SourcePath logo
SourcePath LabsEvidence-first AI for structured authority documents

Structured documents, preserved as evidence-bearing graphs

This walkthrough shows how SourcePath ingests real Federal XML, preserves hierarchy, extracts cross-references, and turns missing targets into an explicit ingestion backlog. It is an example of the ingestion pipeline, not the product home page itself.

Start Here: Tiny Orientation Demo

This is the smallest useful example: one short document, a couple of sections, and unresolved references that become a backlog. It is the quickest way to understand what the larger outputs are doing.

Real Federal Ingestion Outputs

These artifacts come from the real GovInfo XML files. Each document has guided focused views for reliable rendering, plus larger full-volume maps for scale. The browser maps omit paragraph nodes for readability, while the preview CSVs preserve detailed row-level structure.

Real-stack backlog

Outstanding references discovered while ingesting the two Federal volumes.

Why this matters

The point of this example is not generic text search. It is to show that the system can preserve authority-bearing structure, prefer traceable evidence over semantic resemblance, and keep the next ingestion steps visible instead of implicit. The document-map pages prefer pre-rendered local SVGs, so the walkthrough does not depend on a Mermaid CDN.