@absolutejs/audience

In-house audience & affinity intelligence for AbsoluteJS apps — the ownable parts of what an audience-intelligence SaaS sells, run on your own provider and embedding model.

• inferPsychographics — communication style, values, and motivations of a person or brand from public signals (the "how to approach them" layer).
• profileAffinity — a structured interest/brand affinity profile, embedded into a vector.
• affinityOverlap — a measured audience-overlap score between two profiles (cosine of their embedded affinities) plus shared topics/brands. Pure, no model call.
• cosineSimilarity — the underlying vector math, exported.

Like @absolutejs/partnership, every primitive is a pure function of typed input plus an injected AI call. Affinity work also takes an injected embed call, so the package never picks a provider or an embedding model — your app supplies both, with its own metering and caching.

Install

bun add @absolutejs/audience

Wiring

import type { AudienceContext, GenerateObject } from "@absolutejs/audience";
import { meteredGenerateObjectAI } from "./usage/meteredAI";
import { aiProvider } from "./integrations/aiProvider";
import { embedTexts } from "./integrations/ragStore";

export const audienceCtx = (userSub?: string | null): AudienceContext => ({
  embed: (texts) => embedTexts(texts, "passage"),
  generateObject: ((req) =>
    meteredGenerateObjectAI({ ...req, provider: aiProvider, userSub })) as GenerateObject,
});

Measuring audience overlap

import { profileAffinity, affinityOverlap } from "@absolutejs/audience";

const me = await profileAffinity({ name: "Me", signals: { niche, offer } }, audienceCtx());
const them = await profileAffinity({ name: company, signals: { summary, industry } }, audienceCtx());

const { score, sharedTopics, rationale } = affinityOverlap(me, them);
// score ∈ [0,1] — a *measured* overlap to use wherever you'd otherwise
// have an LLM guess (e.g. a Trust & Fit "audience overlap" dimension).

For a cheap inline path (no affinity extraction), embed two short descriptor strings yourself and call cosineSimilarity directly.

License

BSL 1.1 → Apache 2.0 on the Change Date (see LICENSE). Build your own apps and SaaS on top of it; you may not offer it as a competing hosted audience-intelligence service.