Integrating Custom Agents with Built-in Agents

With AI Refinery, you can seamlessly access LLM, VLM, and Embedding models using the standard AIRefinery API. The Authenticator object in AI Refinery takes care of authentication for AIRefinery, ensuring a smooth integration process. In this tutorial, we will demonstrate how to utilize the standard AIRefinery API with your Custom Agent, and use it along with a built-in utility agent.

Objective

Combine custom and built-in agents using the AI Refinery SDK to create and run a simple AI system that helps users plan parties.

Steps

1. Configuration file

As a first step, you simply need to create a yaml file with all the required configuration.

1. The Recommender Agent is the agent that you will design to use the AIRefinery API to provide general recommendation.

2. The Party Planner Agent uses the PlanningAgent from the AIRefinery™ Library that is capable of providing concrete planning schemes for the user. The Party Planner Agent will use the chat history (context: - "chat_history") stored in the AIRefinery™ memory dedicated for your project to provide a concrete plan according to the user query.

utility_agents:
  - agent_class: CustomAgent
    agent_name: "Recommender Agent"
    agent_description: |
      The Recommender Agent is a specialist in item recommendations. For instance,
      it can provide users with costume recommendations, items to purchase, food,
      decorations, and so on. 
    config: {}

  - agent_class: PlanningAgent
    agent_name: "Party Planner"
    agent_description: |
      The Party Planner agent is specialized in helping users planning their parties.
      For example, how to organize a halloween party, christmas party, and so on.
      Don't call this agent for item recommendations.
    config:
      output_style: "markdown"
      contexts:
        - "chat_history"

super_agents: []

orchestrator:
  agent_list:
    - agent_name: "Party Planner"
    - agent_name: "Recommender Agent"

2. Python file

Now, you can start the development of your assistant using:

1. AIRefinery API to enable the LLM capabilities of your Custom Agent.

2. DistillerClient to take advantage of the other features of AIRefinery™.

import os

from air import AsyncAIRefinery, DistillerClient, login
from dotenv import load_dotenv

load_dotenv()  # This loads your ACCOUNT and API_KEY from your local '.env' file

auth = login(
    account=str(os.getenv("ACCOUNT")),
    api_key=str(os.getenv("API_KEY")),
)

distiller_client = DistillerClient()

project = "party_project"

distiller_client.create_project(config_path="config.yaml", project=project)


async def recommender_agent(query: str) -> str:
    global auth

    prompt = """Given the query below, your task is to provide the user with useful and cool
       recommendation followed by a one-sentence justification.\n\nQUERY: {query}"""

    prompt = prompt.format(query=query)

    airefinery_client = AsyncAIRefinery(**auth.openai())

    response = await airefinery_client.chat.completions.create(
        messages=[{"role": "user", "content": prompt}],
        model="meta-llama/Llama-3.1-70B-Instruct",
    )

    return response.choices[0].message.content


executor_dict = {"Recommender Agent": recommender_agent}

response = distiller_client.interactive(
    project=project, uuid="test_user", executor_dict=executor_dict
)