npm install @boundaryml/baml

npx baml-cli init

rm baml_src/resume.baml

cp ./walkthrough/01-agent.baml baml_src/agent.baml

// ./walkthrough/01-agent.baml
class DoneForNow {
  intent "done_for_now"
  message string 
}

client<llm> Qwen3 {
  provider "openai-generic"
  options {
    base_url env.BASETEN_BASE_URL
    api_key env.BASETEN_API_KEY 
  }
}

function DetermineNextStep(
    thread: string 
) -> DoneForNow {
    client Qwen3

    // use /nothink for now because the thinking tokens (or streaming thereof) screw with baml (i think (no pun intended))
    prompt #"
        {{ _.role("system") }}

        /nothink 

        You are a helpful assistant that can help with tasks.

        {{ _.role("user") }}

        You are working on the following thread:

        {{ thread }}

        What should the next step be?

        {{ ctx.output_format }}
    "#
}

test HelloWorld {
  functions [DetermineNextStep]
  args {
    thread #"
      {
        "type": "user_input",
        "data": "hello!"
      }
    "#
  }
}

npx baml-cli generate

export BAML_LOG=debug

cp ./walkthrough/01-cli.ts src/cli.ts

// ./walkthrough/01-cli.ts
// cli.ts lets you invoke the agent loop from the command line

import { agentLoop, Thread, Event } from "./agent";

export async function cli() {
    // Get command line arguments, skipping the first two (node and script name)
    const args = process.argv.slice(2);

    if (args.length === 0) {
        console.error("Error: Please provide a message as a command line argument");
        process.exit(1);
    }

    // Join all arguments into a single message
    const message = args.join(" ");

    // Create a new thread with the user's message as the initial event
    const thread = new Thread([{ type: "user_input", data: message }]);

    // Run the agent loop with the thread
    const result = await agentLoop(thread);
    console.log(result);
}

src/index.ts
+import { cli } from "./cli"
+
 async function hello(): Promise<void> {
     console.log('hello, world!')
 
 async function main() {
-    await hello()
+    await cli()
 }
 

cp ./walkthrough/01-index.ts src/index.ts

cp ./walkthrough/01-agent.ts src/agent.ts

// ./walkthrough/01-agent.ts
import { b } from "../baml_client";

// tool call or a respond to human tool
type AgentResponse = Awaited<ReturnType<typeof b.DetermineNextStep>>;

export interface Event {
    type: string
    data: any;
}

export class Thread {
    events: Event[] = [];

    constructor(events: Event[]) {
        this.events = events;
    }

    serializeForLLM() {
        // can change this to whatever custom serialization you want to do, XML, etc
        // e.g. https://github.com/got-agents/agents/blob/59ebbfa236fc376618f16ee08eb0f3bf7b698892/linear-assistant-ts/src/agent.ts#L66-L105
        return JSON.stringify(this.events);
    }
}

// right now this just runs one turn with the LLM, but
// we'll update this function to handle all the agent logic
export async function agentLoop(thread: Thread): Promise<AgentResponse> {
    const nextStep = await b.DetermineNextStep(thread.serializeForLLM());
    return nextStep;
}

  function DetermineNextStep(thread: string) -> DoneForNow {
      client Qwen3
      // ...

client "openai/gpt-4o"

export BASETEN_API_KEY=...
export BASETEN_BASE_URL=...

npx tsx src/index.ts hello

{
  intent: 'done_for_now',
  message: 'Hello! How can I assist you today?'
}