DeepThought

DeepThought – AI-Powered Crypto Chat Bot (⌐ ͡■ ͜ʖ ͡■)

DeepThought brings Web3 directly to WhatsApp, turning complicated blockchain interactions into easy, everyday chats—just like messaging your bestie. Forget juggling wallets, cryptic UIs, or lengthy tutorials. Just type your request, and DeepThought handles everything behind the scenes.

Here’s what makes DeepThought awesome:

Natural Chat Interactions: Ask stuff like "What's my Ethereum balance?", "Mint an NFT on Flow," or "Roll a d20 for my on-chain RPG," all from WhatsApp. Easy-peasy, no sweat.

Multi-Chain Magic: Supports Ethereum, Flow, Hedera, Rootstock (RSK), and even oracle integrations (like Pyth)—with more chains on the way.

On-Chain Gaming: Dive straight into text-based RPGs, roll dice, or engage with mini-games without ever leaving your chat.

Secure Wallet Management: Powered by simulated EIP-7702 delegation, DeepThought safely performs actions using temporary "sub-wallets," keeping your main wallet super secure.

Smart AI Backend: Powered by OpenAI GPT-4.1, DeepThought smartly interprets your commands, translating casual text into precise blockchain actions. Plus, we're aiming for a privacy-first, self-hosted AI future.

Privacy By Design: We never share your private keys or transaction history beyond what's strictly needed. Future updates will bring end-to-end encryption and zero-knowledge proofs.

Quick & Modular: DeepThought’s modular structure means easily adding new chains, DeFi protocols, NFT marketplaces, and DAOs. Fast and hassle-free expansions incoming!

DeepThought isn't just tech—it's your chill crypto companion, bridging the gap between Web3 complexity and everyday conversations. No stress, no crypto jargon, just chat.

Ready to speak crypto? Dive in with DeepThought! ദ്ദി( • ᴗ - ) ✧

How DeepThought is Built:

DeepThought is essentially a complex WhatsApp chatbot designed to simplify interactions with Web3 technologies. Here’s a breakdown of its architecture and the technologies used:

The Core Interface is WhatsApp Users interact with DeepThought entirely through WhatsApp. This is the primary entry point for all commands and requests.

The AI Brain (Current Hackathon Setup) is OpenAI’s GPT-4.1 as the natural language processing engine. It uses function-calling capabilities. When a user sends a message (e.g., “Mint a Flower NFT”), GPT-4.1 interprets the intent (mint) and identifies the relevant parameters (Flower NFT). If the intent matches a predefined function/feature in DeepThought's toolset, the backend then constructs the necessary blockchain transaction. The user is then alerted to confirm the action. Only the "bare minimum" context is sent to GPT-4.1 to understand the request; full wallet data remains within DeepThought's secure backend.

Blockchain Interaction & Wallet Management: Our Vision (EIP-7702) is to use EIP-7702 for account abstraction (long term). This would allow users to authorize "sub-wallets" with limited permissions and a small gas allowance, enabling DeepThought to execute approved actions without the user's main private key. Approval would ideally be done via Pass-Keys.

The most notable "hacky" part, since native EIP-7702 support isn't widely available on blockchains yet, is that the project simulates this functionality for the hackathon. When a user "authorizes" DeepThought, the backend uses a wallet provided and funded by the DeepThought team. This simulated interaction is restricted to Test networks only. A crucial point for the hackathon build is: "Your Full Private Key Never shared with AI, since we currently have to work around the non existing EIP, we do have our own private key added to the code so users dont have to connect."

This means the team has embedded a private key for the bot's operations on testnets, so users don't need to connect their own wallets during the demo.

Specific Blockchain & Partner Technology Integrations (Current Capabilities for EthGlobal Prague 2025):

• Blockscout Integration: Technology: Blockscout API. How it's pieced together: DeepThought queries the Blockscout API to fetch transaction data for a given address or known name. It then generates a screenshot styled with Blockscout branding and delivers this image to the user in WhatsApp.

Benefit: Provides users with familiar, Blockscout-branded transaction history directly in chat, leveraging Blockscout's robust explorer data.

• Hedera NFT Minting & Smart Contract Deployment: Technologies: Hedera network, IPFS. How it's pieced together: DeepThought deploys a pre-generated smart contract on Hedera. An algorithm (with its art generation script published on IPFS) creates unique planet-like flower patterns based on wallet addresses for the NFTs. Users can mint one NFT per wallet via WhatsApp. The smart contract generates a public IPFS link for the art, tied to the user's wallet, and issues the NFT. An image of the NFT and its info is sent to the user in WhatsApp. Benefit: Demonstrates AI-driven NFT generation and smart contract interaction on Hedera, using IPFS for decentralized art storage, all managed via chat.

• Rootstock (RSK) On-Chain Text-Based RPG: Technology: Rootstock (RSK) blockchain. How it's pieced together: A smart contract powering a simple text-based RPG is deployed on Rootstock. Users interact with this contract through WhatsApp to start games and make choices. The AI (GPT-4.1) uses minimal information stored on-chain by the RPG contract to generate and describe "visually stunning and intriguing scenarios." Benefit: Showcases how LLMs can enhance limited on-chain data from RSK to create richer experiences, demonstrating a symbiosis between expensive on-chain storage and generative AI.

• Pyth Network Verifiable Randomness: Technology: Pyth Network's entropy contract on the Sanko testnet. How it's pieced together: DeepThought deployed its own random number contract on Sanko that calls Pyth's entropy contract (using the default provider on Sanko). Users request dice rolls in WhatsApp using natural language (e.g., "roll a normal dice") or D&D notation (e.g., "1d20"). The AI processes the natural language, converting it to D&D notation for the backend. The backend uses Pyth via the deployed contract to generate random numbers, and the results are sent back to WhatsApp. Benefit: Leverages Pyth's verifiable RNG for fair and transparent dice rolls in on-chain gaming scenarios, made accessible through natural language.

######## Piecing it All Together (Simplified Flow): User sends a command/request via WhatsApp. The DeepThought backend receives the message. OpenAI GPT-4.1 (with function calling) interprets the user's intent. Based on the intent, the backend: Interacts with the relevant blockchain (Hedera, RSK, Pyth on Sanko testnet) using the team-provided and funded wallet (the "spoofed EIP-7702"). Calls external APIs (like Blockscout). Utilizes IPFS for data storage (like Hedera NFT art). The results (confirmations, data, images, game responses) are formatted and sent back to the user via WhatsApp.

Notable "Hacky" Elements: Spoofed EIP-7702: The most significant workaround is simulating EIP-7702 by using a team-controlled, pre-funded wallet on testnets. This is a practical solution for a hackathon where full EIP-7702 isn't available. Team's Private Key in Code: To facilitate user interaction without requiring them to connect wallets for the demo, the team has "our own private key added to the code." This is standard for quick prototyping in a hackathon but would not be done in a production system. Pre-generated Smart Contract for Hedera: While the NFT art is dynamically generated, the smart contract itself that gets deployed on Hedera is "pre-generated," rather than the AI dynamically creating and deploying varied smart contracts on the fly (which is a future vision).

Focus on Test Networks: Due to the "spoofed" wallet and hackathon context, interactions are limited to test networks.

This setup allows DeepThought to demonstrate its core vision of chat-based Web3 interaction within the constraints of a hackathon like EthGlobal Prague 2025, with clear plans to transition the "hacky" or simulated parts to more robust, native solutions as the underlying technologies mature.