I spent some time studying the differences between the current version of AI16Z's AI Agent framework Eliza and the V2 version revealed by Shawn.
Before the comparison, let's first understand the basic architecture of the current version and V2 version.
As shown in Figure 1, the current version of the Eliza framework consists of an interface layer, a core function layer, and a model layer.
Interface Layer
- Responsible for interaction with external platforms
- Supports multiple platforms such as Discord, Twitter, Telegram, etc.
- Provide API interface for other systems to call
Core Layer
- Memory system: provides retrievable storage and document processing capabilities
- Behavior system: manage custom behaviors and action execution
- Configuration system: manage configuration through .env and elizaConfig.yaml
Model Layer
- Supports multiple large language models, including OpenAI, Anthropic, Llama, Grok, etc.
- Provide a unified model interface abstraction
Each layer interacts with each other through standard interfaces, supporting a high degree of modularity and scalability.
On the 15th, Shawn revealed that Eliza v2 is currently under development and the alpha version is expected to be released in 2 to 3 weeks.
Eliza v2 will unify the message bus and simplify the client; unified Agent wallet; the model system adopts a registry and overlay mode; enhance the extensible common core framework; update community plug-ins; and complete 100% test coverage.
The architecture design diagram of V2 is speculated as follows:
Let me explain the main improvements of Eliza v2 and where they fit into the architecture:
Improvements to the interface layer: Unified Message Bus
- Change the original scattered client integration into a unified message processing system
- Each platform (Discord/Twitter, etc.) becomes an adapter for the message bus
- Simplifies client development and maintenance
- Provide a more consistent message handling experience
Improvements to the core functional layer: unified proxy wallet
- Integrate the original decentralized wallet configuration
- Provide unified multi-chain wallet management
- Simplify cross-chain operations
Extensible core framework: redesign the core interface to make it more general
- Enhance the scalability of the framework
- Better plugin integration
Community plugin update: Improved plugin system
- Improve plugin management
- 100% test coverage ensures quality
Improvements to the model layer: Introducing a model registry mechanism
- Support dynamic registration and overwriting of models
- The models are divided into two categories: LLMs and Embeddings
- More flexible model calling method
The architecture of this new version is more modular and unified, the interactions between different components are clearer, and it also provides a better foundation for future expansion.
Finally, if analyzed from an investment perspective, Eliza v2 will bring new growth opportunities for functional extension plug-ins such as AgentTank.