nativ-vercel-ai

Nativ localization tools for the Vercel AI SDK. Give any AI agent the ability to translate, search translation memory, and manage localization settings — all using your team's brand voice and style guides.

Install

npm install nativ-vercel-ai ai zod

Quick start

import { generateText } from "ai";
import { openai } from "@ai-sdk/openai";
import { createNativTools } from "nativ-vercel-ai";

const nativTools = createNativTools(); // reads NATIV_API_KEY from env

const { text } = await generateText({
  model: openai("gpt-4o"),
  tools: nativTools,
  maxSteps: 5,
  prompt: 'Translate "Welcome back!" into French, German, and Japanese.',
});

console.log(text);

Configuration

// Pass options explicitly
const tools = createNativTools({
  apiKey: "nativ_...",
  baseUrl: "https://api.usenativ.com",  // default
  timeout: 120_000,                      // default: 2 minutes
});

// Or set environment variable
// NATIV_API_KEY=nativ_...
const tools = createNativTools();

Get your API key at dashboard.usenativ.com → Settings → API Keys.

Available tools

| Tool | Description | |------|-------------| | nativ_translate | Translate text with cultural adaptation, TM matching, and rationale | | nativ_translate_batch | Translate multiple texts to the same target language | | nativ_search_translation_memory | Search existing translations for consistency | | nativ_add_translation_memory | Store an approved translation for future reuse | | nativ_get_languages | List configured languages with formality settings | | nativ_get_style_guides | Get all style guides | | nativ_get_brand_voice | Get the brand voice prompt | | nativ_get_tm_stats | Translation memory statistics |

Use with streamText

import { streamText } from "ai";
import { openai } from "@ai-sdk/openai";
import { createNativTools } from "nativ-vercel-ai";

const result = streamText({
  model: openai("gpt-4o"),
  tools: createNativTools(),
  maxSteps: 5,
  prompt: "What languages do we support?",
});

for await (const chunk of result.textStream) {
  process.stdout.write(chunk);
}

Cherry-pick tools

const allTools = createNativTools();

// Only expose translate to the model
const { text } = await generateText({
  model: openai("gpt-4o"),
  tools: {
    nativ_translate: allTools.nativ_translate,
  },
  prompt: 'Say "Thank you" in Korean',
});

Next.js route handler

// app/api/chat/route.ts
import { streamText } from "ai";
import { openai } from "@ai-sdk/openai";
import { createNativTools } from "nativ-vercel-ai";

export async function POST(req: Request) {
  const { messages } = await req.json();

  const result = streamText({
    model: openai("gpt-4o"),
    tools: createNativTools(),
    maxSteps: 10,
    messages,
  });

  return result.toDataStreamResponse();
}

How it works

createNativTools() instantiates a nativ-sdk client and wraps each API method as a Vercel AI SDK tool() with typed Zod schemas. The tools are provider-agnostic — they work with OpenAI, Anthropic, Google, and any model the AI SDK supports.

Related packages

| Package | Description | |---------|-------------| | nativ-sdk | Node.js/TypeScript SDK | | nativ | Python SDK | | langchain-nativ | LangChain tools | | nativ-mcp | MCP server |

License

MIT — see LICENSE.