# Extract Document Chunks, Tables, and Assets

Produce semantic chunks, structured table exports, page Markdown, hierarchy, and embedded assets for downstream processing.

Reader extraction turns supported files into stable intermediate structures. That is useful for search indexes, retrieval systems, review sidecars, data migration, and document inventories.

## Chunk content

`Read-OfficeDocument -As Chunks` emits bounded text chunks with location and provenance. Format-specific options control Word notes, PowerPoint notes, Excel headers and row groups, Markdown heading chunks, hashes, and maximum content sizes. Set `-MaximumChunkCharacters` and `-MaximumTableRows` to the downstream limits.

## Materialize tables

`Read-OfficeDocument -As Tables` returns structured tables. Add `-Bundle` for CSV, Markdown, and JSON representations in memory. `Export-OfficeDocumentAsset -Type Table -OutputDirectory .\Tables` writes deterministic sidecars. Use `Read-OfficeDocument -As Assets` to inspect embedded assets and `Export-OfficeDocumentAsset -OutputDirectory .\Assets` to materialize them.

The [extract-tables recipe](https://github.com/EvotecIT/OfficeIMO/blob/3b680b1b6fb15a268ff5525a4ca6cdd70b2f6613/PowerShell/PSWriteOffice/Examples/Reader/Recipe-Reader-ExtractTables.ps1) creates a Markdown table, reads it as structured data, writes its sidecars, and returns semantic chunks.

Choose limits from the downstream contract. A search preview, an embedding pipeline, and a complete archival export need different chunk and table bounds.

<!-- PowerForge markdown source sha256:00773D3A0D227EA13B1FE583AACEFCD91AD1ED87CF2686D627DB54FD830E20C6; title=True -->

<!-- PowerForge markdown artifact sha256:50197AE3BD9EA7316C6A83BB03D769CBF747A2D49419711DE1DB4B4DA0369DA0 -->
