Architecture

Helva’s architecture as a multi-agent system is designed for modularity, scalability, and seamless interaction with both users and decentralized infrastructure. It can be summarized into the following three main components:

  1. Back-End- that manages operations. Built using Python along with Web3.py, Flask

  2. AI Agent- that makes every operation smarter. Built using Python with Langchain package.

  3. UI- that makes everything accessible from a simple UI. Built using React JS.

Back-End

The Back End layer represents the operational core of Helva and includes the following categories of functions:

  • Smart Contract Interactions This set of tools encompasses all the code and logic required to interact with EVM-based blockchain ecosystems. It enables seamless integration with DeFi protocols such as Quickswap, lending and borrowing platforms, liquid staking services, and native chain functionalities. The functions accessible by Helva are the following: - Token Swaps -LPs Tokens Management - Perpetuals - Lending - Borrowing - Liquid Staking Different DeFi Protocols may be enabled to access those functions. To see the full list, please head to Integrations List section. If you want to apply your token and/or DeFi Protocol, fill this form.

  • Analysis This set of tools is responsible for collecting, processing, and analyze both on-chain and off-chain data related to Polygon and all the tokens listed. It includes advanced pipelines for tracking and retreiving the following insights, divided by category: 1. Technical Indicators -Historical Price on 1m, 1H, 1d, 7d, 30d Candles with up to 1 year historical data -Moving Averages: SMA, EMA, MAC -Volatility Indicators: Bollingher Bands, Average True Range, Std Deviation -Momentum & Volume Indicators: OBV, ROC, Volume RSI, ADX -Oscillators: RSI, Stochastic Oscillator 2. On-Chain & Fundamental Indicators -Historical Total-Value-Locked -Historical 30d Daily Txs -Latest Relevant News These insights, can be called singularly by the users, or used to build a custom strategy or, if you are not comfortable with these indicators, Helva can build one for you. Helva can create two types of strategies/analysis: 1. Technical- Strategies with MACD, RSI, Bollingher Bands etc. 2. On-Chain - Including Historical TVL,Stable Mcap, Daily Transactions History etc.

  • Tools This set of tools enables the access to the most common on-chain functions, accessible natively through the Web3 Library: -Send Native and ERC-20 Tokens -Transactions & Blocks Data Retrieval -Chain Status -Account Balances

AI Agent

The LLM, or the AI Agent, is where Helva transforms raw data, backend logic, and user inputs into intelligent actions and insights. This is achieved through the integration of state-of-the-art language models via the Langchain framework.

Langchain serves as the orchestration layer that connects the AI's reasoning capabilities to Helva’s custom tools—such as DeFi strategy modules, smart contract functions, data analysis pipelines. This integration offers several key advantages:

  • Custom Tool Integration Langchain allows Helva to link AI APIs directly with personalized functions, enabling seamless interaction with DeFi protocols and blockchain infrastructure in a modular and controlled manner.

  • Model Flexibility and Upgradability By abstracting the language model layer, Langchain makes it easy to switch or upgrade the AI provider. For example, Helva can start with GPT and later transition to a more advanced model without major architectural changes.

User Interface

Helva’s User Interface is the gateway through which users interact directly with the platform. Designed to be intuitive and familiar, the UI resembles a typical conversational prompt—much like ChatGPT—but with a Web3 twist: users log in securely via their MetaMask wallet.

The UI is composed of three main components:

  • Conversational Prompt: A clean, chat-style interface where users can type natural language commands to instruct Helva. Whether it's executing a token swap, requesting a trade analysis & execution, or deploying a personalized yield strategy, everything starts with a message.

  • Navigation Menu: A sidebar on the left provides quick access to key features and sections, streamlining the user experience.

  • Chat History: Users can view and revisit all their past conversations with Helva. Prefer a fresh start? A simple action button lets you spin up a new session instantly.

Last updated