Parze Python SDK

Official Python client for the Parze document parsing API.

Installation

pip install parze

Quick Start

from parze import ParzeClient

# Initialize client with your API key
client = ParzeClient(api_key="pk_live_your_key_here")

# Parse a document
result = client.parse("invoice.pdf")
print(result["text"])

# Extract structured data (one step)
schema = {
    "invoice_number": {"type": "string", "description": "Invoice number"},
    "total_amount": {"type": "string", "description": "Total amount"},
    "date": {"type": "string", "description": "Invoice date"}
}
extraction = client.extract(file="invoice.pdf", extraction_schema=schema)
print(extraction["extraction"])

# To avoid double billing, parse first and then extract with job_id
parse_result = client.parse("invoice.pdf")
extraction = client.extract(parse_result["text"], schema, parse_result["job_id"])

# Validate document quality (pre-validation)
validation = client.validate(
    "invoice.pdf",
    validation_type="pre",
    validation_rules={
        "quality_checks": {
            "min_resolution": 150,
            "check_readability": True,
            "check_completeness": True
        }
    }
)
print(validation)

# Get AI-suggested schema
suggested = client.suggest_schema(text)
print(suggested)

API Reference

parse(file, output_format="structured", preserve_tables=True, extraction_mode="auto")

Parse a document into structured text.

Parameters:

• file (str or file object): Path to file or file object
• output_format (str): "structured", "markdown", or "json"
• preserve_tables (bool): Preserve table structure
• extraction_mode (str): "auto", "ocr_only", "llm_only", or "identity_doc"

Returns: Dict with parsed text and metadata

extract(text=None, extraction_schema=None, job_id=None, file=None, extraction_mode="auto", preserve_tables=True)

Extract structured data from a file or parsed text using a schema.

Parameters:

• text (str, optional): Document text (from parse)
• extraction_schema (dict, required): Schema defining fields to extract
• job_id (str, required if text is provided): Job ID from parse response
• file (str or file object, optional): Path to file or file object (if provided, parse runs internally)
• extraction_mode (str, optional): "auto", "ocr_only", "llm_only", or "identity_doc" (file-based only)
• preserve_tables (bool, optional): Preserve table structure during parsing (file-based only)

Returns: Dict with extracted data and confidence scores

suggest_schema(text)

Get AI-suggested extraction schema based on document text.

Parameters:

• text (str): Document text

Returns: Dict with suggested schema

text_to_schema(description)

Convert natural language description to extraction schema.

Parameters:

• description (str): Natural language description of fields

Returns: Dict with generated schema

validate(files, validation_type="pre", validation_rules=None, extraction_schema=None, job_id=None)

Validate document quality (pre) or extracted data (post).

Parameters:

• files (str, file object, or list): Path(s) to file(s) or file objects
• validation_type (str): "pre" or "post"
• validation_rules (dict, optional): Validation rules payload
• extraction_schema (dict, optional): Required for post-validation
• job_id (str, optional): Job ID from parse

Returns: Dict with validation results

Get API Key

Get your API key from platform.parze.ai