Connect. Normalize. Preserve. Search.

FEED is a source-adapter framework for public information. This site runs a live nutrition-research collect. The same architecture runs other authorized public-data missions. Open FEED.

1. 01 · Connect

   One adapter interface. Official APIs, open datasets, a robots-respecting crawler, a firehose, licensed feeds. Each adapter carries its terms, rate limit, and cost.

2. 02 · Normalize

   One record. Source, platform, credit, time, text, media, engagement. Facets stay attached to the source observation.

3. 03 · Deduplicate

   URL key. Media hash. Text hash and simhash. A row seen again updates last-seen and adds an engagement snapshot. Nothing is overwritten.

4. 04 · Quality

   Credit. URL. Spam. Topic gate. Source mix. The dataset passes or it holds. Every run is logged with a recipe hash so it can be rerun.

5. 05 · Preserve

   Searchable storage with daily rollups. Public read of the live data. Scheduled pulls keep it current.

6. 06 · Export

   A keyed export for a pipeline. Request a FEED key. We review each request.

A source is not a scraper. Instagram, Facebook, and X enter through their official APIs or a licensed feed, priced on the row. Public pages enter through a crawler that honours robots.txt and never logs in. Open research datasets show what platform rows look like before a firehose is bought.