SectorsData Processing

Data Processing

Moving data between systems, normalizing it, and keeping it queryable is repetitive, rule-based work, whether that's a one-off backfill or a job that runs on a schedule. A flow's database and browser nodes are built to automate exactly that.

Where it fits

  • A flow can read or scrape from a source, transform it, and write it to a database in one run, instead of a person copying data between systems by hand.
  • Table, Elasticsearch, and RAG databases cover three different shapes of data (typed rows, full-text search, and semantic retrieval) without standing up a separate external database for each.
  • A flow can be triggered on a schedule instead of on demand, so a recurring extraction or sync job runs unattended.

What this looks like

  • A BROWSER node that keeps a single Chromium session alive across a flow run, so it can log in once and then navigate and scrape across several steps, with SSRF-safe URL handling built in.
  • A DATABASE_WRITE node normalizing scraped or ingested data into a Table database, or a DATABASE node indexing it into an Elasticsearch cluster or a RAG database for later retrieval.
  • A UTILITY node for a sandboxed custom transform step, like reformatting, deduping, or validating, that the built-in node types don't cover, running in a locked-down, single-use container.

Frequently Asked Questions