Replicate TypeScript Library

The Replicate TypeScript library provides convenient access to the Replicate API from TypeScript or JavaScript.

Documentation

API reference documentation is available here.

Installation

npm install replicate-fern

Usage

The library needs to be configured with your account's API token, which is available in the Replicate Dashboard.

import { ReplicateClient } from "replicate";

const client = new ReplicateClient({
    token: "YOUR_TOKEN",
});

const prediction = await client.predictions
    .create({
        version: "da77bc59ee60423279fd632efb4795ab731d9e3ca9705ef3341091fb989b7eaf",
        input: {
            prompt: "A photo of an astronaut riding a horse",
        },
    })
    .wait();

console.log(prediction);

Polling Predictions

When creating predictions, you can use the .wait() method to automatically poll until the prediction is complete. You can customize the polling behavior:

const prediction = await client.predictions
    .create({
        version: "da77bc59ee60423279fd632efb4795ab731d9e3ca9705ef3341091fb989b7eaf",
        input: {
            prompt: "A photo of an astronaut riding a horse",
        },
    })
    .wait({
        polling: {
            intervalMs: 1_000, // Poll every second (default: 3000ms)
            timeoutMs: 30_0000, // Timeout after 5 minutes (default: no timeout)
        },
        // other common request options
    });

Request And Response Types

The SDK exports all request and response types as TypeScript interfaces. Simply import them with the following namespace:

import { Replicate } from "replicate";

const request: Replicate.GetCollectionsRequest = {
    ...
};

Error Handling

When the API returns a non-success status code (4xx or 5xx response), a subclass of ReplicateError will be thrown:

import { ReplicateError } from "replicate";

try {
    await client.deployments.create({
        // ...deployment options
    });
} catch (err) {
    if (err instanceof ReplicateError) {
        console.log(err.statusCode);
        console.log(err.message);
        console.log(err.body);
        console.log(err.rawResponse);
    }
}

Advanced Usage

Additional Headers

You can send additional headers with your requests:

const response = await client.deployments.create(
    {
        // ...deployment options
    },
    {
        headers: {
            "X-Custom-Header": "custom value",
        },
    },
);

Retries

The SDK is instrumented with automatic retries with exponential backoff. A request will be retried as long as the request is deemed retryable and the number of retry attempts has not grown larger than the configured retry limit (default: 2).

A request is deemed retryable when any of the following HTTP status codes is returned:

• 408 (Timeout)
• 429 (Too Many Requests)
• 5XX (Internal Server Errors)

Use the maxRetries request option to configure this behavior.

const response = await client.deployments.create(..., {
    maxRetries: 0 // override maxRetries at the request level
});

Timeouts

The SDK defaults to a 60 second timeout. Use the timeoutInSeconds option to configure this behavior.

const response = await client.deployments.create(..., {
    timeoutInSeconds: 30 // override timeout to 30s
});

Aborting Requests

The SDK allows users to abort requests at any point by passing in an abort signal.

const controller = new AbortController();
const response = await client.deployments.create(..., {
    abortSignal: controller.signal
});
controller.abort(); // aborts the request

Access Raw Response Data

The SDK provides access to raw response data, including headers, through the .withRawResponse() method. The .withRawResponse() method returns a promise that results to an object with a data and a rawResponse property.

const { data, rawResponse } = await client.deployments.create(...).withRawResponse();

console.log(data);
console.log(rawResponse.headers['X-My-Header']);

Runtime Compatibility

The SDK is compatible with the following runtimes:

• Node.js 18+
• Vercel
• Cloudflare Workers
• Deno v1.25+
• Bun 1.0+
• React Native

Customizing the Fetch Client

The SDK provides flexibility in customizing the underlying HTTP client. You can provide your own implementation:

import { ReplicateClient } from "replicate";

const client = new ReplicateClient({
  token: "YOUR_TOKEN",
  fetcher: // provide your implementation here
});

Custom Fetch Requests

The SDK provides a low-level fetch method for making custom HTTP requests to the Replicate API:

const response = await client.fetch(
    "/v1/models/meta/llama-2-70b-chat",
    {
        method: "GET",
        cache: "no-cache",
        credentials: "same-origin",
    },
    {
        timeoutInSeconds: 30, // Timeout after 30 seconds
        maxRetries: 3, // Retry up to 3 times
        headers: {
            // Add custom headers
            "Custom-Header": "value",
        },
        abortSignal: signal, // Optional abort signal
    },
);

const data = await response.json();

The fetch method mimics native fetch but also automatically handles:

• Authentication
• Base URL resolution
• Retries with exponential backoff
• Request timeouts
• Custom headers

Contributing

While we value open-source contributions to this SDK, this library is generated programmatically. Additions made directly to this library would have to be moved over to our generation code, otherwise they would be overwritten upon the next generated release. Feel free to open a PR as a proof of concept, but know that we will not be able to merge it as-is. We suggest opening an issue first to discuss with us!

On the other hand, contributions to the README are always very welcome!