@absolutejs/discoverv0.0.9alphaCommerce & GrowthB2B decision-maker discovery from free open-data adapters with LLM plus web search as the fallback.
Finds the right person to contact at a company — the decision-maker lookup PDL and Apollo charge per person-search for. Orchestration is provider-agnostic: it consults importable open-data DatasetSource adapters first (free), then falls back to an injected web search plus LLM extraction only when it needs more. Results are deduped and ranked by confidence, and a failing source, search, or LLM contributes nothing rather than throwing.
bun add @absolutejs/discoverFind the right person to contact at a company — the decision-maker PDL/Apollo charge per person-search for. Provider-agnostic orchestration: it consults importable dataset adapters first (free), then falls to LLM + web only when it needs more.
Licensed under BSL 1.1 (converts to Apache 2.0 on 2030-06-06). Use it to build your product; you may not host it as a competing prospecting/sales- intelligence SaaS. See LICENSE.
Open-data sources are importable adapters implementing one interface, so seeded public data is free and the paid LLM/web path is the fallback:
@absolutejs/dataset-gleif, -sec-edgar, -github, … each implement this over a public dataset and normalize to one shape. Adapters are pure importers of public data. The self-collected, self-healing dataset (learned email patterns, the cross-member graph, outcomes) is deliberately not an adapter — it stays private; it's the product.
discover stays provider-agnostic — you inject the capabilities:
With no deps it returns only what the adapters provide. With search + extract it discovers from the open web (the LLM is told to use only people who appear in the results — never to invent names). Results are deduped and ranked by confidence. Never throws — a failing source / search / LLM just contributes nothing.
discoverContacts takes a company plus a role intent and returns ranked contacts with title, confidence, LinkedIn URL, source, and a reason.
Open-data sources implement one DatasetSource interface (findPeople, findCompany) and normalize to one shape, so seeded public data stays free.
You inject search (Brave, SerpAPI, RAG) and extract (any LLM completion) via DiscoverDeps; the package never picks a provider.
The LLM is instructed to use only people who appear in the search results — never to invent names — and results are deduped and confidence-ranked.
discoverContacts never throws: a failing source, search, or LLM call simply contributes nothing to the result set.
A withCache wrapper is exported for memoizing expensive source and search calls.
Outcomes
Find the right person to contact at a company — the decision-maker PDL/Apollo charge per person-search for. Provider-agnostic orchestration: it consults importable dataset adapters first (free), then falls to LLM + web only when it needs more.
Open-data sources are importable adapters implementing one interface, so seeded public data is free and the paid LLM/web path is the fallback:
discover stays provider-agnostic — you inject the capabilities:
Hardening checklist
Follow in order
# @absolutejs/discover
import { discoverContacts } from "@absolutejs/discover";
await discoverContacts(
{ company: "Acme", domain: "acme.com", roleIntent: "head of partnerships" },
{ sources: [secEdgar, gleif], search: braveSearch, extract: askClaude },
);
// → [{ fullName: "Jane Doe", title: "Head of Partnerships", confidence: 90,
// linkedinUrl: "...", source: "...", reason: "listed as partnerships lead" }]Open-data sources are importable adapters implementing one interface, so seeded public data is free and the paid LLM/web path is the fallback:
type DatasetSource = {
name: string;
findPeople?: (q: DatasetQuery) => Promise<NormalizedPerson[]>;
findCompany?: (q: { name?; domain? }) => Promise<NormalizedCompany | null>;
};discover stays provider-agnostic — you inject the capabilities:
type DiscoverDeps = {
search?: (query: string) => Promise<WebSearchResult[]>; // Brave, SerpAPI, RAG…
extract?: (prompt: string) => Promise<string>; // any LLM completion
sources?: DatasetSource[]; // importable adapters
};Dataset adapters are consulted first for free; the injected search and extract capabilities only run when the adapters come up short.
import { discoverContacts } from '@absolutejs/discover';
const contacts = await discoverContacts(
{
company: 'Acme',
domain: 'acme.com',
roleIntent: 'head of partnerships'
},
{ extract: askClaude, search: braveSearch, sources: [secEdgar, gleif] }
);
// → [{ fullName: 'Jane Doe', title: 'Head of Partnerships',
// confidence: 90, linkedinUrl: '…', source: '…',
// reason: 'listed as partnerships lead' }]Every open-data adapter — GLEIF, SEC EDGAR, GitHub — implements this one interface over a public dataset and normalizes to one shape.
type DatasetSource = {
name: string;
findPeople?: (q: DatasetQuery) => Promise<NormalizedPerson[]>;
findCompany?: (q: {
name?: string;
domain?: string;
}) => Promise<NormalizedCompany | null>;
};Open-data sources are importable adapters implementing the DatasetSource interface, published from the dataset-adapters monorepo (Apache-2.0). Seeded public data is free; the paid LLM/web path is the fallback.
Search the declarations exported by the current package type files. Expand a symbol to inspect its source-backed signature.
These playbooks show where this package fits, how to verify the combined system, and what changes before production.
Current package surface
Import surface · click to copy