indoxMiner 0.0.4

Indox Data Extraction

IndoxMiner

IndoxMiner is a powerful Python library for extracting structured information from unstructured text and documents using Large Language Models (LLM). It provides a flexible schema-based approach to define and validate the information you want to extract.

Features

• 🔍 Extract structured data from text and PDFs
• 📄 Support for multiple document formats
• 🔧 Customizable extraction schemas
• ✅ Built-in validation rules
• 📊 Easy conversion to pandas DataFrames
• 🤖 Integration with OpenAI models
• 🎯 Type-safe field definitions
• 🔄 Async support for better performance

Installation

pip install indoxminer

Quick Start

Basic Text Extraction

from indoxminer import (
    ExtractorSchema,
    Field,
    FieldType,
    ValidationRule,
    OutputFormat,
    Extractor,
    OpenAi
)

# Initialize OpenAI
llm_extractor = OpenAi(api_key="your-api-key", model="gpt-4-mini")

# Define extraction schema
schema = ExtractorSchema(
    fields=[
        Field(
            name="product_name",
            description="Product name",
            field_type=FieldType.STRING,
            rules=ValidationRule(min_length=2)
        ),
        Field(
            name="price",
            description="Price in USD",
            field_type=FieldType.FLOAT,
            rules=ValidationRule(min_value=0)
        )
    ]
)

# Create extractor
extractor = Extractor(llm=llm_extractor, schema=schema)

# Extract information
text = """
MacBook Pro 16-inch with M2 chip
Price: $2,399.99
In stock: Yes
"""
result = await extractor.extract(text)

# Convert to DataFrame
df = extractor.to_dataframe(result)

PDF Document Processing

from indoxminer import DocumentProcessor, ProcessingConfig

# Initialize document processor
processor = DocumentProcessor(["invoice.pdf"])

# Process documents with configuration
documents = processor.process(
    config=ProcessingConfig(
        hi_res_pdf=True
    )
)

# Define complex schema
schema = ExtractorSchema(
    fields=[
        Field(
            name="bill_to",
            description="Bill To",
            field_type=FieldType.STRING,
            rules=ValidationRule(min_length=2)
        ),
        Field(
            name="date",
            description="date",
            field_type=FieldType.DATE,
        ),
        Field(
            name="amount",
            description="price in usd",
            field_type=FieldType.FLOAT,
        ),
    ],
    output_format=OutputFormat.JSON
)

# Extract information
results = await extractor.extract(documents)

# Handle results and validation
valid_data = results.get_valid_results()

if not results.is_valid:
    for chunk_idx, errors in results.validation_errors.items():
        print(f"Chunk {chunk_idx} has errors: {errors}")

Core Components

ExtractorSchema

Defines the structure of data to be extracted:

• fields: List of Field objects defining what to extract
• output_format: Desired output format (JSON, etc.)

Field

Defines individual data fields to extract:

• name: Field identifier
• description: Field description for the LLM
• field_type: Data type (STRING, FLOAT, INTEGER, DATE)
• rules: Validation rules for the field

ValidationRule

Sets constraints for extracted data:

• min_length: Minimum string length
• max_length: Maximum string length
• min_value: Minimum numeric value
• max_value: Maximum numeric value
• pattern: Regex pattern for validation

DocumentProcessor

Handles document processing:

• Supports multiple file formats
• Configurable processing options
• High-resolution PDF support

Configuration

ProcessingConfig

config = ProcessingConfig(
    hi_res_pdf=True,  # Enable high-resolution PDF processing
    # Add other configuration options as needed
)

Error Handling

The library provides comprehensive validation and error handling:

• Validation errors are collected per chunk
• Easy access to valid and invalid results
• Detailed error messages for debugging

Output Formats

Supported output formats:

• JSON
• Pandas DataFrame
• Raw dictionary

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Support

For issues and feature requests, please use the GitHub issue tracker.