discover()

First contact with a row in your CSV. Resolve the company. Find the canonical site. Confirm it is alive.

Inputs and outputs

In. Lead row from CSV (any of: company name, domain, LinkedIn URL, email).

Out. Canonical company record: legal name, primary domain, country, alive flag, redirect chain.

Current version

v1.4.2 Forkable through the prompt library.

Model defaults

Default: Groq Llama-3.1-70B for name resolution. Anthropic Haiku for ambiguous cases. You can override per stage in your routing config.

How it fails

Most common: holding-company names that resolve to a parent group with no real product. We tag those parent_resolution_ambiguous and skip the row by default. You can override per-pack.

Evals

Resolution accuracy on a labelled set of 2,400 messy rows: 96.1% at v1.4.2, up from 91.7% at v1.3.0. The full eval set ships with the prompt and is editable. Eval format.