Name the document and its version
A file name may be useful to a person and unclear to an ingestion pipeline. Add a title, category and version when the source provides one.
Do not invent a version to fill a field. A missing version is useful information: it tells the next reviewer what still needs checking against the source.
Keep references beside the content
When a passage is edited or deduplicated, its source should remain traceable. Precidoc exports page or section references with chunks, including multiple references when identical chunks are merged.
The built-in Markdown sample uses section references. A digital PDF can use page references. Preserve whichever your document actually provides.
Document → reviewed text → chunk Chunk → source section or page Merged chunk → all retained source references
Choose a format for the next step
JSON is useful when another tool will read structured document, chunk and preparation fields. Markdown is useful for a person reviewing the text and report.
Precidoc exports a preparation package, not a connection to your vector database. Check the destination schema and adapt the fields before ingestion. Both formats should retain the material you need for a later review.
Review, then download
Check the source text, table values and metadata before approving the result. An edit or a change to chunk settings requires another review.
The current document session lives in browser memory. Download the export before refreshing or closing the tab. Keep it with the source document so the next person can reproduce your checks.