate HTML Version
+
+In addition to PPTX, generate a single-file HTML presentation with embedded figures (horizontal-slide style):
+
+```python
+import sys
+sys.path.insert(0, '<SKILL_ROOT>/scripts')
+from generate_html import HtmlPPT
+
+html = HtmlPPT(title="Academic Report", theme="molecular")
+
+# API identical to ChemistryPPT
+html.add_title_slide("Title CN", title_en="Title EN", authors="...", journal="...")
+html.add_section_slide("Part 1")
+html.add_content_slide("Key Point", ["bullet 1", "bullet 2"])
+html.add_figure_slide("Figure Title", figure_path="figures/p3_fig1.png",
+                       bullets=["description"], figure_label="Figure 1",
+                       layout="figure_right")
+html.add_summary_slide("Summary", ["finding 1", "finding 2"])
+html.add_thankyou_slide()
+
+html.save("output/presentation.html")
+```
+
+**HTML features**:
+- Single self-contained file, figures embedded as base64
+- Horizontal slide navigation: keyboard ← → Home End, scroll wheel, touch swipe, dot navigation
+- 4 academic color themes (academic / molecular / green / nature)
+- Responsive design for projectors and mobile
+- No local server or dependencies needed
+
+### Step 7: Verify and Review Reports
+
+After generating PPTX, run verification:
+
+```python
+from pptx import Presentation
+
+prs = Presentation("output/presentation.pptx")
+print(f"Slide count: {len(prs.slides)}")
+
+# Check all images are present
+for i, slide in enumerate(prs.slides):
+    for shape in slide.shapes:
+        if shape.shape_type == 13:  # Picture
+            print(f"  Slide {i+1}: image {shape.image.content_type}")
+```
+
+The `ChemistryPPT` class includes built-in error tracking:
+
+```python
+ppt.save("output/presentation.pptx")
+ppt.save_report("output/presentation.pptx")  # generates JSON report
+
+# Or access directly
+report = ppt.get_report()
+print(report["missing_images"])  # list of image paths not found
+print(report["warnings"])        # non-fatal issues
+print(report["errors"])          # actual errors
+```
+
+**Report fields**:
+- `missing_images`: all paths passed to `add_figure_slide()` that don't exist
+- `errors`: actual image insertion failures
+- `warnings`: non-fatal issues
+- `slide_types`: slide type statistics
+
+### Error Handling & Fallback Mechanisms
+
+**Figure extraction** (`extract_charts.py`):
+1. Multi-strategy: vector graphics → embedded images → page rendering fallback
+2. Compatible with PyMuPDF 1.19+ and 1.23+ (auto-switch `cluster_drawings` / `get_drawings` with manual clustering)
+3. Multi-tolerance retry (3→6→10→15→20) when default clustering finds too few figures
+4. Auto-generated JSON extraction report via `--report` flag
+
+**Paper analysis** (`analyze_paper.py`):
+1. Encoding-safe: Windows GBK/ASCII fallback to prevent Unicode crashes
+2. Confidence-annotated paper classification (high/medium/low)
+3. Structured JSON output: `--json report.json` for workflow integration
+4. Weighted keyword detection: ×3 weight in first 1/3 of text (methods/results), ×1 in references
+
+**PPT building** (`create_ppt.py`):
+1. Missing images auto-recorded (build continues), summarized on `save()`
+2. Image insertion exceptions caught (corrupt/unsupported formats)
+3. `save_report()` exports complete build log
+
+**Common issues and resolutions**:
+
+| Issue | Cause | Resolution |
+|-------|-------|------------|
+| `cluster_drawings` not found | PyMuPDF < 1.23 | Auto-fallback to `get_drawings()` manual clustering |
+| 0 vector figures extracted | Special PDF rendering | Multi-tolerance retry, ultimately page rendering |
+| Windows encoding crash | Unicode chars (e.g. `−`, `₂`) | `_safe_print()` auto-fallback to ASCII |
+| Paper type misclassification | Characterization keywords in references | Weighted detection (body ×3), confidence annotation |
+| Missing image files | Incorrect extraction or path | Recorded in `missing_images`, slide shows placeholder |
+
+---
+
+## Chemistry Academic Visual Standards
+
+See `references/visual_style.md` for full details. Summary below:
+
+### Color Schemes
+
+| Name | Primary | Accent | Best for |
+|------|---------|--------|----------|
+| **Academic Classic** (default) | `#003366` deep blue | `#B41E1E` dark red | General chemistry |
+| **Molecular Tech** | `#1A5276` steel blue | `#E74C3C` bright red | Computational/materials |
+| **Green Chemistry** | `#1E5631` deep green | `#D4A017` gold | Catalysis/energy/environment |
+| **Nature Style** | `#222222` near-black | `#0066CC` blue | CNS journal presentations |
+
+### Typography
+
+- Title: Bold sans-serif, 28–36pt
+- Body: Regular sans-serif, 16–20pt
+- English/numbers: Arial / Helvetica
+- Chemistry formulas: monospace or appropriate serif
+
+### Slide Layout
+
+- 16:9 widescreen (13.333" × 7.5")
+- Margins ≥ 0.5"
+- Left-aligned body text; titles may be centered
+- Figures get priority space; explanatory text is concise
+
+### Chemistry-Specific Elements
+
+- **Reaction schemes**: structure images or text-based with → connectors
+- **Data tables**: comparison tables for catalyst performance / computational parameters
+- **Mechanism diagrams**: keep originals, annotate key steps (i, ii, iii...)
+- **Energy diagrams**: label key TS and intermediate energies on free energy profiles
+- **Characterization data**: label key peaks (XRD), valence states (XPS), lattice spacings (TEM)
+
+---
+
+## Output Files
+
+Default output in `output/`:
+
+```
+output/
+├── presentation.pptx          # Final PPTX
+├── presentation.html          # (optional) Single-file HTML deck
+├── presentation_report.json   # PPTX build report
+├── figures/                   # Extracted figures
+│   ├── p2_fig1.png
+│   ├── p4_fig2.png
+│   └── ...
+└── qa_report.md               # Quality check report (optional)
+```
+
+---
+
+## Scripts
+
+| Script | Function |
+|--------|----------|
+| `scripts/extract_charts.py` | Multi-strategy figure extraction (vector + embedded + fallback) |
+| `scripts/analyze_paper.py` | PDF structure analysis, paper type classification, JSON output |
+| `scripts/convert_to_images.py` | PDF page to high-res image conversion |
+| `scripts/create_ppt.py` | Main script: ChemistryPPT class for academic PPTX creation |
+| `scripts/generate_html.py` | HTML generator: single-file horizontal-slide web deck |
+
+---
+
+## Chemistry Paper Type Adaptations
+
+### Catalysis Chemistry
+Focus: preparation methods, characterization (XRD/TEM/XPS/BET), activity/selectivity/stability data, TOF, activation energy, in-situ characterization, DFT-assisted mechanism
+
+### Materials Chemistry
+Focus: synthesis strategy, morphology control, structure characterization, physicochemical properties, application performance, structure-property relationships
+
+### Organic Synthesis Chemistry
+Focus: synthetic route, substrate scope, condition optimization, mechanistic probe experiments, selectivity control
+
+### Computational/Theoretical Chemistry
+Focus: computational methods and parameters, model validity, energy/structure data, comparison with experiment/benchmarks, atomic-level mechanism
+
+### Energy/Battery Chemistry
+Focus: materials design, electrochemical performance (capacity/rate/cycling), in-situ characterization, interface chemistry, degradation mechanisms
+
+### Environmental/Atmospheric Chemistry
+Focus: reaction kinetics, product analysis, mechanistic pathways, environmental implications, model calculations
+
+---
+
+## Prohibitions
+
+- Do not generate placeholder content ("Please manually add XX", "Fill in XX here")
+- Do not force all papers into the same template — adapt structure to actual paper content
+- Do not ignore correct formatting of chemical formulas / element symbols (subscripts, italics, etc.)
+- Do not create text-only result slides when figures are available
+- Do not fabricate data or conclusions not present in the paper
+- Do not produce a text-only outline instead of a usable presentation

package/assets/academic_template.html

@@ -0,0 +1,197 @@
+<!DOCTYPE html>
+<html lang="zh-CN">
+<head>
+<meta charset="UTF-8">
+<meta name="viewport" content="width=device-width, initial-scale=1.0">
+<title>__TITLE__</title>
+<style>
+  :root {
+    --primary: __PRIMARY__;
+    --primary-rgb: __PRIMARY_RGB__;
+    --accent: __ACCENT__;
+    --bg: __BG__;
+    --bg-light: __BG_LIGHT__;
+    --text: __TEXT__;
+    --muted: __MUTED__;
+    --section-bg: __SECTION_BG__;
+    --section-text: __SECTION_TEXT__;
+    --table-stripe: __TABLE_STRIPE__;
+    --slide-width: 100vw;
+    --slide-height: 100vh;
+  }
+
+  * { box-sizing: border-box; margin: 0; padding: 0; }
+
+  html, body {
+    width: 100%; height: 100%; overflow: hidden;
+    background: var(--bg); color: var(--text);
+    font-family: "Segoe UI", "Noto Sans SC", "Microsoft YaHei", Arial, sans-serif;
+    -webkit-font-smoothing: antialiased;
+  }
+
+  /* Slide container */
+  #deck {
+    position: fixed; inset: 0;
+    display: flex; flex-wrap: nowrap;
+    transition: transform 0.55s cubic-bezier(0.4, 0, 0.2, 1);
+    z-index: 10; will-change: transform;
+    width: __NSLIDES__vw;
+  }
+
+  /* Single slide */
+  .slide {
+    width: 100vw; height: 100vh;
+    flex: 0 0 100vw; position: relative;
+    padding: 5vh 5vw 7vh 5vw;
+    display: flex; flex-direction: column;
+    overflow-y: auto; overflow-x: hidden;
+  }
+  .slide.light { background: var(--bg); color: var(--text); }
+  .slide.dark { background: var(--section-bg); color: var(--section-text); }
+  .slide.accent-bg { background: var(--bg-light); color: var(--text); }
+
+  /* Title slide */
+  .slide-title { justify-content: center; align-items: flex-start; padding-top: 12vh; }
+  .slide-title .accent-line { width: 8vw; min-width: 80px; height: 3px; background: var(--accent); margin-bottom: 3vh; }
+  .slide-title h1 { font-size: clamp(24px, 3.2vw, 44px); font-weight: 700; line-height: 1.25; max-width: 85vw; }
+  .slide-title .en-title { font-size: clamp(14px, 1.5vw, 22px); color: var(--muted); margin-top: 1.5vh; max-width: 85vw; }
+  .slide-title .meta { margin-top: 6vh; font-size: clamp(12px, 1.1vw, 16px); color: var(--muted); }
+  .slide-title .meta span { display: block; margin-bottom: 0.4vh; }
+  .slide-title .page-num { position: absolute; bottom: 3vh; right: 3vw; font-size: 12px; color: var(--muted); }
+
+  /* Section divider */
+  .slide-section { justify-content: center; align-items: flex-start; padding-top: 20vh; }
+  .slide-section .accent-bar { width: 3px; height: 12vh; min-height: 60px; background: var(--accent); position: absolute; left: 3vw; top: 50%; transform: translateY(-50%); }
+  .slide-section h2 { font-size: clamp(22px, 3.6vw, 48px); font-weight: 700; margin-left: 2vw; line-height: 1.2; }
+  .slide-section .sub { font-size: clamp(12px, 1.3vw, 18px); opacity: 0.7; margin-left: 2vw; margin-top: 1.5vh; }
+
+  /* Content slide */
+  .slide-content { padding-top: 8vh; }
+  .slide-content h3 { font-size: clamp(16px, 2.2vw, 32px); font-weight: 700; color: var(--primary); margin-bottom: 0.8vh; }
+  .slide-content .title-line { width: 5vw; min-width: 40px; height: 2px; background: var(--accent); margin-bottom: 2.5vh; }
+  .slide-content .subtitle { font-size: clamp(11px, 1vw, 14px); color: var(--muted); margin-bottom: 2vh; }
+  .slide-content .bullets { list-style: none; max-width: 90%; }
+  .slide-content .bullets li {
+    font-size: clamp(13px, 1.35vw, 18px); line-height: 1.6;
+    padding: 0.8vh 0; border-bottom: 1px solid rgba(var(--muted), 0.12);
+  }
+  .slide-content .notes { position: absolute; bottom: 2.5vh; font-size: 10px; color: var(--muted); max-width: 85%; }
+  .slide-content .page-num { position: absolute; bottom: 3vh; right: 3vw; font-size: 12px; color: var(--muted); }
+
+  /* Figure slide */
+  .slide-figure { padding-top: 6vh; }
+  .slide-figure h3 { font-size: clamp(16px, 2.2vw, 32px); font-weight: 700; color: var(--primary); margin-bottom: 0.5vh; }
+  .slide-figure .title-line { width: 5vw; min-width: 40px; height: 2px; background: var(--accent); margin-bottom: 2vh; }
+  .slide-figure .fig-area { display: flex; gap: 2vw; flex: 1; min-height: 0; align-items: flex-start; }
+  .slide-figure .fig-area.top { flex-direction: column; }
+  .slide-figure .fig-area img { max-width: 100%; max-height: 55vh; object-fit: contain; border-radius: 2px; }
+  .slide-figure .fig-text { flex: 1; font-size: clamp(12px, 1.25vw, 16px); line-height: 1.6; min-width: 220px; }
+  .slide-figure .fig-text li { padding: 0.6vh 0; list-style: none; border-bottom: 1px solid rgba(var(--muted), 0.10); }
+  .slide-figure .fig-caption { font-size: 10px; color: var(--muted); margin-top: 0.8vh; }
+  .slide-figure .page-num { position: absolute; bottom: 3vh; right: 3vw; font-size: 12px; color: var(--muted); }
+
+  /* Table slide */
+  .slide-table { padding-top: 8vh; }
+  .slide-table h3 { font-size: clamp(16px, 2.2vw, 32px); font-weight: 700; color: var(--primary); margin-bottom: 0.5vh; }
+  .slide-table .title-line { width: 5vw; height: 2px; background: var(--accent); margin-bottom: 2.5vh; }
+  .slide-table table { width: 100%; border-collapse: collapse; font-size: clamp(11px, 1.1vw, 15px); }
+  .slide-table th { background: var(--primary); color: #fff; padding: 1.2vh 1vw; text-align: center; font-weight: 600; }
+  .slide-table td { padding: 0.9vh 1vw; text-align: center; border-bottom: 1px solid rgba(var(--muted), 0.15); }
+  .slide-table tr:nth-child(even) td { background: var(--table-stripe); }
+  .slide-table .page-num { position: absolute; bottom: 3vh; right: 3vw; font-size: 12px; color: var(--muted); }
+
+  /* Summary slide */
+  .slide-summary { background: var(--bg-light); padding-top: 10vh; }
+  .slide-summary .top-bar { width: 100%; height: 4px; background: var(--primary); position: absolute; top: 0; left: 0; }
+  .slide-summary h3 { font-size: clamp(18px, 2.6vw, 36px); font-weight: 700; color: var(--primary); margin-bottom: 3vh; }
+  .slide-summary .bullets { list-style: none; max-width: 88%; }
+  .slide-summary .bullets li { font-size: clamp(13px, 1.4vw, 18px); line-height: 1.55; padding: 1.2vh 0; }
+  .slide-summary .page-num { position: absolute; bottom: 3vh; right: 3vw; font-size: 12px; color: var(--muted); }
+
+  /* Thank you slide */
+  .slide-thanks { justify-content: center; align-items: center; text-align: center; background: var(--section-bg); color: var(--section-text); }
+  .slide-thanks h2 { font-size: clamp(28px, 4.5vw, 56px); font-weight: 700; }
+  .slide-thanks .sub { font-size: clamp(14px, 1.6vw, 22px); opacity: 0.65; margin-top: 2vh; }
+
+  /* Navigation dots */
+  #nav-dots { position: fixed; bottom: 2.5vh; left: 50%; transform: translateX(-50%); display: flex; gap: 8px; z-index: 100; }
+  #nav-dots .dot { width: 8px; height: 8px; border-radius: 50%; background: var(--muted); opacity: 0.35; cursor: pointer; transition: all 0.2s; }
+  #nav-dots .dot.active { opacity: 1; background: var(--accent); transform: scale(1.3); }
+
+  /* Page counter */
+  #page-counter { position: fixed; bottom: 2.8vh; right: 2.5vw; font-size: 11px; color: var(--muted); z-index: 100; font-family: monospace; }
+
+  /* Keyboard hint */
+  .kb-hint { position: fixed; bottom: 2.5vh; left: 2.5vw; font-size: 10px; color: var(--muted); z-index: 100; opacity: 0.5; }
+
+  /* Responsive */
+  @media (max-width: 768px) {
+    .slide { padding: 4vh 4vw 8vh 4vw; }
+    .slide-figure .fig-area { flex-direction: column !important; }
+    .slide-figure .fig-area img { max-height: 35vh; }
+  }
+</style>
+</head>
+<body>
+
+<div id="deck">
+  __SLIDES__
+</div>
+
+<div id="nav-dots">__NAV_DOTS__</div>
+<div id="page-counter">1 / __TOTAL__</div>
+<div class="kb-hint">← → 翻页 · Click dots · Scroll</div>
+
+<script>
+(function(){
+  var total = __TOTAL__;
+  var deck = document.getElementById('deck');
+  var dots = document.querySelectorAll('#nav-dots .dot');
+  var counter = document.getElementById('page-counter');
+  var current = 0;
+
+  deck.style.width = (total * 100) + 'vw';
+
+  function goTo(idx) {
+    if (idx < 0 || idx >= total) return;
+    current = idx;
+    deck.style.transform = 'translateX(' + (-current * 100) + 'vw)';
+    dots.forEach(function(d, i) { d.classList.toggle('active', i === current); });
+    counter.textContent = (current + 1) + ' / ' + total;
+  }
+
+  // Keyboard
+  document.addEventListener('keydown', function(e) {
+    if (e.key === 'ArrowRight' || e.key === 'ArrowDown' || e.key === ' ') { e.preventDefault(); goTo(current + 1); }
+    if (e.key === 'ArrowLeft' || e.key === 'ArrowUp') { e.preventDefault(); goTo(current - 1); }
+    if (e.key === 'Home') { e.preventDefault(); goTo(0); }
+    if (e.key === 'End') { e.preventDefault(); goTo(total - 1); }
+  });
+
+  // Dot clicks
+  dots.forEach(function(d) { d.addEventListener('click', function() { goTo(parseInt(this.dataset.idx)); }); });
+
+  // Touch swipe
+  var touchStartX = 0;
+  document.addEventListener('touchstart', function(e) { touchStartX = e.touches[0].clientX; });
+  document.addEventListener('touchend', function(e) {
+    var dx = e.changedTouches[0].clientX - touchStartX;
+    if (Math.abs(dx) > 50) { goTo(current + (dx < 0 ? 1 : -1)); }
+  });
+
+  // Scroll wheel
+  var wheelTimeout;
+  document.addEventListener('wheel', function(e) {
+    e.preventDefault();
+    if (wheelTimeout) return;
+    wheelTimeout = setTimeout(function() { wheelTimeout = null; }, 900);
+    goTo(current + (e.deltaY > 0 ? 1 : -1));
+  }, { passive: false });
+
+  // Initial
+  goTo(0);
+})();
+</script>
+
+</body>
+</html>