AI in Actions

​

The Universal AI Interface

The AI interface is a set of functions that allow you to interact with the AI. It provides a way to send queries to the AI, and to receive responses through the ActivePieces server.

This allows piece users to use AI in their flows, without having to worry about the underlying implementation details like authentication and rate limits.

​

Import the AI Interface

To use the AI interface, you need to import it from the @activepieces/pieces-common package.

import { AI } from '@activepieces/pieces-common';

​

Initialize the AI Instance

To initialize the AI instance in your actions, you need to provide the following parameters:

• provider: The AI provider to use. This is a string that represents the AI provider. For example, openai or anthropic.
• server: The server context object. This is an object that contains information about the server, such as the API URL and token.

  // In your action
  async run(context) {
    const ai = AI({ provider: 'openai', server: context.server })
    ...
  }

​

Chat with the AI

To chat with the AI, you can use the ai.chat.text method. This method takes an array of messages and a callback function that will be called with the response from the AI.

const response = await ai.chat.text({
  model: "gpt-4o",
  messages: [ // The context of the conversation
    {
      role: "system",
      content: "You are a helpful assistant.",
    },
    {
      role: "user",
      content: "What is the weather like today?",
    }
  ],
  creativity: 100, // Optional, default is 100 .. this a number between 0 and 100 that represents the creativity of the AI
  maxTokens: 2000, // Optional, default is 2000 .. this is the maximum number of tokens that the AI can generate
})