Spaces:
Sleeping
Sleeping
File size: 11,926 Bytes
73c6377 4a17bbc 73c6377 4a17bbc 73c6377 4a17bbc 73c6377 4a17bbc 73c6377 4a17bbc 73c6377 4a17bbc 73c6377 4a17bbc 73c6377 4a17bbc 73c6377 4a17bbc 73c6377 4a17bbc 73c6377 4a17bbc 73c6377 4a17bbc 73c6377 4a17bbc 73c6377 4a17bbc |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 |
# Import required libraries
import os
from pathlib import Path
from typing import List
from langchain.schema import Document
from .config import logger
from llama_parse import LlamaParse
from llama_index.core import SimpleDirectoryReader
def load_pdf_documents(pdf_path: Path, api_key: str = None) -> List[Document]:
"""
Load and process PDF documents from medical guidelines using LlamaParse.
Excellent for borderless tables and complex medical document layouts.
Extracts disease and provider from directory structure.
Directory structure expected: data/new_data/PROVIDER/file.pdf
Example: data/new_data/SASLT/SASLT_2021.pdf
Args:
pdf_path: Path to the PDF file
api_key: LlamaCloud API key. If None, reads from LLAMA_CLOUD_API_KEY env variable
Get your API key from: https://cloud.llamaindex.ai/api-key
Returns:
List of Document objects with metadata (source, disease, provider, page_number)
"""
try:
# Validate file exists
if not pdf_path.exists():
raise FileNotFoundError(f"PDF file not found at {pdf_path}")
# Extract provider from directory structure
path_parts = pdf_path.parts
disease = "HBV" # Default disease for this system
provider = "unknown"
# Find provider: it's the parent directory of the PDF file
if len(path_parts) >= 2:
provider = path_parts[-2] # Parent directory (e.g., SASLT)
# If provider is 'new_data', it means file is directly in new_data folder
if provider.lower() == "new_data":
provider = "unknown"
# Get API key from parameter or environment variable
llama_api_key = api_key or os.getenv("LLAMA_CLOUD_API_KEY")
if not llama_api_key:
raise ValueError(
"LlamaCloud API key not found. Please provide api_key parameter or set "
"LLAMA_CLOUD_API_KEY environment variable. "
"Get your key from: https://cloud.llamaindex.ai/api-key"
)
# Initialize LlamaParse with optimized settings for medical documents
parser = LlamaParse(
api_key=llama_api_key,
result_type="markdown", # or "text" for plain text
verbose=True,
language="en",
# Medical document optimizations
parsing_instruction=(
"This is a medical guideline document. "
"Pay special attention to tables (including borderless tables), "
"clinical recommendations, dosage information, and reference citations. "
"Preserve table structure and maintain hierarchical headings."
),
# Advanced options for better table handling
invalidate_cache=False, # Use cache for faster re-processing
do_not_cache=False,
fast_mode=False, # Use deep parsing for better accuracy
# Split by page for proper page numbering
split_by_page=True, # This is the key parameter!
)
# Parse the PDF file
logger.info(f"Parsing PDF with LlamaParse: {pdf_path.name}")
# Use SimpleDirectoryReader with LlamaParse
file_extractor = {".pdf": parser}
reader = SimpleDirectoryReader(
input_files=[str(pdf_path)],
file_extractor=file_extractor
)
# Load documents - each page will be a separate document when split_by_page=True
llama_documents = reader.load_data()
# Convert to LangChain Document format
documents = []
for doc_idx, llama_doc in enumerate(llama_documents):
# When split_by_page=True, each llama_doc represents one page
# Check if page number exists in metadata, otherwise use index
page_num = llama_doc.metadata.get('page_number', doc_idx + 1)
processed_doc = Document(
page_content=llama_doc.text.strip(),
metadata={
"source": pdf_path.name,
"disease": disease,
"provider": provider,
"page_number": page_num,
"document_index": doc_idx,
# Preserve any additional metadata from LlamaParse
**{k: v for k, v in llama_doc.metadata.items()
if k not in ['source', 'disease', 'provider', 'page_number', 'document_index']}
}
)
documents.append(processed_doc)
logger.info(f"Loaded {len(documents)} pages from PDF: {pdf_path.name} (Disease: {disease}, Provider: {provider})")
return documents
except Exception as e:
logger.error(f"Error loading PDF documents from {pdf_path}: {str(e)}")
raise
def load_pdf_documents_advanced(
pdf_path: Path,
api_key: str = None,
premium_mode: bool = False
) -> List[Document]:
"""
Advanced version with premium features for complex medical documents.
Args:
pdf_path: Path to the PDF file
api_key: LlamaCloud API key
premium_mode: Use premium GPT-4o mode for highest accuracy (costs more)
Returns:
List of Document objects with enhanced metadata
"""
try:
if not pdf_path.exists():
raise FileNotFoundError(f"PDF file not found at {pdf_path}")
path_parts = pdf_path.parts
disease = "HBV"
provider = path_parts[-2] if len(path_parts) >= 2 else "unknown"
if provider.lower() == "new_data":
provider = "unknown"
llama_api_key = api_key or os.getenv("LLAMA_CLOUD_API_KEY")
if not llama_api_key:
raise ValueError("LlamaCloud API key required")
# Advanced parser configuration
parser = LlamaParse(
api_key=llama_api_key,
result_type="markdown",
verbose=True,
language="en",
# Premium mode uses GPT-4o for better accuracy
premium_mode=premium_mode,
# Detailed parsing instructions for medical content
parsing_instruction=(
"Medical guideline document with complex tables. Instructions:\n"
"0. Keep the original text intact without changing anything\n"
"1. Preserve all table structures, especially borderless tables\n"
"2. Maintain hierarchical organization of sections and subsections\n"
"3. Keep dosage tables and treatment algorithms intact\n"
"4. Preserve reference numbers and citations\n"
"5. Identify and mark clinical recommendation levels\n"
"6. Extract figures and their captions accurately"
),
# Extract structured data
take_screenshot=True, # Capture page screenshots for reference
# Table-specific optimizations
invalidate_cache=False,
do_not_cache=False,
fast_mode=False,
# Critical: split by page for accurate page numbering
split_by_page=True,
)
file_extractor = {".pdf": parser}
reader = SimpleDirectoryReader(
input_files=[str(pdf_path)],
file_extractor=file_extractor
)
logger.info(f"Parsing PDF with LlamaParse (Premium: {premium_mode}): {pdf_path.name}")
llama_documents = reader.load_data()
documents = []
for doc_idx, llama_doc in enumerate(llama_documents):
# Get page number from metadata or use index
page_num = llama_doc.metadata.get('page_number', doc_idx + 1)
# Enhanced metadata
metadata = {
"source": pdf_path.name,
"disease": disease,
"provider": provider,
"page_number": page_num,
"document_index": doc_idx + 1,
"parser": "llamaparse",
"premium_mode": premium_mode
}
# Add additional metadata from LlamaIndex document
if hasattr(llama_doc, 'metadata'):
# Merge additional metadata, avoiding duplicates
for key, value in llama_doc.metadata.items():
if key not in metadata:
metadata[key] = value
processed_doc = Document(
page_content=llama_doc.text.strip(),
metadata=metadata
)
documents.append(processed_doc)
logger.info(f"Loaded {len(documents)} pages from PDF: {pdf_path.name}")
return documents
except Exception as e:
logger.error(f"Error loading PDF documents from {pdf_path}: {str(e)}")
raise
# Batch processing function for multiple PDFs
def load_multiple_pdfs(
pdf_directory: Path,
api_key: str = None,
file_pattern: str = "*.pdf"
) -> List[Document]:
"""
Load multiple PDF files from a directory.
Args:
pdf_directory: Directory containing PDF files
api_key: LlamaCloud API key
file_pattern: Glob pattern for PDF files (default: "*.pdf")
Returns:
List of all documents from all PDFs
"""
all_documents = []
pdf_files = list(pdf_directory.glob(file_pattern))
logger.info(f"Found {len(pdf_files)} PDF files to process")
for pdf_path in pdf_files:
try:
documents = load_pdf_documents(pdf_path, api_key=api_key)
all_documents.extend(documents)
logger.info(f"Successfully processed: {pdf_path.name}")
except Exception as e:
logger.error(f"Failed to process {pdf_path.name}: {str(e)}")
continue
logger.info(f"Total documents loaded: {len(all_documents)}")
return all_documents
def load_markdown_documents(md_path: Path) -> List[Document]:
"""
Load and process Markdown medical guidelines.
Extracts disease and provider from directory structure.
Directory structure expected: data/new_data/PROVIDER/file.md
Example: data/new_data/SASLT/guidelines.md
Args:
md_path: Path to the Markdown file
Returns:
List of Document objects with metadata (source, disease, provider, page_number)
"""
try:
# Validate file exists
if not md_path.exists():
raise FileNotFoundError(f"Markdown file not found at {md_path}")
# Extract provider from directory structure
# Structure: data/new_data/PROVIDER/file.md
path_parts = md_path.parts
disease = "HBV" # Default disease for this system
provider = "unknown"
# Find provider: it's the parent directory of the markdown file
if len(path_parts) >= 2:
provider = path_parts[-2] # Parent directory (e.g., SASLT)
# If provider is 'new_data', it means file is directly in new_data folder
if provider.lower() == "new_data":
provider = "unknown"
# Read markdown content
with open(md_path, 'r', encoding='utf-8') as f:
content = f.read()
# Create document with minimal metadata for RAG
doc = Document(
page_content=content,
metadata={
"source": md_path.name,
"disease": disease,
"provider": provider,
"page_number": 1
}
)
logger.info(f"Loaded Markdown document: {md_path.name} (Disease: {disease}, Provider: {provider})")
return [doc]
except Exception as e:
logger.error(f"Error loading Markdown document from {md_path}: {str(e)}")
raise
|