@anyr/ai-sdk-provider

The AnyRouter provider for the Vercel AI SDK.

Call every model on the AnyRouter gateway through the standard AI SDK interface — generateText, streamText, generateObject, embedMany, generateImage — with one key and one provider.

Install

npm install @anyr/ai-sdk-provider ai zod
# or
pnpm add @anyr/ai-sdk-provider ai zod

Quickstart

import { createAnyRouter } from "@anyr/ai-sdk-provider"
import { generateText } from "ai"

// Reads ANYROUTER_API_KEY from the environment when omitted.
const anyrouter = createAnyRouter()

const { text } = await generateText({
  model: anyrouter("openai/gpt-4o-mini"),
  prompt: "What is AnyRouter?",
})
console.log(text)

Usage

generateText

import { createAnyRouter } from "@anyr/ai-sdk-provider"
import { generateText } from "ai"

const anyrouter = createAnyRouter({ apiKey: process.env.ANYROUTER_API_KEY })

const { text } = await generateText({
  model: anyrouter("anthropic/claude-opus-4-5"),
  system: "You are a helpful assistant.",
  prompt: "Explain the butterfly effect in one paragraph.",
})

streamText

import { createAnyRouter } from "@anyr/ai-sdk-provider"
import { streamText } from "ai"

const anyrouter = createAnyRouter()

const { textStream } = await streamText({
  model: anyrouter("openai/gpt-4o"),
  prompt: "Write a haiku about the ocean.",
})

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

Reasoning models

Reasoning models (e.g. deepseek/deepseek-r1) return their thinking as a separate stream — the provider surfaces it as the AI SDK's standard reasoning parts, so reasoning/reasoningText and the full-stream reasoning-* parts work out of the box.

import { createAnyRouter } from "@anyr/ai-sdk-provider"
import { generateText } from "ai"

const anyrouter = createAnyRouter()

const { text, reasoningText } = await generateText({
  model: anyrouter("deepseek/deepseek-r1"),
  prompt: "How many r's are in strawberry?",
})

console.log(reasoningText) // the model's chain of thought
console.log(text) // the final answer

generateObject (structured output)

import { createAnyRouter } from "@anyr/ai-sdk-provider"
import { generateObject } from "ai"
import { z } from "zod"

const anyrouter = createAnyRouter()

const { object } = await generateObject({
  model: anyrouter("openai/gpt-4o-mini"),
  schema: z.object({
    title: z.string(),
    summary: z.string(),
    tags: z.array(z.string()),
  }),
  prompt: "Summarise the latest AI news in a structured format.",
})

console.log(object.title, object.tags)

Tool calling

import { createAnyRouter } from "@anyr/ai-sdk-provider"
import { generateText, tool } from "ai"
import { z } from "zod"

const anyrouter = createAnyRouter()

const { text, toolCalls } = await generateText({
  model: anyrouter("openai/gpt-4o"),
  tools: {
    get_weather: tool({
      description: "Get the current weather in a city.",
      parameters: z.object({ city: z.string() }),
      execute: async ({ city }) => ({ temperature: 22, unit: "C", city }),
    }),
  },
  prompt: "What is the weather in Tokyo?",
})

Embeddings

import { createAnyRouter } from "@anyr/ai-sdk-provider"
import { embedMany } from "ai"

const anyrouter = createAnyRouter()

const { embeddings } = await embedMany({
  model: anyrouter.embeddingModel("openai/text-embedding-3-small"),
  values: ["Hello world", "How are you?"],
})

console.log(embeddings[0]) // [0.123, -0.456, ...]

Image generation

import { createAnyRouter } from "@anyr/ai-sdk-provider"
import { generateImage } from "ai"

const anyrouter = createAnyRouter()

const { image } = await generateImage({
  model: anyrouter.imageModel("openai/dall-e-3"),
  prompt: "A serene mountain lake at dawn, oil painting.",
  size: "1024x1024",
})

// image.base64 — base64-encoded PNG or URL depending on the model

Provider options (routing, session, extra body)

import { createAnyRouter } from "@anyr/ai-sdk-provider"
import { generateText } from "ai"

// Set routing preferences at the provider level
const anyrouter = createAnyRouter({
  provider: { only: ["openai"], quantizations: ["bf16"] },
  sessionId: "user-session-abc",
})

// Or override them per-call
const { text } = await generateText({
  model: anyrouter("openai/gpt-4o"),
  prompt: "Hello!",
  experimental_providerMetadata: {
    anyrouter: {
      provider: { order: ["openai", "azure"] },
      sessionId: "per-request-session",
      extraBody: { logprobs: true },
    },
  },
})

Configuration

import { createAnyRouter } from "@anyr/ai-sdk-provider"

const anyrouter = createAnyRouter({
  // API key — falls back to ANYROUTER_API_KEY env var (resolved lazily at request time)
  apiKey: "<your-anyrouter-key>",

  // API base URL — defaults to https://anyrouter.dev/api/v1
  baseURL: "https://anyrouter.dev/api/v1",

  // Extra headers merged into every request
  headers: { "X-Custom-Header": "value" },

  // Custom fetch implementation (e.g. for logging or proxying)
  fetch: myCustomFetch,

  // Default routing preference forwarded as `provider` in the request body
  provider: { only: ["openai"] },

  // Session ID forwarded as `session_id` — useful for grouping in analytics
  sessionId: "my-session",

  // Extra fields merged into every request body (per-call overrides these)
  extraBody: { safe_mode: true },
})

Provider methods

| Method | Returns | Description | |--------|---------|-------------| | anyrouter(modelId) | LanguageModelV3 | Shorthand for anyrouter.chat(modelId) | | anyrouter.chat(modelId) | LanguageModelV3 | Chat / instruction language model | | anyrouter.languageModel(modelId) | LanguageModelV3 | Alias for .chat() (ProviderV3 interface) | | anyrouter.embeddingModel(modelId) | EmbeddingModelV3 | Text embedding model | | anyrouter.imageModel(modelId) | ImageModelV3 | Image generation model | | anyrouter.client | AnyRouter | Raw @anyr/sdk client (escape hatch) |

Default export

import { anyrouter } from "@anyr/ai-sdk-provider"

// Uses ANYROUTER_API_KEY from the environment
const { text } = await generateText({
  model: anyrouter("openai/gpt-4o-mini"),
  prompt: "Hello!",
})

Built on @anyr/sdk.

License

MIT © Duyet Le