Verify claims against original sources by tracing the complete extraction chain.
Follow any claim, output, or extracted result back through the processing chain to its original source. Provenance tracking ensures that knowledge doesn't drift from truth as it moves through extraction, synthesis, and publication.
| Trace | Verify | Trust |
|---|---|---|
| Follow any output back through the extraction chain to its original source. | Compare extracted outputs with original context for accuracy. | Build a knowledge base where every claim is backed by inspectable evidence. |
An extracted summary says something specific — but was it accurate to the original source? Provenance tracing lets you follow any claim back to the exact message, document, or conversation it came from.
When content is extracted, summarised, reorganised, and published, each step can introduce subtle distortions. Provenance tracking makes drift visible by maintaining the full chain from source to output.
If you publish from your corpus, readers need to trust the sources. Provenance links give every published claim a verifiable origin — not just a bibliography, but an inspectable source chain.
Every extracted output carries a chain: Original Source → Extraction → Verification → Publication. Each link is navigable — click through to see the original context, the extraction parameters, and the reviewer assessment.
Provenance isn't just a link — it includes the surrounding context. When you trace a code block to its source conversation, you see the discussion that produced it, the alternatives that were considered, and the reasoning behind the final version.
When extracted outputs change over time — through editing, re-extraction, or manual modification — the system flags divergence from the original source. Drift is visible and correctable.
The primary surface for checking extracted outputs against source material, with evidence panels and validation workflows.
Explore Verification Workspace →
Graph provenance connects entities and relationships to their source documents and extraction context.