Exporting Your Data

Your data is yours. Thought Cove can export a Cove’s contents as a Portable Knowledge Package — an open, vendor-neutral format any downstream tool can load.

What’s in a package

An export is two files: an Apache Parquet file containing your structured, embedded chunks, and a manifest (manifest.json) describing the package.

• The Parquet file holds each chunk’s text, its 768-dimension embedding, page number, section header, and metadata.
• The manifest records the schema version, the embedding model used, a content hash for integrity, and the row count.

Because it’s Parquet plus a manifest, the package loads directly into data tools, notebooks, and downstream AI systems without any proprietary reader.

Export a Data Pond

Choose the domain and version you want to package and export it; you’ll receive a single .zip containing the Parquet file and its manifest.

Use it downstream

Point your own pipeline at the Parquet file to run search, analysis, or your own model over the same embedded chunks. The content hash in the manifest lets you verify the package hasn’t changed since export.