Monthly Progress Report [Tech] — April 2024
At ZkAGI, we’re constantly pushing the boundaries of innovation to deliver cutting-edge solutions for the crypto and AI communities. In our latest development cycle, our engineers have been hard at work enhancing the tech stack behind our privacy-preserving ZkSurfer. Let’s dive in and explore the exciting new features and improvements we’ve rolled out.
🔧 Backend Infrastructure:
🚀 Our Zynapse API now provides developers with an innovative messenger integration to streamline their development:
- Leveraging Telethone, we facilitated direct messaging to members, streamlined member scraping processes, and fortified authentication mechanisms.
- Email privacy protection has been enhanced by employing app passwords instead of conventional passwords, thereby adding an additional layer of security for email API endpoints.
💡 Click here to embark on a journey through our documentation and test out these enchanting endpoints for yourself
🚀 Implementation of RAG for Leo Code Gen:
- Our developers concentrated on integrating RAG (Retrieval-Augmented Generation) for Leo Code Generation.
- We employed a combination of Ollama, llamaindex, and Qdrant vector databases to optimize the performance and efficiency of our system.
- By opting for RAG instead of fine-tuning, we experienced faster results while requiring fewer computational resources.
- Our team curated a custom dataset specifically tailored for LEO Code by referring to Leo code documentation. This custom dataset ensured the accuracy and relevance needed for our model training process.
💡 Click here to learn more about the Leo Code Generation feature of our ZkSurfer application
🚀 While the application is active, it retains all chat details; however, upon closure, no information is saved except for Telegram authentication. Furthermore, sensitive node details such as private keys and passwords are never transmitted to the AI model. Instead, they are securely transferred to the relevant node for setup purposes.
🎨 Frontend Enhancement:
🚀 Utilised the Context API and Zustand aiming to elevate user experience, optimize performance, and address state management bugs within shared components effectively.
🚀 Implemented Clipboard API in Leo Code Generation feature to facilitate easier code copying for developers.
🚀 Implemented dynamic UI updates to conceal sensitive information, ensuring data privacy and enhancing privacy in ZkSurfer chat application:
- Enabled form field triggers inside the chat window to mask sensitive information entry.
- On form submit, upon successful API call, integrated functionality to display a success message within the same input field, providing users with immediate feedback and enhancing the overall user experience.
So, there you have it, folks — a sneak peek into the innovative fusion of crypto and AI we’ve been crafting. Stay tuned for more updates, and remember: With great coding power comes an even greater responsibility for privacy and security, values that we at ZkAGI prioritize in our services for developers.
