LANGCHAIN — What Is Exploring Genworlds?
Computers are good at following instructions, but not at reading your mind. — Donald Knuth
Exploring GenWorlds is an open-source development framework for multi-agent systems. It enables developers to create environments with specialized AI agents and shared objects to execute complex tasks and achieve common goals.
World
The World keeps track of all the Agents, Objects, and World properties. Below is a YAML example of a world definition:
world_definition:
base_args:
websocket_url: ws://localhost:7456/ws
world:
id: world
class: genworlds.worlds.world_2d.world_2d.World2D
name: All-In Podcast
description: The four "besties" discuss any topic you want.
locations:
- roundtable
objects:
- id: mic1
agents:
- id: jason_calacanisSimulation Socket
The Simulation Socket serves as the communication backbone, enabling parallel operation of the World, Agents, and Objects.
Agents
Agents are autonomous entities that perceive their surroundings and make decisions to achieve specific goals. Here is an example of a Navigation Brain constructor:
NavigationBrain(
openai_api_key=openai_api_key,
name=name,
role=role,
background=background,
personality=personality,
topic_of_conversation=topic_of_conversation,
constraints=constraints,
evaluation_principles=evaluation_principles,
n_of_thoughts=3,
)Components of an Agent
Agents have Brains to control their thinking process. For example, the Event Filler Brain generates JSON parameters required for an action the Agent executes in a World.
Memories
Agents have current and pregenerated memories. Current memories store recent and relevant events, while pregenerated memories are injected based on their relevance to the Agent’s current goals.
Agent Coordination
Agents react to each other using shared Objects. For example, the “Token Bearer” uses a Token in their inventory to communicate and signal to other Agents whose turn it is to perform an action.
Use Case Highlight — RoundTable
RoundTable is a podcast simulation that showcases the coordination capabilities of Agents in GenWorlds. It uses Objects, Agent inventories, and pre-generated memories to create dynamic discussions between Agents with specific personalities, memories, and expertise.
The GenAI Ecosystem
The GenWorlds framework is designed with modularity, flexibility, and composability, allowing developers to create their own applications. GenWorlds also offers a platform and tools for easy access to monetizing applications in a Gen-AI marketplace.