​
Overview

While most LLM providers follow a similar chat-like structure, there are subtle differences in how prompts are formatted and processed. PromptL addresses these differences by providing Adapters for each major provider, ensuring that your prompts are correctly formatted and seamlessly integrated.

Currently, PromptL supports:

More providers will be supported in the future, and you can even create your own custom adapters for unsupported platforms.

​
Why Adapters?

Adapters handle provider-specific definition differences, such as:

  • Message structure: OpenAI uses role-based messages, while Anthropic uses a user/assistant prefix.
  • API integration: Adapters ensure compatibility with the provider’s API.

By using an adapter, you don’t need to worry about these differences—PromptL handles them for you.

​
Getting Started with Adapters

Here’s how to use an adapter in your project. For this example, we’ll use OpenAI:

import { render, Adapter } from 'promptl-ai';

const { messages, config } = await render({
  prompt,
  adapter: Adapters.openai, // Specify the adapter
});

​
Supported Providers

​
OpenAI (default)

The OpenAI Adapter, which is selected by default, formats prompts to match OpenAI’s chat-completion API, including support for models like gpt-4 and gpt-3.5.

​
Anthropic

The Anthropic Adapter ensures compatibility with Anthropic’s API.

Additional providers will be supported in the future. Check back for updates!

​
Extending Adapters

If you’re working with an unsupported provider, you can create your own adapter. Adapters are simple functions that transform PromptL’s messages and config into the format required by your provider.

An adapter is defined as an object with two functions: { fromPromptl, toPromptl }. Each function takes an object with messages and config properties and returns the same object with transformed data.

To see the structure of messages used in PromptL, check out the GitHub PromptL Repository

​
Example: Custom Adapter

const CustomAdapter = {
  fromPromptl: ({ messages, config }) => {
    // Transform PromptL messages to your provider's format
    const formattedMessages = messages.map((msg) => ({
      role: msg.type,
      content: msg.text,
    }));

    return { messages: formattedMessages, config };
  },

  toPromptl: ({ messages, config }) => {
    // Transform your provider's messages to PromptL format
    const formattedMessages = messages.map((msg) => ({
      type: msg.role,
      text: msg.content,
    }));

    return { messages: formattedMessages, config };
  },
};

Pass your custom adapter to the render function:

const { messages, config } = await render({
  prompt,
  adapter: CustomAdapter,
});

​
Contribute or Request Support

We’re constantly working to support more providers. If you’d like to request a specific provider or contribute an adapter, check out our GitHub repository.

​
Summary

Adapters make it easy to use PromptL with different LLM providers by handling provider-specific formatting and configuration. Whether you’re using OpenAI, Anthropic, or another platform, PromptL ensures seamless integration. Get started with a supported adapter or build your own for maximum flexibility.

Quick StartOpenAI