kritmatta-lead-scorer

Single source of truth for lead qualification scoring. Consumed by both the Vega app (hire-chat-next-app) and the enrichment worker (enrichment-worker) so the scoring logic can never drift between them.

What's in it

• scorer.js - weight-profile scorer with per-ICP signal_overrides (the app UI and the worker score identically). Hiring intent is read from either the website scrape (tier2.hiring_signals.hiring_velocity) or a native LinkedIn-jobs signal (enrichment_data.hiring_signal, from the app's hiringSignalService) - so a company hiring a relevant role scores even before any website enrichment.
• decay.js - intent-signal time decay.
• reasoningRewriter.js - optional LLM rephrasing of the deterministic reasoning. The AI call is dependency-injected so this package has no app/worker-specific requires.

Install

Public package on npmjs.com - no auth or .npmrc needed:

npm install kritmatta-lead-scorer

Usage

App (deterministic, no AI):

const { scoreLeadFromEnrichment } = require('kritmatta-lead-scorer');
const { score } = scoreLeadFromEnrichment(lead, icpCriteria);

Worker (with LLM-rewritten reasoning - inject callAI once at startup):

const scorer = require('kritmatta-lead-scorer');
const { callAI } = require('./utils/ai');
scorer.setCallAI(callAI);            // enable reasoning rewriting
await scorer.scoreLeadsMultiIcp(leads, job, supabase);

If setCallAI is never called, scoreLeadsMultiIcp still works and just returns the deterministic reasoning (that's the app's mode).

Publishing (maintainers)

Public npm. Bump version in package.json, then:

npm login          # once, to npmjs.com
npm publish        # public by default (unscoped name)