Jul 10, 2024 • 3 min read
Valory Unveils 'Propel Genie' at EthCC: Supercharging the Creation of Agents for Olas & Beyond
Brussels, July 10, 2024 – During this year’s Ethereum Community Conference (EthCC), Valory announced the introduction of Propel Genie at the “Agents Unleashed” event. Genie is an AI agent that, acting as a software engineer, generates autonomous AI agents, thereby revolutionizing the developer experience for Builders in the Olas ecosystem and beyond.
Valory has been at the forefront of integrating AI and blockchain technology to create open-source and user-ownable autonomous agents since 2021. Valory is a co-creator of Olas, the platform for co-owning AI, which has generated swarms of AI agents making over 900,000 transactions.
With the launch of Genie, Valory unveils an agent-generating agent, drastically simplifying the developer experience and pushing the boundaries for what’s possible in the autonomous agent space. Genie will be made available via Valory’s Propel platform, a platform offering development and deployment solutions for autonomous agents.
What is Propel Genie?
Genie aims to 10x the developer experience for creators of autonomous agents. Tailored for initial use by Builders in the Olas ecosystem, Genie utilizes generative AI to take design specifications in natural language and turn them into fully functional autonomous AI agent code, ready for deployment on Olas and beyond.Key Features
- Prompt to Full-Agent Execution Through Genie’s intuitive interface, users begin by defining their agent in natural language. Genie translates these descriptions into a technical roadmap, known as a Finite State Machine (FSM) specification, which outlines the essential structure for the agent architecture. This roadmap becomes the basis for further AI-generated code that forms the skeleton of the new agent. Users then have the chance to refine and approve this foundation, simplifying the development process and ensuring that the end product aligns perfectly with their vision, making it quicker and easier to move from idea to reality.
- Interactive Development Environment Genie’s user-friendly framework allows users to design specific components of agents around their own logic. This hands-on approach ensures that the agent perfectly aligns with their specifications.
- Advanced Learning Capabilities Genie leverages multiple SOTA large language models to enhance the quality of user-provided prompts. This ensures the development plan is accurate and optimized for the intended functionality.
- Rapid Prototyping Genie enables rapid prototyping through automated workflows to facilitate the swift creation and iteration of autonomous agents. This allows developers to quickly build, test, and refine their agent ideas, significantly speeding up the development cycle. By streamlining the prototyping process, Genie reduces the time taken to build and test prototypes and enables developers to adapt more dynamically to evolving requirements.
- Adaptive Multi-Agent Collaboration Genie's innovative architecture employs multiple language models in a collaborative multi-agent system, ensuring a comprehensive and adaptive evaluation at every stage of development. This integration allows for a dynamic feedback loop that continuously refines and optimizes the agent creation process. By leveraging the strengths of various specialized agents, Genie not only enhances the precision of the development but also ensures that the final agent is robust and reliable.
Benefits
- Fast-Track Development Genie drastically reduces the development timeline by automating the initial phases of code generation and structuring. What previously could take months can now be accomplished in days.
- Targeted Precision Each output is tailored to the user’s specific requirements, offering a high degree of precision and customization in the development of agents.
- Optimized Efficiency By automating the initial stages of code generation and structuring, Genie allows developers to focus on refining and customizing their projects, boosting overall efficiency.