granite-docling-demo / granite_docling.py
fmeres's picture
Initial commit: Granite Docling 258M online demo
40e7f18
"""
Granite Docling 258M Implementation
This module provides an interface to the IBM Granite Docling 258M model
for document processing and conversion tasks.
"""
import os
import logging
import time
from pathlib import Path
from typing import Union, Optional, Dict, Any
from docling.document_converter import DocumentConverter, PdfFormatOption
from docling.datamodel.base_models import InputFormat
from docling.datamodel.pipeline_options import (
PdfPipelineOptions,
VlmPipelineOptions,
ResponseFormat,
AcceleratorDevice,
vlm_model_specs
)
from docling.pipeline.vlm_pipeline import VlmPipeline
# Additional imports for fast document analysis
try:
import fitz # PyMuPDF for fast PDF metadata extraction
PYMUPDF_AVAILABLE = True
except ImportError:
PYMUPDF_AVAILABLE = False
try:
from PIL import Image
PIL_AVAILABLE = True
except ImportError:
PIL_AVAILABLE = False
# Set up logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
class GraniteDocling:
"""
A wrapper class for the IBM Granite Docling 258M model.
This class provides an easy-to-use interface for document processing
using the Granite Docling model through the Docling framework.
"""
def __init__(
self,
model_type: str = "transformers",
artifacts_path: Optional[str] = None
):
"""
Initialize the Granite Docling processor.
Args:
model_type: Model type - "transformers" or "mlx"
artifacts_path: Path to cached model artifacts
"""
self.model_type = model_type.lower()
self.artifacts_path = artifacts_path
# Choose the appropriate model configuration
if self.model_type == "mlx":
self.vlm_model = vlm_model_specs.GRANITEDOCLING_MLX
else:
self.vlm_model = vlm_model_specs.GRANITEDOCLING_TRANSFORMERS
# Initialize the document converter
self._setup_converter()
def _setup_converter(self):
"""Set up the document converter with Granite Docling configuration."""
# Set up VLM pipeline options using the pre-configured Granite Docling model
pipeline_options = VlmPipelineOptions(vlm_options=self.vlm_model)
# Configure PDF processing options
pdf_options = PdfFormatOption(
pipeline_cls=VlmPipeline,
pipeline_options=pipeline_options,
)
# If artifacts path is specified, add it to PDF pipeline options
if self.artifacts_path:
pdf_pipeline_options = PdfPipelineOptions(artifacts_path=self.artifacts_path)
pdf_options.pipeline_options = pdf_pipeline_options
# Initialize the document converter
self.converter = DocumentConverter(
format_options={
InputFormat.PDF: pdf_options,
}
)
logger.info(f"Initialized Granite Docling with model type: {self.model_type}")
def analyze_document_structure(
self,
source: Union[str, Path],
sample_pages: int = 3,
max_sample_chars: int = 2000
) -> Dict[str, Any]:
"""
Fast document structure analysis without full conversion.
This method provides lightweight document insights including:
- Basic metadata (pages, size, type)
- Structure detection (headers, tables, images)
- Content sampling from first few pages
- Performance optimized for large documents
Args:
source: Path to the document
sample_pages: Number of pages to sample for content analysis
max_sample_chars: Maximum characters to extract for preview
Returns:
Dictionary containing document analysis and structure information
"""
start_time = time.time()
try:
source_path = Path(source)
logger.info(f"Analyzing document structure: {source}")
# Initialize analysis result
analysis_result = {
"source": str(source),
"file_name": source_path.name,
"file_size_mb": round(source_path.stat().st_size / (1024 * 1024), 2),
"analysis_time_seconds": 0,
"document_type": source_path.suffix.lower(),
"structure_detected": {},
"content_preview": "",
"metadata_extraction": {},
"processing_approach": "fast_analysis"
}
# PDF-specific fast analysis
if source_path.suffix.lower() == '.pdf' and PYMUPDF_AVAILABLE:
analysis_result.update(self._analyze_pdf_structure(source, sample_pages, max_sample_chars))
# Image file analysis
elif source_path.suffix.lower() in ['.png', '.jpg', '.jpeg', '.bmp', '.tiff'] and PIL_AVAILABLE:
analysis_result.update(self._analyze_image_structure(source))
# For other formats, use docling but with limited sampling
else:
analysis_result.update(self._analyze_other_format_structure(source, sample_pages, max_sample_chars))
analysis_result["analysis_time_seconds"] = round(time.time() - start_time, 2)
logger.info(f"Document analysis completed in {analysis_result['analysis_time_seconds']} seconds")
return analysis_result
except Exception as e:
logger.error(f"Error analyzing document structure {source}: {str(e)}")
return {
"source": str(source),
"error": str(e),
"analysis_time_seconds": round(time.time() - start_time, 2),
"processing_approach": "fast_analysis_failed"
}
def _analyze_pdf_structure(self, source: Union[str, Path], sample_pages: int, max_sample_chars: int) -> Dict[str, Any]:
"""Fast PDF structure analysis using PyMuPDF."""
try:
doc = fitz.open(str(source))
total_pages = doc.page_count
# Extract metadata
metadata = doc.metadata
# Sample pages for structure analysis
pages_to_sample = min(sample_pages, total_pages)
sample_text = ""
headers_found = []
tables_detected = 0
images_detected = 0
text_density_avg = 0
for page_num in range(pages_to_sample):
page = doc[page_num]
# Get text content
page_text = page.get_text()
sample_text += page_text[:max_sample_chars // pages_to_sample] + "\n"
# Detect structure elements
text_dict = page.get_text("dict")
# Count images
images_detected += len(page.get_images())
# Estimate text density
text_density_avg += len(page_text.strip()) / max(1, page.rect.width * page.rect.height) * 10000
# Simple header detection (large/bold text)
for block in text_dict.get("blocks", []):
if "lines" in block:
for line in block["lines"]:
for span in line.get("spans", []):
text = span.get("text", "").strip()
if text and len(text) < 100: # Potential header
font_size = span.get("size", 12)
font_flags = span.get("flags", 0)
# Check if text looks like a header (large font or bold)
if font_size > 14 or (font_flags & 2**4): # Bold flag
headers_found.append(text)
# Simple table detection (look for aligned text patterns)
tables_detected += self._estimate_tables_in_page_text(page_text)
doc.close()
text_density_avg = round(text_density_avg / pages_to_sample, 2) if pages_to_sample > 0 else 0
return {
"total_pages": total_pages,
"pages_analyzed": pages_to_sample,
"metadata_extraction": {
"title": metadata.get("title", ""),
"author": metadata.get("author", ""),
"creation_date": metadata.get("creationDate", ""),
"modification_date": metadata.get("modDate", "")
},
"structure_detected": {
"headers_found": len(set(headers_found)),
"sample_headers": list(set(headers_found))[:5],
"estimated_tables": tables_detected,
"images_detected": images_detected,
"text_density": text_density_avg,
"has_text": len(sample_text.strip()) > 50
},
"content_preview": sample_text[:max_sample_chars].strip()
}
except Exception as e:
logger.warning(f"PyMuPDF analysis failed, falling back: {e}")
return self._analyze_other_format_structure(source, sample_pages, max_sample_chars)
def _analyze_image_structure(self, source: Union[str, Path]) -> Dict[str, Any]:
"""Fast image file analysis."""
try:
with Image.open(source) as img:
return {
"total_pages": 1,
"pages_analyzed": 1,
"metadata_extraction": {
"format": img.format,
"mode": img.mode,
"size": f"{img.size[0]}x{img.size[1]}",
"has_exif": bool(getattr(img, '_getexif', lambda: None)())
},
"structure_detected": {
"content_type": "image",
"requires_ocr": True,
"estimated_text_content": "unknown_until_ocr"
},
"content_preview": f"Image file: {img.format} format, {img.size[0]}x{img.size[1]} pixels"
}
except Exception as e:
logger.warning(f"Image analysis failed: {e}")
return {
"total_pages": 1,
"structure_detected": {"content_type": "image", "analysis_failed": str(e)},
"content_preview": "Image analysis failed"
}
def _analyze_other_format_structure(self, source: Union[str, Path], sample_pages: int, max_sample_chars: int) -> Dict[str, Any]:
"""Lightweight analysis for other formats using minimal docling processing."""
try:
# Use docling but process minimally - just get basic structure
result = self.converter.convert(source=str(source))
document = result.document
# Get basic info without full markdown conversion
total_pages = len(document.pages) if hasattr(document, 'pages') else 1
# Sample first few pages only
pages_to_analyze = min(sample_pages, total_pages)
sample_content = ""
if hasattr(document, 'pages'):
for i in range(pages_to_analyze):
if i < len(document.pages):
page = document.pages[i]
# Get text content from page without full markdown processing
if hasattr(page, 'text'):
sample_content += str(page.text)[:max_sample_chars // pages_to_analyze] + "\n"
# If we still don't have content, do a quick markdown export of first portion
if not sample_content:
full_content = document.export_to_markdown()
sample_content = full_content[:max_sample_chars]
# Quick structure analysis
headers_found = [line.strip() for line in sample_content.split('\n') if line.strip().startswith('#')]
table_lines = [line for line in sample_content.split('\n') if '|' in line and line.strip()]
return {
"total_pages": total_pages,
"pages_analyzed": pages_to_analyze,
"structure_detected": {
"headers_found": len(headers_found),
"sample_headers": headers_found[:5],
"estimated_tables": len([line for line in table_lines if line.count('|') > 1]),
"has_markdown_structure": len(headers_found) > 0 or len(table_lines) > 0
},
"content_preview": sample_content.strip()
}
except Exception as e:
logger.warning(f"Docling lightweight analysis failed: {e}")
return {
"total_pages": 1,
"structure_detected": {"analysis_method": "file_info_only"},
"content_preview": "Unable to analyze document structure"
}
def _estimate_tables_in_page_text(self, text: str) -> int:
"""Estimate number of tables in text by looking for aligned patterns."""
lines = text.split('\n')
potential_table_lines = 0
for line in lines:
# Look for lines with multiple whitespace-separated columns
parts = line.strip().split()
if len(parts) >= 3: # At least 3 columns
# Check if parts look like tabular data (numbers, short text)
if any(part.replace('.', '').replace(',', '').isdigit() for part in parts):
potential_table_lines += 1
# Rough estimate: every 5+ aligned lines might be a table
return potential_table_lines // 5
def convert_document(
self,
source: Union[str, Path],
output_format: str = "markdown"
) -> Dict[str, Any]:
"""
Convert a document using the Granite Docling model.
Args:
source: Path to the document or URL
output_format: Output format (currently supports 'markdown')
Returns:
Dictionary containing the conversion result and metadata
"""
try:
logger.info(f"Converting document: {source}")
# Convert the document
result = self.converter.convert(source=str(source))
document = result.document
# Extract the converted content
if output_format.lower() == "markdown":
content = document.export_to_markdown()
else:
content = str(document)
# Prepare result dictionary
conversion_result = {
"content": content,
"source": str(source),
"format": output_format,
"pages": len(document.pages) if hasattr(document, 'pages') else 1,
"metadata": {
"model_type": self.model_type,
"model_config": str(self.vlm_model.__class__.__name__)
}
}
logger.info(f"Successfully converted document with {conversion_result['pages']} pages")
return conversion_result
except Exception as e:
logger.error(f"Error converting document {source}: {str(e)}")
raise
def convert_to_file(
self,
source: Union[str, Path],
output_path: Union[str, Path],
output_format: str = "markdown"
) -> Dict[str, Any]:
"""
Convert a document and save the result to a file.
Args:
source: Path to the input document or URL
output_path: Path where the converted document will be saved
output_format: Output format (currently supports 'markdown')
Returns:
Dictionary containing the conversion result and metadata
"""
# Convert the document
result = self.convert_document(source, output_format)
# Save to file
output_path = Path(output_path)
output_path.parent.mkdir(parents=True, exist_ok=True)
with open(output_path, 'w', encoding='utf-8') as f:
f.write(result["content"])
result["output_path"] = str(output_path)
logger.info(f"Saved converted document to: {output_path}")
return result
def batch_convert(
self,
sources: list,
output_dir: Union[str, Path],
output_format: str = "markdown"
) -> list:
"""
Convert multiple documents in batch.
Args:
sources: List of document paths or URLs
output_dir: Directory to save converted documents
output_format: Output format for all documents
Returns:
List of conversion results
"""
output_dir = Path(output_dir)
output_dir.mkdir(parents=True, exist_ok=True)
results = []
for source in sources:
try:
# Generate output filename
source_path = Path(source)
if output_format.lower() == "markdown":
output_filename = source_path.stem + ".md"
else:
output_filename = source_path.stem + f".{output_format}"
output_path = output_dir / output_filename
# Convert and save
result = self.convert_to_file(source, output_path, output_format)
results.append(result)
except Exception as e:
logger.error(f"Failed to convert {source}: {str(e)}")
results.append({
"source": str(source),
"error": str(e),
"success": False
})
return results
def download_models():
"""Download the required Granite Docling models."""
try:
import subprocess
logger.info("Downloading Granite Docling models...")
subprocess.run([
"docling-tools", "models", "download-hf-repo",
"ibm-granite/granite-docling-258M"
], check=True)
logger.info("Models downloaded successfully!")
except subprocess.CalledProcessError as e:
logger.error(f"Failed to download models: {e}")
raise
except FileNotFoundError:
logger.error("docling-tools not found. Please install docling first.")
raise
if __name__ == "__main__":
# Example usage
granite = GraniteDocling()
# Example conversion (replace with actual document path)
# result = granite.convert_document("path/to/document.pdf")
# print(result["content"])