Data Source

Stores what feeds it in the resolved platform or authorized MongoDB store: raw pages, Markdown pages and LLM results, with the saved document ids and counts.

Option	Meaning
save_images	Also keep page images in your storage: up to 20 per page and 200 per run, each up to 10 MB, public hosts only. Pages reference them in blobs.
Where to keep the results	Your database by default, or one of your MongoDB resources

Tip

Use two Data Sources: one fed by the LLM Processor (the records) and one fed by the URL Fetcher with save_images on (the pages and their images).

Never put keys or connection strings in a node. Stores are chosen by resource id.

Input, output, and execution

Connect a branch providing raw_pages, pages, or llm_result. The node writes those artifacts to the corresponding collections and returns database_result with saved ids, counts, and execution/node provenance. It does not require a dummy LLM stage for raw-only storage.

Keep artifacts without images

{"save_images":false}

save_images defaults to false. A branch with no persistable artifacts normally fails with dataSource requires source artifacts to persist. Storage connection failures can fail the node; inspect its output and run error before treating data as saved. Use MongoDB Store for keyed structured records and Query to read saved artifacts. Connection objects in the engine schema are injected internally.