Version Packages

[github-actions[bot]]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

@statelyai/[email protected]

Minor Changes

• #26 2d63f30 Thanks @davidkpiano! - The createAgent(…) function now returns agent actor logic instead of the state machine actor logic. That means its return value can now be used directly in actors for a state machine, and it will perform tool calls and choose the correct event to send back to the machine:

  import { createAgent } from "../src";
  import { z } from "zod";
  import { setup, createActor } from "xstate";
  import OpenAI from "openai";
  
  const openai = new OpenAI({
    apiKey: process.env.OPENAI_API_KEY,
  });
  
  const agent = createAgent(openai, {
    model: "gpt-3.5-turbo-16k-0613",
    events: {
      "agent.thought": z.object({
        text: z.string().describe("The text of the thought"),
      }),
    },
  });
  
  const machine = setup({
    actors: { agent },
  }).createMachine({
    initial: "thinking",
    states: {
      thinking: {
        invoke: {
          src: "agent",
          input: "Think about a random topic, and then share that thought.",
        },
        on: {
          "agent.thought": {
            actions: ({ event }) => console.log(event.text),
            target: "thought",
          },
        },
      },
      thought: {
        type: "final",
      },
    },
  });
  
  const actor = createActor(machine).start();
  

  See the examples directory for more detailed examples

Patch Changes

• #26 3f00f5e Thanks @davidkpiano! - defineEvents was removed. Use the events property in createAgent({ … }) instead:

  import { z } from "zod";
  import { createAgent } from "@statelyai/agent";
  
  const agent = createAgent({
    model: "gpt-4-1106-preview",
    events: {
      "agent.getWeather": z.object({
        location: z.string().describe("The location to get the weather for"),
      }),
      "agent.reportWeather": z.object({
        location: z
          .string()
          .describe("The location the weather is being reported for"),
        highF: z.number().describe("The high temperature today in Fahrenheit"),
        lowF: z.number().describe("The low temperature today in Fahrenheit"),
        summary: z.string().describe("A summary of the weather conditions"),
      }),
      "agent.doSomethingElse": z
        .object({})
        .describe(
          "Do something else, because the user did not provide a location"
        ),
    },
  });
  

• #28 dbdea98 Thanks @davidkpiano! - Reduces API to one function: createAgent(…). This creates an agent that:

  • Is actor logic for an agent that makes decisions based on a goal
  • Has .fromText(…) and .fromTextStream(…) helpers
  • Wraps the ai library