Ingest a document
Parsed → chunked → embedded → graph-extracted with provenance.
Tip: extraction runs one model call per chunk — a whole book (~46 chunks) takes several minutes. You'll see live progress here.
Documents0
Select one to inspect or query.
Generate from a template
Upload a template; each field
[LIKE_THIS] and section {{ … }} is filled from the selected source document — grounded, not invented.Source:none — pick one in the Documents panel →