Nearer

NEAR Protocol introduced Chain Signatures in March 2024 during BUIDL Asia in Seoul, a feature that allows NEAR accounts to control multiple addresses across various blockchains using a decentralized Multi-Party Computation (MPC) network. This enables users to sign transactions on different blockchains with one NEAR account, simplifying cross-chain interactions. However, despite the innovation, there isn’t a widely adopted tool or wallet that fully leverages this capability. Most available wallets focus on managing a single EVM address or work only with Ethereum. There is no existing solution that provides users (especially non-technical ones) with a seamless way to create, manage, and interact with multiple EVM addresses from a single NEAR account, while performing typical blockchain activities like sending funds or staking.

This results in:

1. Complexity for Users: Users have to manage multiple EVM addresses manually, which complicates tasks like transferring funds, staking, and bridging assets across chains.
2. Staking Challenges: Users must individually search for the best APY across different blockchains, making the process inefficient and time-consuming.
3. User Experience Gaps: Current solutions lack user-friendly AI integration to simplify the experience for both Web2 and Web3 users, limiting accessibility.

Project Overview: Our project solves these problems by creating an AI-powered platform that leverages NEAR Chain Signatures to manage multiple EVM addresses with ease. Users can perform standard blockchain tasks such as creating derived wallet addresses, sending transactions, staking, and asset management—all driven by a natural language interface powered by AI (like ChatGPT). The platform provides an intuitive UI/UX that caters to both Web2 and Web3 users, making it easier for anyone to interact with blockchain technologies.

Additionally, the platform integrates a machine learning model to automate and optimize staking decisions. By analyzing real-time and historical data from the Pyth Network, the AI identifies the best staking opportunities and automatically moves assets to the chain with the highest APY. To ensure the system remains secure and abuse-free, World ID is integrated to verify that each user is a unique individual.

How it is made: NEAR Protocol: NEAR’s Chain Signatures technology allows users to manage multiple EVM wallets from a single NEAR account. This simplifies the complexity of handling multiple blockchain addresses and enables cross-chain transactions efficiently. The platform allows seamless cross-chain operations, such as transferring funds and staking, by leveraging NEAR’s Chain Signatures and interactions with multiple EVM chains. This provides users with a smooth experience while moving assets across blockchains.

Phala Network's Red Pill Contract: The Red Pill contract is integrated with OpenAI’s language models, allowing users to interact with the platform using natural language. Commands like "list wallets" or "send funds" are translated into actionable blockchain transactions through the AI interface. The AI and machine learning models are hosted on-chain using Phala’s network, ensuring decentralized, secure, and transparent processes.

Pyth Network: Real-time and historical price data from Pyth Network are fed into a Long Short-Term Memory (LSTM) machine learning model, which predicts the best staking opportunities across various EVM chains, optimizing users’ asset management.

World ID Integration: To prevent system abuse, World ID is integrated for user verification. This ensures that each user is a verified individual, eliminating the risk of bots or malicious actors exploiting the platform.

How these benefit our project: The use of NEAR Protocol's Chain Signatures with the MPC (Multi-Party Computation) model was pivotal in making the entire platform feasible, setting it apart from other solutions. The MPC model in NEAR allows for seamless management of multiple EVM addresses under a single account, a feature that drastically simplifies blockchain interactions for users. Unlike traditional wallets or decentralized applications (dApps) that require managing separate private keys or seed phrases for each address, NEAR's Chain Signatures and MPC network enable a user to control numerous EVM accounts effortlessly, using a single NEAR account and signature process.

This capability allowed us to introduce features like derived wallet management, where users can store and manage multiple wallet addresses across chains like Ethereum, Polygon, and Optimism with ease. Each derived wallet is managed through NEAR’s MPC model, making it possible to sign transactions on different chains while maintaining the same account structure. Other available blockchain solutions would require manual handling of multiple wallets, private keys, and bridges, creating complexity for users. NEAR’s MPC model eliminates this complexity by allowing users to store and manage EVM addresses without worrying about the backend technicalities, enhancing the overall user experience.

In addition, Phala Network’s Red Pill contract proved to be an invaluable asset to the project. Phala Network is relatively low-cost and extremely easy to integrate. The Red Pill contract, running AI models on-chain, offers a seamless interaction with OpenAI, which enabled us to deploy the natural language processing (NLP) interface that drives the platform's user interactions. Without Phala’s flexible and cost-effective infrastructure, running an AI-powered DeFi platform with real-time, on-chain AI processing would have been either prohibitively expensive or technically unfeasible. This allowed us to build not only a scalable platform but also one that can easily be maintained and expanded.

Phala's integration made it possible to perform tasks such as training and running an LSTM model for predicting staking opportunities directly on-chain. The platform’s affordability and ease of use made it highly practical to deploy machine learning functionalities that normally would require off-chain computation, ensuring that our solution remains decentralized and secure. These partner technologies provided a foundation that helped us simplify complex blockchain processes, making the platform both user-friendly and technically advanced.

Notable Technical Challenges and Solutions: Derived Wallet Management: A particularly hacky part of the project is the management of derived wallet addresses. Although all addresses initially exist, storing and interacting with them required a novel approach. We used a mechanism where a small test fund is sent to an address derived from a NEAR Chain Signature account, ensuring the address is active. This is achieved by iterating through the wallet’s derivation paths until an unused one is found. The system identifies this path and stores the address, allowing users to manage multiple wallets effortlessly. This method allows seamless wallet storage without manual intervention.

Cross-Chain Asset Exchange without Bridges or Exchanges: Another notable feature is the ability to bridge or swap assets without relying on traditional bridges, decentralized exchanges (DEXs), or centralized exchanges (CEXs). This is done by leveraging NEAR’s MPC (Multi-Party Computation) model. Two users can exchange assets by sending them to each other’s derived addresses, after which the MPC model grants them access to these new addresses. This eliminates the need for a third-party service to swap assets, making the process both decentralized and highly efficient.

On-Chain LSTM Model: A unique achievement in this project was the creation of an LSTM model entirely on-chain using Phala’s Red Pill agent. Typically, LSTM models rely on TensorFlow, which is not natively available in TypeScript. To overcome this limitation, we manually coded the LSTM model using mathematical formulas and logic entirely in TypeScript. This allowed us to run the model on-chain, giving users real-time AI-driven insights into the best staking opportunities without needing off-chain computational resources.