TL;DR: Spectral launches Syntax, your go-to copilot for the Onchain Agent Economy, now available in Beta. The Syntax Agent translates your natural language commands into ready-to-use Solidity code that can be compiled, debugged, and deployed directly within our chat interface. Additionally, discover and choose from a variety of community-curated agents tailored to your needs.
At Spectral, our mission has always been to decentralize machine intelligence, fostering a future where users access AI in a transparent and performance-verified manner. Today, we're taking a significant step forward in realizing that mission with the launch of Spectral Syntax, the first truly Onchain Agent in Web3.
Syntax: the first truly Onchain Agent in Web3
With fast-paced advancements in AI, particularly in Large Language Models (LLMs), the blend of AI with Web3 is crafting an intriguing yet emerging story. Although the idea of integrating AI agents within Web3 feels like a natural evolution and has sparked considerable discussion, we're just beginning to see the practical application of these concepts come to life.
Before delving into Syntax, it's essential to define an Onchain Agent. At Spectral, an Onchain Agent is designed to autonomously write and deploy Solidity code to the blockchain, owning a digital wallet with private keys to sign transactions independently. This allows it to perform blockchain actions similar to a human, managing contracts and assets based on its programmed logic. Loosely similar to GPTs, Onchain Agents also have their own identity and can autonomously figure out ways to interact with deployment infrastructure.
In that context, Spectral Syntax is the first true manifestation of Onchain Agents in Web3. We’re highly confident in empowering Web3 users by giving them the ability to build and use Agents for performing a variety of day-to-day onchain tasks.
Trained and fine tuned to write production grade Solidity
Image: Converting natural language into production grade Solidity
Most code generation models, such as CodeLlama, GPT-4 Code Interpreter, and DeepSeek, are not specifically designed with Solidity developers in mind. Compared to Python and JavaScript, these models are trained on significantly less Solidity code, resulting in variable performance when tasked with writing Solidity.
At Spectral, we have fine-tuned open-source models using several techniques, including Parameter Efficient Fine-Tuning (PEFT), Low-Rank Adaptation (LoRA), and DeepSpeed, along with a custom-built dataset comprising over 7 million Solidity contracts. These fine-tuned models are combined with a Retrieval-Augmented Generation (RAG) system to enhance responses with knowledge of Web3 projects, such as protocol documentation and GitHub repositories.
As a result, Spectral Syntax produces accurate and gas-optimized Solidity code for a variety of prompts ranging from low to medium complexity, and it outperforms baseline models on our Solidity evaluation dataset. As our modeling efforts continue to evolve, we expect Syntax to deliver optimized code for increasingly complex problems. These Solidity models are explicitly designed to assist developers in writing optimized, modern, well-tested, and secure code.
An Agentic Coding Experience. Even for those who don't know Solidity
Following the release of ChatGPT and GitHub Copilot, it seems fairly intuitive to ask AI to write code; we’ve seen this product feature being used across all strata of society, right from computer science grad students using it in exams, to professional developers using it to optimize their shipping velocity.
Spectral Syntax takes one step further by creating an Agentic experience: just ask the agent what you want to build, and the agent will not just write the code for you but also deploy it onchain for you at the click of a button.
Because our models are fine-tuned for Solidity, the Agent is your coding partner. If you’re a novice at Solidity, or don’t know it at all, you can ask Syntax to build the Web3 dApp you envision. Think of “Hey Syntax, launch an ERC20 coin for my project called Mypersonalcoin and mint 100M tokens of it” or “Create a decentralized voting contract in Solidity that allows token holders to vote on proposals based on their token balance”. If you're a professional developer, even better, because Syntax can accelerate and optimize your bandwidth for you by writing large libraries and components. Think, “Hey Syntax, take a look at my GitHub repo below. Can you write a set of ERC4337 Smart Wallet abstractions that create a new smart wallet every time a user joins?”
Image: Compiling code using Spectral Syntax
Image: Testing code using Spectral Syntax
Syntax also gives you the ability to copy, edit, compile, and test the AI-generated code - it's essentially an Agent-driven development environment to manifest your Web3 dApp ideas through interactions in natural language. These features also help you determine the functional relevance and usability of the code generated while also giving you control to quickly edit some snippets.
What truly sets Syntax apart is its capability to deploy code directly onchain, executing deployments through Foundry by writing instructions that Foundry can process, including code compilation, onchain deployment, and signing transactions.
Image: Deploying code onchain using Syntax
Pick the best Agents for your task
Image: Quick access sidebar to start a conversation with any of the popular agents
Syntax is pioneering an accessible onchain Agent Economy, inviting users to select agents tailored for their specific Web3 tasks. Users can either interact with the foundational agent or opt for one of the specialized agents, each adept in distinct tasks and operations—like the Code Forker assistant, ideal for those aiming to fork and customize a codebase. Additionally, users can discern an agent's expertise and trustworthiness by reviewing its category listing and the popularity reflected through its user engagement.Initially, these Agents will be curated and built by Spectral; however, any user will be able to create their own agent and monetize it in future releases of our product.
Experience the Agent Economy.
We invite you to join Syntax and start conversing with your agent! While Syntax is still in Beta, you can use it for a variety of use cases and applications by prompting various instructions to the agent. Or to get started swiftly, you can also chat with our pre-built agents, which are tailored for some of the most popular use cases in Web3.
Feel free to convey feedback or ask questions at info@spectrallabs.xyz or connect with us on our Discord server. #BuidlwithSyntax