@vectrion/core

The central SDK client for the Vectrion AI runtime infrastructure. Provides the Vectrion class with middleware pipeline orchestration, provider abstraction, and Zod-based structured output validation.

📖 Part of the Vectrion SDK. View the live documentation and architectural specifications at vectrion.vercel.app.

Installation

npm install @vectrion/core @vectrion/types zod

Usage

import { Vectrion } from "@vectrion/core";
import { GoogleProviderAdapter } from "@vectrion/provider-google";
import { z } from "zod";

const ai = new Vectrion({
    providers: [
        new GoogleProviderAdapter({ apiKey: process.env.GOOGLE_API_KEY! }),
    ],
});

// Text generation
const result = await ai.generate({
    model: "gemini-2.0-flash",
    prompt: "Explain TypeScript in one sentence",
});
console.log(result.data);
console.log(result.response.usage); // { promptTokens, completionTokens, totalTokens }
console.log(result.response.cost); // { totalCostUsd }

// Structured output with Zod schema
const user = await ai.generate({
    model: "gemini-2.0-flash",
    prompt: 'Extract: "Jane is 30 and lives in NYC"',
    schema: z.object({ name: z.string(), age: z.number(), city: z.string() }),
});
console.log(user.data); // { name: "Jane", age: 30, city: "NYC" } — fully typed

// Middleware
ai.use(async (ctx, next) => {
    console.log(`Sending: ${ctx.request.prompt}`);
    await next();
    console.log(
        `Provider: ${ctx.response?.provider}, Latency: ${ctx.response?.latencyMs}ms`,
    );
});

API

new Vectrion(config)

• config.providers — Array of ProviderAdapter instances
• config.router — Optional RouterEngine (default: sequential fallback)

ai.generate(params)

• params.model — Model identifier string
• params.prompt — Input prompt
• params.schema — Optional Zod schema for structured output
• Returns { data, response } where data is typed to the schema

ai.use(middleware)

• Adds middleware to the onion pipeline (pre/post hooks around provider execution)

License

MIT