How to Use Claude AI API in JavaScript?

Claude AI is an AI assistant created by Anthropic to be helpful, harmless, and honest. It offers a powerful conversational API that developers can leverage to build a wide range of AI-powered applications.

In this comprehensive guide, we will explore how to use the Claude AI API with JavaScript to integrate Claude’s advanced natural language processing capabilities into your Node.js and front-end web applications.

Prerequisites

• Node.js installed on your development machine
• Basic knowledge of JavaScript programming
• Claude AI API key (register on Anthropic website to obtain)

Install Claude AI NPM Package

We will use the official claude-ai NPM package to integrate with Claude API in our JavaScript code.

Run the below command to install it:

npm install claude-ai

This will download and install the package into the node_modules directory to be used in our app.

Import Claude AI Package

In your JavaScript code, import the installed Claude AI package with:

const { Configuration, OpenAIApi } = require("claude-ai");

This will import the Configuration and OpenAIApi classes that we need to initialize connection with Claude API.

Initialize Configuration

Before we can make requests, we need to initialize a Configuration object with our API key:

const configuration = new Configuration({
  apiKey: YOUR_CLAUDE_API_KEY,
});

Replace YOUR_CLAUDE_API_KEY with the actual API key you copied from your Claude account.

This configuration will be used to authenticate and connect to Claude API.

Create OpenAIApi Instance

With the configuration initialized, we can now instantiate the OpenAIApi class:

const openai = new OpenAIApi(configuration);

The openai object provides methods to call different Claude API endpoints.

Call the Completions Endpoint

The most commonly used Claude endpoint for generating text is createCompletion.

Let’s make our first API call to get a text completion:

const response = await openai.createCompletion({
  model: "claude-instructions-x",
  prompt: "What is Claude AI?",
  max_tokens: 100,
});

This sends the prompt “What is Claude AI?” to Claude model “claude-instructions-x” asking for a 100 token (roughly 100 word) completion.

The full response containing the generated text is stored in response.

Print the Response

To print the full response JSON, we can simply console.log it:

console.log(response.data);

This will output the completion text, model used, number of tokens, and other metadata.

Specifically the choices.text field contains the actual generated text from our prompt that we usually want.

Access the Generated Text

Instead of the full response, we likely just want the completed text itself:

const aiText = response.data.choices[0].text; 

console.log(aiText);

Now aiText contains just the string of text Claude AI generated for our prompt!

We can display it on our webpage or process it further in our code.

Send Additional Parameters

We can customize the completions by sending additional parameters to Claude API:

const response = await openai.createCompletion({
  model: "claude-instructions-x",
  prompt: "How do I make pasta?",
  temperature: 0.5,
  max_tokens: 300,
  top_p: 1,
  frequency_penalty: 0, 
  presence_penalty: 0
});

This adjusts parameters like temperature, number of tokens, top p filtering algorithm, frequency penalty, and presence penalty – which will tailor the type of generated completions.

Experiment with different values to see impacts on the text Claude returns!

Handle Errors Gracefully

We should handle any potential errors from the API gracefully in our code:

try {
  const response = await openai.createCompletion(promptOptions);  
  // ... rest of code
} catch (error) {
  if (error.response) {
    console.error(error.response.status, error.response.data); 
  } else {
    console.error(`Error calling Claude AI: ${error.message}`);
  }
}

This will catch errors returned if Claude API requests fail and log details about the issue instead of a cryptic exception that crashes our app.

Create Chatbot Conversations

In addition to single turn text completions, we can use Claude to power chatbots by chaining together multiple API calls with conversation history:

const conversation = [];

conversation.push({role: "system", content: "You are a helpful assistant named Claude."});

conversation.push({role: "user", content: "Hello Claude! How are you today?"});

const response = await openai.createChatCompletion({
  model: "claude-instructor",  
  messages: conversation
});

conversation.push({
  role: "assistant", 
  content: response.data.choices[0].message
})

This keeps track of a conversation context array that we include in the request. Claude uses previous messages to inform next responses!

We can keep chaining these requests to have a multi-turn conversation.

Call other Claude Models

So far we focused on Claude’s text generation capabilities using the claude-instructions models.

Claude also offers other models with different capabilities we can leverage:

Text Summary

Get a concise summary:

const response = await openai.createCompletion({
  model: "claude-explainer",
  prompt: LONG_TEXT_HERE,
  max_tokens: 300,
});

Sentiment Analysis

Analyze emotional sentiment of text:

const response = await openai.createCompletion({
  model: "claude-tone-analyzer",
  prompt: "I really enjoyed that movie!",
});

Language Translation

Translate text to another language:

const response = await openai.createCompletion({
  model: "claude-translator",
  prompt: "Hello my friend",
  parameters: {
    target_language: "es"  
  } 
});

And more models for codes, SEM/SEO, and questions coming soon!

Call the Classifications Endpoint

In addition to text generation completions, Claude offers a createClassification endpoint for content moderation and analysis.

Classify if text is toxic:

const response = await openai.createClassification({
  model: "claude-moderator",
  input: "That movie was trash!" 
});

Response will include a toxicity score between 0-1. Useful for comments, chatbots, and UGC.

Next Steps

And that covers the basics of using the Claude AI API with JavaScript!

Some next things to try:

• Build a Claude-powered chatbot
• Create a Claude content summarizer web app
• Classify toxicity scores of user inputs
• Generate creative stories, lyrics, scripts and more!

The possibilities are endless with what you can build. Claude’s API makes it easy to integrate advanced AI into any JavaScript project.

Check out the Claude AI documentation for even more details, tips, and examples!

Conclusion

The Claude AI API opens up an exciting world of possibilities for JavaScript developers. With simple API calls, we can tap into industrial-strength natural language generation, conversation management, content analysis, and more to build smart applications.

Whether creating a helpful chatbot, generating compelling text, or analyzing sentiment – Claude’s models elevate JavaScript code with AI superpowers.

By handling configuration, authentication, and responses gracefully, Claude integrates seamlessly into any Node.js or front-end stack. So start prototyping ideas with the robust Claude API today to offer next-generation AI capabilities in your JavaScript apps that will delight users!

FAQs