Why AI needs to be decentralized
As AI increasingly drives content creation, decision-making, and online interactions, critical questions of transparency and accountability continue to emerge. SubQuery addresses these issues by storing all agent interactions and reasoning on-chain, ensuring that data is immutable, traceable, and auditable.
Key issues SubQuery’s decentralized approach addresses include:
Single point of failure risk : Decentralized infrastructure improves the resilience of the system.
Opaque decision-making process : On-chain storage introduces an accountability mechanism for agent behavior.
Scalability of Autonomous Agents : Provides developers with tools to quickly prototype and deploy AI agents.
SubQuery's core focus is to deliver real application value rather than relying on hype, aiming to build trustworthy AI-driven systems.
Introduction to SubQuery AI Framework
SubQuery's AI framework is designed to make creating and managing AI agents simple and intuitive. Developers can start with a sample project that includes all the necessary components.
The main features of the framework include:
Modular design
Developed in TypeScript, combined with a modern and secure runtime environment based on Deno (Deno is Node.js reimagined).
Open Source Model
Support for integration with popular open source models, such as Llama 3B, provides developers with flexibility and independence from proprietary tools.
System prompts
Developers can define the agent's behavior, specifying its role, personality, and the purpose of each interaction.
Functional tools
Allows agents to perform specific tasks, such as making API calls, interacting with blockchain data, or executing smart contracts.
Live Demo: Build a Custom AI Agent
In the demo, SubQuery showed a step-by-step example that highlighted the simplicity of the framework. The agent, called “Greeter AI,” was designed to ask the user for their name and greet them with a unique joke or pun.
This interesting example highlights how system prompts and lightweight configuration can shape agent behavior. For more advanced use cases, the demo also explores functional tools, showing their ability to handle complex tasks such as reversing user input or retrieving blockchain data.
Real-world application case: knowledge-based transparent proxy
One prominent example of SubQuery is “Argument,” an AI agent designed to engage in debates on social media platforms.
The notable feature of Argument is its on-chain memory function:
Every interaction is recorded on-chain , including the reasoning and decision-making process.
Public access rights allow anyone to view the Argument's "thought process" through stored transactions and responses.
This approach solves a key problem in AI systems, namely that the decision-making process is often a “black box.” Storing AI reasoning on-chain allows developers and users to trace back interaction records and identify potential errors, thereby eliminating ambiguous issues such as “hallucinations.”
Expand your knowledge with RAG support
Another demo highlighted SubQuery’s use of RAG (Retrieval Augmented Generation) to power an AI support bot that was trained on its comprehensive documentation.
Through RAG, SubQuery's AI achieves the following functions:
Handling large knowledge bases : Use LanceDB to convert documents into vector embeddings for efficient querying.
Accurately process user queries : Provide detailed answers from specific document sources.
Enhance usability : Provide sarcastic, simplified, or technical-style explanations with customizable prompts.
This application is a real example of how the framework can support developers in navigating a complex ecosystem, making the tools more accessible to a wider audience.
Empowering developers and innovators
SubQuery's focus goes beyond its internal use cases and aims to empower the Web3 developer community , inspiring them to imagine and develop their own AI-driven solutions. The framework provides the flexibility to explore novel ideas, whether for automating decentralized processes, building interactive dApps, or creating entirely new agent-driven platforms.
As James of SubQuery says: “We provide the tools to let your imagination run wild. The only limit is your creativity.”
Conclusion: Shaping the Future of Decentralized AI
SubQuery's AI framework demonstrates the potential of combining artificial intelligence with blockchain technology. By focusing on transparency, accountability, and real-world utility, SubQuery is redefining what is possible in the AI 3.0 era.
As decentralized agents proliferate, SubQuery’s tools will play a vital role in ensuring this new wave of innovation is both powerful and responsible. Whether through playful debate agents like Argument, or powerful AI-powered bots, SubQuery is laying the foundation for a new generation of transparent and resilient AI systems.
