@idriszade/process-enrich-apify

Pipeline-kit Process<RawContact, EnrichedContact> adapter for per-contact enrichment via Apify actors. Defaults target pipelinelabs/lead-scraper-apollo-zoominfo-lusha-ppe; the row mapper and input builder are pluggable for other lead-search actors.

Install

pnpm add @idriszade/process-enrich-apify @idriszade/core apify-client zod

Usage

import { createEnrichApifyProcess } from '@idriszade/process-enrich-apify';

const enrich = createEnrichApifyProcess({
  actorId: 'pipelinelabs/lead-scraper-apollo-zoominfo-lusha-ppe',
  defaultInput: { seniorityIncludes: ['c_suite', 'vp', 'director'] },
});

// Inside a kit pipeline, use via Pipeline.from(...).through(enrich)...
// Or call directly:
const result = await enrich.run(
  { email: '[email protected]', title: 'VP Sales' },
  ctx,
);
if (result.error === null) console.log(result.data.companyName);

Token resolution

Reads APIFY_TOKEN first, falls back to APIFY_API_TOKEN (pursuit's name) — pass apifyToken in config to override either env path.

Cost

PPE actor at $0.001/lead with totalResults: 1 per enrichment ≈ $0.001/contact. Within kit's $0.10/run cap by 100x.

Schema discipline

Both RawContactSchema and EnrichedContactSchema use mode='before'-equivalent .catch(undefined) defaults per feedback_pydantic_boundary_coercion.md — boundary validators must coerce, never reject batches.

Supported actor row shapes

The default row mapper targets two known schemas:

1. Nested company.{...} (Apollo / pipelinelabs PPE actor) — top-level email, fullName, title, seniority, plus company: { name, domain, industry, employees, location: { country } }.
2. Flat snake_case (legacy ZoomInfo / Lusha-derived actors) — person_email, full_name / name, job_title, company_name, company_domain, etc.

If you point this adapter at an actor with a different output schema, supply a custom rowMapper:

const enrich = createEnrichApifyProcess({
  actorId: 'your-org/custom-actor',
  rowMapper: (row) => ({
    email: row.contactEmail,
    companyName: row.organizationName,
    // ...etc
  }),
});

When the default mapper resolves NO top-level identity fields (email, fullName, companyName, companyDomain), it logs a warning to console.warn with the unrecognized row keys. Downstream Zod validation then handles the row gracefully — actor schema drift never throws, only warns.

MCP server

Ships a stdio MCP server alongside the kit primitive:

npx -y @idriszade/process-enrich-apify

Exposes a single tool enrich_contact_apify. Wire into Claude Desktop / Cursor / Smithery via mcp.json.

License

MIT.