Reruns Post-Extraction

Overview

After extracting a document, you don’t need to start from scratch to adjust your results. The Pulse Playground supports several types of reruns that let you iterate on processing settings, schemas, and splits without re-uploading or re-extracting the document.

Types of Reruns

1. Full Re-Extraction

Re-processes the entire document from scratch with new extraction settings. This re-uploads the original file and runs a fresh extraction. When to use:

You want to change extraction settings (page range, figure extraction, chunking)
The original extraction had issues you want to fix with different parameters

How it works in the Playground:

Full re-extraction creates a new extraction — it does not modify the original. You’ll be navigated to the new extraction result automatically.

API equivalent:

# Re-extract with different settings
result = client.extract(
    file=open("document.pdf", "rb"),
    pages="1-10",           # different page range
    async_=True
)

2. Schema-Only Rerun

Applies a new or modified schema to an existing extraction without re-processing the document. This is the most common type of rerun — fast and cost-effective because extraction is not repeated. When to use:

You want to try a different schema on the same document
You need to fix or refine extracted fields
You want to extract additional fields from an already-processed document

How it works in the Playground:

Schema reruns are fast because they operate on the already-extracted content stored in Pulse — no document re-processing needed. You can iterate on schemas as many times as you want.

API equivalent:

# Apply a new schema to an existing extraction
schema_result = client.schema(
    extraction_id="existing-extraction-id",
    schema_config={
        "input_schema": {
            "type": "object",
            "properties": {
                "total_amount": {"type": "number"},
                "due_date": {"type": "string", "format": "date"}
            }
        },
        "schema_prompt": "Extract payment details"
    }
)

print(schema_result.schema_output)

Each schema rerun creates a new schema version, letting you compare results across different schema configurations.

3. Split Rerun

Re-runs the document split with different topics on an existing extraction. Useful when you want to reorganize how the document is sectioned. When to use:

You want to try different topic definitions
You want to add or remove topics
The initial split didn’t assign pages correctly and you want to adjust topic descriptions

How it works in the Playground:

API equivalent:

# Re-split with different topics
split_result = client.split(
    extraction_id="existing-extraction-id",
    split_config={
        "split_input": [
            {"name": "Executive Summary", "description": "High-level overview"},
            {"name": "Financial Details", "description": "Revenue, costs, margins"},
            {"name": "Risk Factors", "description": "Identified risks and mitigations"}
        ]
    }
)

print(split_result.split_output)

Rerun Workflow Summary

Rerun Type	Re-processes document?	Creates new extraction?	Speed
Full Re-Extraction	✅ Yes	✅ Yes	Slow (full pipeline)
Schema-Only	❌ No	❌ No	Fast (seconds)
Split Rerun	❌ No	❌ No	Fast (seconds)

Schema Versioning

Each schema rerun creates a versioned result. In the Playground, you can:

Switch between versions using the version selector in the Schema tab
Compare results across different schema configurations
Roll back to a previous version if a new schema didn’t produce better results

This makes it easy to iteratively refine your schema without losing previous work.

Choose Your Workflow

Understanding the main pipeline configurations

Schema Endpoint

API reference for the /schema endpoint

Platform Reference

Pipelines

Forms

Production Workflow

Library & Exports

Reruns Post-Extraction

Overview

Types of Reruns

1. Full Re-Extraction

2. Schema-Only Rerun

3. Split Rerun

Rerun Workflow Summary

Schema Versioning

Choose Your Workflow

Schema Endpoint

​Overview

​Types of Reruns

​1. Full Re-Extraction

​2. Schema-Only Rerun

​3. Split Rerun

​Rerun Workflow Summary

​Schema Versioning

​Related

Choose Your Workflow

Schema Endpoint

Overview

Types of Reruns

1. Full Re-Extraction

2. Schema-Only Rerun

3. Split Rerun

Rerun Workflow Summary

Schema Versioning

Related