Source-Available Tool

Local-First PDF Structure Extraction for AI Pipelines

Sourcetrace PDF is powered by pdfstruct, a source-available PDF extraction layer for converting born-digital PDFs into traceable layout-aware JSON and Markdown.

It is designed for private AI pipelines where page evidence, text position, reading order, tables, annotations, metadata, and diagnostics matter before retrieval or model reasoning.

Discuss Commercial Use View Document Structure Review

What It Preserves

Traceable Layout-Aware JSON and Markdown

Page boundaries

Text runs

Glyph positions where enabled

Lines and paragraph candidates

Reading-order candidates

Table candidates with confidence and evidence

Annotations

Images and image metadata

Form fields where supported

Metadata

Diagnostics

JSON and Markdown export

What It Does Not Pretend

PDF Structure Is Inferred

PDF is not a semantic document format.

A PDF page is often a set of positioned drawing operations. Tables, headings, columns, and reading order may not exist explicitly inside the file.

Sourcetrace PDF exposes confidence, evidence, and diagnostics rather than pretending layout recovery is certain.

Fit

Best Fit

Legal document bundles
Financial statements
Contracts
Policy documents
Board packs
Born-digital PDFs
Private RAG pipelines
Local-first AI ingestion
Document auditability workflows

Not Best Fit

Scanned documents where OCR is the primary problem
Image-only PDFs
Generic cheap document conversion at cents-per-page pricing
Use cases where cloud extraction is already acceptable and sufficient
Workflows that do not need traceability or layout evidence

Licence

Commercial Source-Available Licence

Sourcetrace PDF is intended to be source-available.

Free permitted use

Local evaluation
Personal use
Academic research
Non-production internal testing
Proof-of-concept assessment

Commercial licence required

Production use
Client work
Regulated workflows
SaaS or hosted API use
Redistribution
Embedding in another commercial product
Processing documents for third parties
Usage above the free evaluation threshold

Recommended licence model: Business Source License 1.1 with later conversion to Apache-2.0 after the defined change date.

Deployment Model

Local First, Not Hosted First

Sourcetrace PDF is intended to run where the documents already live.

• Python library
• Command-line tool
• Docker container
• Client-hosted API wrapper
• Integration inside private AI ingestion pipelines

A hosted API is not the default model because sensitive document workflows often require local control, private processing, and clear custody of source material.

Commercial Use Through Lumen & Lever

Commercial Use Is Handled Through Lumen & Lever

Available commercial paths include professional licence, team licence, embedded/OEM licence, Document Structure Review, custom extractor pack, and Structural AI Architecture Sprint where the issue extends beyond extraction.

Discuss Commercial Use