seed

To seed external sources, you convert raw inputs like URLs and local files into structured inbox items with provenance metadata, ensuring traceability before downstream processing. This handles input acquisition, format detection, and writes the item to an inbox directory.

Provenance metadata records the source URI, acquisition time, method, and content hash for inbox items. You need it to ensure complete traceability of raw inputs and to support deduplication, keeping your processing pipeline clean and verifiable.

Yes, you can batch process multiple inputs including pasted text and clipboard data across environments. Batch seeding creates structured inbox items with automatic metadata extraction while performing deduplication to avoid duplicate entries in your processing inbox.

Deduplication works by generating a content hash for each seeded item during provenance generation. This hash identifies duplicate entries from URLs, local files, or pasted text, preventing redundant inbox items and keeping downstream workflows clean.

The best way to extract title, author, and date is through automatic metadata extraction during the seeding process. When you seed sources into the inbox, the system detects formats and pulls this metadata automatically to create structured, traceable items.