iAgentes dyorAI

Crypto Everyone

Inspiration

The inspiration for Dyorai is personal and direct: our mothers.

We (D.S. Duarte, a programmer, and Mario Filho, in marketing) watched our families master complex digital tools like instant bank transfers (PIX in Brazil) and online shopping, but remain terrified of Web3.

They hear "Bitcoin" or "Wallet" and see an insurmountable wall of complexity, fear, and technical jargon. We realized the biggest barrier to mass adoption isn't technology; it's trust and education.

DyorAI was born from a simple question: "How can we build an AI agent that is patient, empathetic, and human enough to be the 'private tutor' our mothers wish they had to navigate this new world?" We didn't want to build another trading bot; we wanted to build a confidence-builder.

What we learned

This hackathon taught us that in Web3, empathy is as important as security. A user's first five minutes are critical. If they feel confused or intimidated, they are lost forever.

Technically, we learned the immense power of combining specialized AI services. Using Convai for a dynamic personality and ElevenLabs for a trustworthy, calm voice isn't a gimmick; it's a core feature. It directly combats user fear in a way a simple text-bot cannot.

We also validated that seamless onboarding is non-negotiable. Integrating Thirdweb confirmed that removing all friction from wallet creation is the first, and most important, victory in a new user's journey.

How we built it

DyorAI is built on a modern, modular AI and Web3 stack, designed for empathy and security:

AI Agent Core:

Convai: For the agent's personality, memory, and managing the conversational flow.

ElevenLabs: For realistic, empathetic, and low-latency voice generation.

LLMs (Various): As the core knowledge "brain," grounded by a curated RAG (Retrieval-Augmented Generation) database to prevent hallucinations and ensure accuracy.

Web3 & Infrastructure:

Thirdweb: To handle all wallet connection and creation logic securely and simply

React: For the clean, responsive, and simple frontend interface.

Supabase: For a scalable backend, user authentication, and database.

Circle (USDC): As the core asset for our first guided tutorials, providing a stable, understandable entry point for users new to digital assets.

Challenges we faced

Balancing Simplicity vs. Power: The biggest challenge was not adding features. It's easy to build a complex dashboard. It's incredibly hard to build a simple, single-purpose agent that feels human. We had to constantly cut features to stay true to our "mother" persona.

AI Integration & Latency: Integrating multiple AI services (voice, personality, LLM) in real-time while ensuring a low-latency, natural conversation flow was a significant technical hurdle during the limited timeframe of the hackathon.

Grounding the AI for Safety: Our top priority was preventing the AI from giving "hallucinated" or dangerous financial advice. Implementing a robust RAG system to keep the agent factual, safe, and strictly focused on our curated knowledge base was a major challenge that required constant testing.