zenpreference

A semantic preference store for AI applications. Store key-value preferences with natural language content and retrieve them by semantic similarity.

Features

• Semantic retrieval — Find preferences by meaning, not just exact keys
• Key routing — Known keys (e.g. theme, packageManager) get exact matches first
• SQLite storage — Persistent storage with optional lazy loading
• AI SDK embeddings — Uses any AI SDK–compatible embedding model

Installation

bun add zenpreference @openrouter/ai-sdk-provider

Usage

import { createOpenRouter } from "@openrouter/ai-sdk-provider";
import { PreferenceStore } from "zenpreference";

const openrouter = createOpenRouter({ apiKey: process.env.OPENROUTER_API_KEY! });
const embeddingModel = openrouter.textEmbeddingModel("openai/text-embedding-3-small");

const store = new PreferenceStore({
  embeddingModel,
  filePath: "./.preferences.db",
  knownKeys: ["theme", "packageManager", "language", "framework"],
});

await store.add({
  key: "theme",
  value: "dark",
  content: "User prefers dark mode for all applications",
});

const prefs = await store.getRelevant("user's color scheme preference");

API

• PreferenceStore — Main store class
• add(preference) — Store a preference
• getByKey(key) — Get by exact key
• getRelevant(query) — Semantic search over preferences

License

MIT