✨ Is your feature request related to a problem?
When using Marker to extract content from PDFs, the JSON output contains rich structured data — including tables, section headers, paragraphs, and metadata. However, there is currently no built-in way to persist this data into a database or export it in a tabular format for downstream use (analytics, reporting, pipelines, etc.).
Users who want to use Marker as part of a data pipeline are forced to write their own parsing logic on top of the JSON output every time.
Describe the solution you'd like
A new ExportConverter (or similar) that takes Marker's existing JSON output and allows exporting to:
SQLite — one table per document section type (e.g. tables, text_blocks, headers), zero external dependencies
CSV — one file per extracted table, named by page and position
Optionally: a simple HTML or Markdown summary report of the document structure
The converter would follow the existing converter pattern in the codebase and would be fully local, no API keys or cloud services required.
🧩 Additional Context
This would make Marker more useful as a drop-in component in document processing pipelines, without requiring users to build their own export layer. It also opens the door to batch processing workflows where multiple PDFs are processed and their content aggregated into a single database.
Happy to implement this if the maintainers are open to it. Would love feedback on the preferred approach before starting.
✨ Is your feature request related to a problem?
When using Marker to extract content from PDFs, the JSON output contains rich structured data — including tables, section headers, paragraphs, and metadata. However, there is currently no built-in way to persist this data into a database or export it in a tabular format for downstream use (analytics, reporting, pipelines, etc.).
Users who want to use Marker as part of a data pipeline are forced to write their own parsing logic on top of the JSON output every time.
Describe the solution you'd like
A new ExportConverter (or similar) that takes Marker's existing JSON output and allows exporting to:
SQLite — one table per document section type (e.g. tables, text_blocks, headers), zero external dependencies
CSV — one file per extracted table, named by page and position
Optionally: a simple HTML or Markdown summary report of the document structure
The converter would follow the existing converter pattern in the codebase and would be fully local, no API keys or cloud services required.
🧩 Additional Context
This would make Marker more useful as a drop-in component in document processing pipelines, without requiring users to build their own export layer. It also opens the door to batch processing workflows where multiple PDFs are processed and their content aggregated into a single database.
Happy to implement this if the maintainers are open to it. Would love feedback on the preferred approach before starting.