LLM-ChatBot-Document / utils /document_viewer.py
FauzanAriyatmoko's picture
feat: Implement an interactive document viewer with citation highlighting and structured PDF text extraction.
a86d063
"""
Document viewer component for displaying PDFs with citation highlighting
"""
from typing import List, Dict, Optional
import json
class DocumentViewer:
"""Handle document viewing with citation highlighting"""
def __init__(self):
"""Initialize document viewer"""
self.current_document = None
self.current_highlights = []
def render_document(self, html_content: str, filename: str, highlight_paragraphs: List[str] = None) -> str:
"""
Render document with optional paragraph highlighting
Args:
html_content: HTML content of the document
filename: Name of the document
highlight_paragraphs: List of paragraph IDs to highlight
Returns:
Enhanced HTML with highlighting
"""
if not html_content:
return self._render_empty_state()
# Add highlighting script and marks
if highlight_paragraphs:
for para_id in highlight_paragraphs:
# Add highlighted class to specific paragraphs
html_content = html_content.replace(
f'<p class="paragraph" id="{para_id}"',
f'<p class="paragraph highlighted-citation" id="{para_id}"'
)
# Wrap with viewer container and add controls
enhanced_html = f"""
<div class="document-viewer-wrapper">
{self._create_viewer_controls(filename)}
<div class="document-viewer-content" id="doc-viewer-content">
{html_content}
</div>
</div>
<script>
// Auto-scroll to first highlighted paragraph
document.addEventListener('DOMContentLoaded', function() {{
const firstHighlight = document.querySelector('.highlighted-citation');
if (firstHighlight) {{
setTimeout(function() {{
firstHighlight.scrollIntoView({{ behavior: 'smooth', block: 'center' }});
}}, 300);
}}
}});
// Smooth scroll when clicking citation links
function scrollToParagraph(paraId) {{
const element = document.getElementById(paraId);
if (element) {{
// Remove previous highlights
document.querySelectorAll('.highlighted-citation').forEach(el => {{
el.classList.remove('highlighted-citation');
}});
// Add highlight to clicked citation
element.classList.add('highlighted-citation');
// Scroll to element
element.scrollIntoView({{ behavior: 'smooth', block: 'center' }});
// Flash effect
element.style.animation = 'none';
setTimeout(() => {{
element.style.animation = 'highlight-flash 1s ease';
}}, 10);
}}
}}
</script>
"""
return enhanced_html
def _create_viewer_controls(self, filename: str) -> str:
"""
Create viewer control bar
Args:
filename: Current document filename
Returns:
HTML for controls
"""
return f"""
<div class="viewer-controls">
<div class="viewer-title">
<span class="doc-icon">📄</span>
<span class="doc-name">{filename}</span>
</div>
<div class="viewer-actions">
<button class="viewer-btn" onclick="document.getElementById('doc-viewer-content').style.fontSize='0.9em';" title="Zoom Out">
🔍−
</button>
<button class="viewer-btn" onclick="document.getElementById('doc-viewer-content').style.fontSize='1em';" title="Reset Zoom">
🔍
</button>
<button class="viewer-btn" onclick="document.getElementById('doc-viewer-content').style.fontSize='1.1em';" title="Zoom In">
🔍+
</button>
</div>
</div>
"""
def _render_empty_state(self) -> str:
"""
Render empty state when no document is selected
Returns:
HTML for empty state
"""
return """
<div class="document-viewer-empty">
<div class="empty-state-icon">📄</div>
<h3>Tidak Ada Dokumen</h3>
<p>Upload dokumen PDF untuk melihat konten dan sitasi di sini.</p>
</div>
"""
def create_citation_link(self, filename: str, paragraph_ids: List[str], snippet: str, page: int = None) -> str:
"""
Create clickable citation link for chat
Args:
filename: Source document filename
paragraph_ids: List of paragraph IDs this citation refers to
snippet: Text snippet to show
page: Page number (optional)
Returns:
HTML citation link
"""
para_id = paragraph_ids[0] if paragraph_ids else "unknown"
page_info = f" (Hal. {page})" if page else ""
citation_html = f"""
<div class="citation-card" onclick="scrollToParagraph('{para_id}')">
<div class="citation-header">
<strong>📄 {filename}{page_info}</strong>
</div>
<div class="citation-snippet">
"{snippet[:150]}..."
</div>
</div>
"""
return citation_html
def format_sources_with_links(self, sources: List[Dict]) -> tuple[str, List[str]]:
"""
Format sources as interactive citations
Args:
sources: List of source metadata from RAG pipeline
Returns:
Tuple of (HTML string, list of paragraph IDs to highlight)
"""
if not sources:
return "", []
all_paragraph_ids = []
html = "<div class='sources-container'>"
html += "<h4 class='sources-title'>📚 Sumber Referensi:</h4>"
for i, source in enumerate(sources, 1):
filename = source.get('filename', 'Unknown')
chunk_text = source.get('chunk_text', '')
# Parse paragraph IDs if available
paragraph_ids_str = source.get('paragraph_ids', '[]')
try:
if isinstance(paragraph_ids_str, str):
paragraph_ids = json.loads(paragraph_ids_str)
else:
paragraph_ids = paragraph_ids_str if isinstance(paragraph_ids_str, list) else []
except:
paragraph_ids = []
# Parse pages
pages_str = source.get('pages', '[]')
try:
if isinstance(pages_str, str):
pages = json.loads(pages_str)
else:
pages = pages_str if isinstance(pages_str, list) else []
except:
pages = []
page = pages[0] if pages else None
# Track all paragraph IDs for highlighting
all_paragraph_ids.extend(paragraph_ids)
# Create citation link
html += self.create_citation_link(
filename=filename,
paragraph_ids=paragraph_ids,
snippet=chunk_text,
page=page
)
html += "</div>"
return html, list(set(all_paragraph_ids)) # Return unique paragraph IDs
def create_document_selector(self, documents: List[Dict], current_doc: str = None) -> str:
"""
Create dropdown selector for documents
Args:
documents: List of document metadata
current_doc: Currently selected document filename
Returns:
HTML for document selector
"""
if not documents:
return "<p class='no-docs-message'>Belum ada dokumen yang tersedia</p>"
html = """
<div class="document-selector">
<label for="doc-select">Pilih Dokumen:</label>
<select id="doc-select" class="doc-select-dropdown">
"""
for doc in documents:
filename = doc.get('filename', 'Unknown')
num_pages = doc.get('num_pages', 0)
selected = 'selected' if filename == current_doc else ''
html += f"""
<option value="{filename}" {selected}>
{filename} ({num_pages} hal.)
</option>
"""
html += """
</select>
</div>
"""
return html