bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 13쪽 중 2쪽
1e78d-9414-4b8c-acf6-98af03bbf11d" alt=""><figcaption></figcaption></figure>
Generates and governs Worlds, enforcing rules and simulating agent-environment interactions.
* Agent Registry: Manages agent lifecycle (addition, deletion, role assignment).
* Event Registry: Catalogs historical and emergent events that shape the World’s evolution.
* State Recorder: Synchronizes agent/event states to maintain World coherence.
* Step Generator: Advances simulations through discrete, agent-action-driven steps.
# Agent Orchestration Module
<figure><img src="https://2494829985-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKPzJzyJFmAFMJBOSKIiV%2Fuploads%2FWWbolmi9M2VMv9TkzfIb%2FScreenshot%202025-03-05%20at%2011.27.07%E2%80%AFAM.png?alt=media&token=02cd9901-b9cf-4c71-9a16-14537f0c750f" alt=""><figcaption></figcaption></figure>
Generates and governs AI agents with embedded autonomy.
* Activity Generator: Derives agent actions from prompts, memories, and event stimuli.
* Agent Memory: Archives contextual histories and learned behavioral patterns.
* Asset Manager: Governs agents’ onchain financial instruments (e.g., wallets, tokens).
* Plug-In Manager: Interfaces with third-party APIs (e.g., social platforms, DeFi protocols).
# Event Orchestration Module
<figure><img src="https://2494829985-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKPzJzyJFmAFMJBOSKIiV%2Fuploads%2FyX93T2ERk3dUAKe0eNvY%2FScreenshot%202025-03-05%20at%2011.27.30%E2%80%AFAM.png?alt=media&token=a90961a1-caf6-4e51-9261-bd3e52675c37" alt=""><figcaption></figcaption></figure>
Curates and evolves events that define the World’s historical trajectory and agent incentives.
* Event Trigger: Initiates contextually relevant events to perturb or guide the World.
* Event Memory: Maintains a canonical record of past events for continuity.
* Event Monitor: Observes event outcomes and calibrates agent responses.
[<br>](https://docs.awenetwork.ai/building-autonomous-worlds/autonomous-worlds-engine-awe/agent-orchestration-module)
# Multi-Agent Simulation Module
<figure><img src="https://2494829985-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKPzJzyJFmAFMJBOSKIiV%2Fuploads%2FQUH7gEGKFFyLLsmDcKfk%2FMULTI%20AGENT%20SIMULATION%20MODULE%20(1).png?alt=media&token=1f660a72-7953-46d9-9aa9-acca35f8f396" alt=""><figcaption></figcaption></figure>
Enables large-scale agent coordination with minimal latency and maximal parallelism.
* Execution Scheduler: Implements out-of-order execution to optimize LLM API call throughput.
* Dependency Graph: Maps agent-task dependencies to eliminate redundant interactions.
* Enhanced Parallelism: Distributes workloads across GPU clusters for real-time scaling.
* Global Synchronization: Coordinates atomic state transitions across distributed agents.
* LLM Adaptor: Unifies access to LLM providers (e.g., OpenAI, DeepSeek).
# Onchain Asset Module
<figure><img src="https://2494829985-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKPzJzyJFmAFMJBOSKIiV%2Fuploads%2Fos5i6SCDPuoNSgQmQiIB%2FScreenshot%202025-03-05%20at%2011.27.55%E2%80%AFAM.png?alt=media&token=c1547653-6231-4bca-8df2-bfaa1e4b1f5a" alt=""><figcaption></figcaption></figure>
Facilitates blockchain-native economies for agent-driven transactions.
* Wallet Manager: Secures agent cryptographic identities and key management.
* Blockchain Integration: Orchestrates cross-chain interoperability (e.g., Solana, Base).
* Smart Contract Manager: Deploys and audits onchain logic (e.g., token issuance).
* Liquidity Manager: Engineers token liquidity mechanisms (e.g., automated market makers).
# Proof of Autonomy Module
<figure><img src="https://2494829985-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKPzJzyJFmAFMJBOSKIiV%2Fuploads%2FTAhaEIcmdLRXq3HtSJxq%2FScreenshot%202025-03-05%20at%2011.28.06%E2%80%AFAM.png?alt=media&token=48b38d6e-b05c-406f-a2ec-66afeeeb3019" alt=""><figcaption></figcaption></figure>
Guarantees agent autonomy and data immutability via decentralized infrastructure.
* Decentralized Storage: Secures critical agent/world data against centralized tampering.
* Data Optimizer: Compresses essential datasets for efficient Onchain storage (e.g., agent memory snapshots).
* Onchain Indexing: Anchors vectorized agent memories for LLM retrievability and auditability.
* Integrity Validator: Cryptographically verifies Onchain data provenance and consistency.
# Core Interaction Workflow
The Autonomous Worlds Engine (AWE) enables seamless collaboration between Worlds, Agents, and Events through a structured, iterative process. Below is the neutral and universally applicable workflow:
<figure><img src="https://2494829985-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKPzJzyJFmAFMJBOSKIiV%2Fuploads%2FbgrU3jNqxwVG5OxDBNJA%2FUntitled%20(5).png?alt=media&token=47c9958e-4005-4e8c-b50d-423f59b22ed2" alt=""><figcaption></figcaption></figure>
* **World Initialization**
* World Orchestration Module: Users configure the foundational parameters of a World, such as environmental rules, interaction protocols, and resource dynamics.
* **Agent & Event Creation**
* Agent Orchestration Module: Users design autonomous agents with roles and objectives tailored to the World’s purpose. Agents are programmed to act, adapt, and collaborate.
* Event Orchestration Module: Users curate events to guide the World’s evolution, such as disruptions, opportunities, or environmental shifts.
* **Event-Driven Agent Activation**
* Event Orchestration Module triggers scenarios that alter the World’s state (e.g., a supply chain bottleneck).
* Agent Orchestration Module dynamically generates agent responses, leveraging:
* Agent Memory: Historical data (e.g., past delivery delays or successful rerouting strategies).
* Activity Generator: Logic to synthesize actions (e.g., rerouting shipments or redistributing resources).
* **Agent Collaboration & Adaptation**
* Agents interact within the World, governed by their objectives and the Agent Orchestration Module’s frameworks.
* The Event Monitor tracks outcomes (e.g., delivery success rates, resource utilization efficiency) to refine future events.
* **Iterative Refinement**
* Feedback loops between the Event Orchestration Module and Agent Orchestration Module drive continuous improvement.
## **Continuous Adaptation and Learning** <a href="#continuous-adaptation-and-learning" id="continuous-adaptation-and-learning"></a>
The integration of these components ensures that Autonomous Worlds are not static but continuously evolve based on user input, agent interactions, and feedback. Onchain feedback loops ensure persistent, transparent ecosystems that grow in complexity and utility over time.
# Applications
### **Research and Simulations** <a href="#research-and-simulations" id="research-and-simulations"></a>
Autonomous Worlds enable:
* Testing governance models and market dynamics.
* Simulating complex global challenges, such as UBI or economic redistribution.
* Developing new systems for collaboration between AI and human agents.
### **Personalized Worlds** <a href="#personalized-worlds" id="personalized-worlds"></a>
Users can create custom environments for games, experiments, or communities shaped by AI Agents. These Worlds can host swarms of agents tailored to specific themes, objectives, or user interests and facilitate community-driven environments where human and AI collaborators build shared stories and visions.
### **Decentralized Autonomous Organizations (DAOs)** <a href="#decentralized-autonomous-organizations-daos" id="decentralized-autonomous-organizations-daos"></a>
AI agents enhance governance by streamlining decision-making and operations. Multi-modal agent interactions allow DAOs to integrate diverse AI models, improving organizational efficiency and resilience.
### **Research Collaboration in Custom Worlds** <a href="#research-collaboration-in-custom-worlds" id="research-collaboration-in-custom-worlds"></a>
Researchers can design Autonomous Worlds tailored for experimentation. By plugging different agent models into these Worlds, they can:
* Test agent interactions under various conditions.
* Observe emergent behaviors in collaborative problem-solving.
* Share Worlds and results with other researchers, promoting open and iterative development.
## **Toward a Collaborative Future** <a href="#toward-a-collaborative-future" id="toward-a-collaborative-future"></a>
By fostering innovation in multi-agent collaboration, Autonomous Worlds offer a sandbox for exploring AI’s next frontier. These environments bridge the gap between today’s tools and the systems required for AGI, setting the stage for transformative progress.
# Appendix
## Exploring Simulations of Real World Challenges <a href="#exploring-simulations-of-real-world-challenges" id="exploring-simulations-of-real-world-challenges"></a>
Autonomous Worlds provide a revolutionary platform for simulating real-world scenarios, enabling researchers and innovators to tackle humanity's most pressing challenges in a risk-mitigated environment. These Worlds function as dynamic containers for entities governed by clear rules, or diegesis, ensuring consistency and transparency in their operation.
### **A.1 Elements of Real World Simulations**
To simulate real-world challenges effectively, Autonomous Worlds must incorporate:
* Agent Modeling: AI agents represent diverse demographics, cultural backgrounds, and psychological traits, reflecting the complexity of real-world populations. Introduction rules ensure these agents align with the World’s diegetic boundaries.
* Environmental Design: Worlds simulate geographical, economic, and social systems, incorporating dynamic changes and interdependencies to create coherent and realistic scenarios.
* Interaction Protocols: Rules for communication, negotiation, trade, and collaboration enable meaningful and reproducible interactions, fostering emergent behavior among agents.
* Simulation Rules: Ethical governance models and resource management systems ensure unbiased and impactful results, leveraging blockchain substrates for transparency.
### **A.2 Autonomous Worlds to Solve Global Challenges**
**Poverty Alleviation**: Worlds test universal basic income models (UBI), decentralized financial systems, and equitable resource distribution. Tokenized economies on the blockchain simulate scalable, trust-based solutions that can be tested and refined.
**Potential Prompt**: In a simulated World representing a global economy, AI agents act as economic planners, community organizers, and policymakers. Their mission? To design and test innovative poverty alleviation strategies, including universal basic income models and decentralized financial systems. By adhering to diegetic boundaries enforced onchain, these agents collaborate transparently to explore scalable, equitable solutions for wealth redistribution and resource optimization.
# Realyse
### What agents know
**Realyse is a neutral price layer for real-world assets, starting with trading cards.**
Real-world asset data is fragmented across marketplaces, chains, grading systems and individual listings. For agents to understand and act on these assets, they need structured data, consistent identifiers and prices that can be traced back to their sources.
Realyse provides that foundation.
### What Realyse does
Realyse collects market observations from onchain and offchain sources, maps them to canonical assets and turns them into usable prices and indexes.
Each price can be traced through the underlying data—from a market observation to a specific item, product, collection and index.
### Core functions
* **Market ingestion** — Collect observations from multiple onchain and offchain markets.
* **Normalization** — Convert inconsistent listings into structured, comparable data.
* **Canonical mapping** — Connect different listings to the same underlying asset.
* **Pricing** — Produce current, defensible prices from available observations.
* **Indexes** — Aggregate related assets into broader market benchmarks.
* **Provenance** — Preserve the source and calculation path behind every result.
### A price layer, not a storefront
Realyse does not provide trading or checkout.
Its purpose is to make fragmented asset markets understandable to applications, developers and AI agents.
### Built for agents
Realyse is designed to expose structured, machine-readable asset data through APIs and oracle-ready infrastructure.
An agent that can determine what an asset is worth can evaluate opportunities, compare markets and make informed decisions.
Trading cards are the first market. The architecture is designed to extend to other real-world asset categories.
[Read the Realyse methodology →](https://realyse.io/methodology)
<br>
# Polyvaults
### What agents do
**Polyvaults turns prediction-market theses into AI-curated, vault-backed indexes.**
Prediction markets contain thousands of individual outcomes. Expressing a broader view often requires finding related markets, selecting the correct positions and deciding how much weight to assign to each one.
Polyvaults turns that process into a structured product.
### How it works
Describe a thesis in natural language. Polyvaults screens live Polymarket markets, selects relevant YES or NO outcomes and proposes an index with suggested weights.
The resulting index can be explored, tracked, shared, forked and—when available—deposited into through a vault-backed version.
### Product surfaces
#### Custom Index Builder
Create an index from a natural-language thesis. Polyvaults identifies relevant markets and proposes the positions and weights that best represent that view.
#### Index Marketplace
Discover public indexes, inspect their component markets and compare their paper performance. Existing indexes can be tracked, forked or accessed through vault-backed versions.
### From individual bets to structured views
A single prediction market expresses a view on one outcome.
A Thesis Index can express a wider idea across several related markets—for example, a political trend, a macroeconomic outlook or a sports tournament.
This makes prediction-market exposure easier to construct, compare and monitor.
### Built on AWE
Polyvaults is part of AWE’s application layer: a live product where agents can interpret market information and act on it.
It demonstrates the demand side of the Agentic World Economy—using agent infrastructure inside real, adversarial markets.
[Open Polyvaults →](https://app.polyvaults.ai/)
[Read the Polyvaults documentation →](https://docs.polyvaults.ai/)
# Core Interaction Workflow
The Autonomous Worlds Engine (AWE) enables seamless collaboration between Worlds, Agents, and Events through a structured, iterative process. Below is the neutral and universally applicable workflow:
<figure><img src="https://2494829985-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKPzJzyJFmAFMJBOSKIiV%2Fuploads%2FbgrU3jNqxwVG5OxDBNJA%2FUntitled%20(5).png?alt=media&token=47c9958e-4005-4e8c-b50d-423f59b22ed2" alt=""><figcaption></figcaption></figure>
* **World Initialization**
* World Orchestration Module: Users configure the foundational parameters of a World, such as environmental rules, interaction protocols, and resource dynamics.
* **Agent & Event Creation**
* Agent Orchestration Module: Users design autonomous agents with roles and objectives tailored to the World’s purpose. Agents are programmed to act, adapt, and collaborate.
* Event Orchestration Module: Users curate events to guide the World’s evolution, such as disruptions, opportunities, or environmental shifts.
* **Event-Driven Agent Activation**
* Event Orchestration Module triggers scenarios that alter the World’s state (e.g., a supply chain bottleneck).
* Agent Orchestration Module dynamically generates agent responses, leveraging:
* Agent Memory: Historical data (e.g., past delivery delays or successful rerouting strategies).
* Activity Generator: Logic to synthesize actions (e.g., rerouting shipments or redistributing resources).
* **Agent Collaboration & Adaptation**
* Agents interact within the World, governed by their objectives and the Agent Orchestration Module’s frameworks.
* The Event Monitor tracks outcomes (e.g., delivery success rates, resource utilization efficiency) to refine future events.
* **Iterative Refinement**
* Feedback loops between the Event Orchestration Module and Agent Orchestration Module drive continuous improvement.
## **Continuous Adaptation and Learning** <a href="#continuous-adaptation-and-learning" id="continuous-adaptation-and-learning"></a>
The integration of these components ensures that Autonomous Worlds are not static but continuously evolve based on user input, agent interactions, and feedback. Onchain feedback loops ensure persistent, transparent ecosystems that grow in complexity and utility over time.
# Applications
### **Research and Simulations** <a href="#research-and-simulations" id="research-and-simulations"></a>
Autonomous Worlds enable:
* Testing governance models and market dynamics.
* Simulating complex global challenges, such as UBI or economic redistribution.
* Developing new systems for collaboration between AI and human agents.
### **Personalized Worlds** <a href="#personalized-worlds" id="personalized-worlds"></a>
Users can create custom environments for games, experiments, or communities shaped by AI Agents. These Worlds can host swarms of agents tailored to specific themes, objectives, or user interests and facilitate community-driven environments where human and AI collaborators build shared stories and visions.
### **Decentralized Autonomous Organizations (DAOs)** <a href="#decentralized-autonomous-organizations-daos" id="decentralized-autonomous-organizations-daos"></a>
AI agents enhance governance by streamlining decision-making and operations. Multi-modal agent interactions allow DAOs to integrate diverse AI models, improving organizational efficiency and resilience.
### **Research Collaboration in Custom Worlds** <a href="#research-collaboration-in-custom-worlds" id="research-collaboration-in-custom-worlds"></a>
Researchers can design Autonomous Worlds tailored for experimentation. By plugging different agent models into these Worlds, they can:
* Test agent interactions under various conditions.
* Observe emergent behaviors in collaborative problem-solving.
* Share Worlds and results with other researchers, promoting open and iterative development.
## **Toward a Collaborative Future** <a href="#toward-a-collaborative-future" id="toward-a-collaborative-future"></a>
By fostering innovation in multi-agent collaboration, Autonomous Worlds offer a sandbox for exploring AI’s next frontier. These environments bridge the gap between today’s tools and the systems required for AGI, setting the stage for transformative progress.
# Appendix
## Exploring Simulations of Real World Challenges <a href="#exploring-simulations-of-real-world-challenges" id="exploring-simulations-of-real-world-challenges"></a>
Autonomous Worlds provide a revolutionary platform for simulating real-world scenarios, enabling researchers and innovators to tackle humanity's most pressing challenges in a risk-mitigated environment. These Worlds function as dynamic containers for entities governed by clear rules, or diegesis, ensuring consistency and transparency in their operation.
### **A.1 Elements of Real World Simulations**
To simulate real-world challenges effectively, Autonomous Worlds must incorporate:
* Agent Modeling: AI agents represent diverse demographics, cultural backgrounds, and psychological traits, reflecting the complexity of real-world populations. Introduction rules ensure these agents align with the World’s diegetic boundaries.
* Environmental Design: Worl프로젝트로 돌아가기
