Skip to main content

​
Goal

Extract different types of structured data from different sections of a long report without forcing one huge schema across the whole document.

​
Sample Document

Use the built-in 10-K Annual Report Platform example. The example includes a saved Split output with topics for exhibits and financial statement schedules, signatures, and certifications.

​
Use This Workflow

Use Extract -> Split -> Schema when sections have different vocabulary, layout, and output requirements.

​
Split Topics

TopicDescription
Financial Statement SchedulesFinancial statement schedules, exhibits, and supporting tables.
SignaturesSignature blocks, officers, titles, and signing dates.
CertificationsOfficer certifications and compliance attestations.

​
Platform Steps

1

Extract the report

Upload the report and run Extract. Use async for long PDFs.
2

Add Split

Add topics with clear names and descriptions. Run Split and inspect page assignments.
3

Add per-topic schemas

Define a different schema for each topic. Keep each schema narrow and specific.
4

Save presets

Save the split config and per-topic schema configs once the workflow works across multiple reports.

​
Python

from pulse import Pulse

client = Pulse(api_key="YOUR_API_KEY")

extract_result = client.extract(
    file_url="https://platform.runpulse.com/api/examples/0514fc05-8b0a-4a3b-9b9b-18ac89fc04e5/pdf",
    async_=True,
)

# In production, poll extract_result.job_id when async_=True, then read the
# completed extraction_id from the job result.
extraction_id = extract_result.extraction_id

split_result = client.split(
    extraction_id=extraction_id,
    split_config={
        "split_input": [
            {"name": "Financial Statement Schedules", "description": "Financial statement schedules, exhibits, and supporting tables"},
            {"name": "Signatures", "description": "Signature blocks, officers, titles, and signing dates"},
            {"name": "Certifications", "description": "Officer certifications and compliance attestations"},
        ]
    },
)

schema_result = client.schema(
    split_id=split_result.split_id,
    split_schema_config={
        "Financial Statement Schedules": {
            "schema": {
                "type": "object",
                "properties": {
                    "schedule_names": {"type": "array", "items": {"type": "string"}},
                    "exhibit_numbers": {"type": "array", "items": {"type": "string"}},
                },
            },
            "schema_prompt": "Extract schedule names and exhibit identifiers.",
        },
        "Signatures": {
            "schema": {
                "type": "object",
                "properties": {
                    "signatories": {
                        "type": "array",
                        "items": {
                            "type": "object",
                            "properties": {
                                "name": {"type": "string"},
                                "title": {"type": "string"},
                                "date": {"type": "string"},
                            },
                        },
                    },
                },
            }
        },
        "Certifications": {
            "schema": {
                "type": "object",
                "properties": {
                    "certifying_officers": {
                        "type": "array",
                        "items": {"type": "string"},
                        "description": "Officers who signed certifications",
                    }
                },
            }
        },
    },
)

print(schema_result.results)

​
Checks

  • Split topics should be mutually exclusive and easy to describe.
  • Inspect page assignments before trusting schema output.
  • Keep per-topic schemas smaller than a single all-purpose schema.
  • If the same topic appears across many documents, save it as a split preset.
  • For regulated review, store the split output and schema version IDs with the downstream record so reviewers can reproduce the exact extraction.

Extract -> Split -> Schema

Full Platform walkthrough.

Chaining Steps

Understand the extraction_id to split_id handoff.

Sample Documents

Use a long sample PDF to test topic splits.