How do you approach ideas, and how much do you stick to them once you start following them? When do you pivot to a new direction, turning the flow of previously shiny ideas in a new direction or wrapping things up under a new umbrella when they outgrow the old one?
These are questions that often challenge Web2 startup founders and, even more so, Web3. FOMO, volatility, and trends are constantly shifting—bull cycles, winters, and who knows where we are now. We started as a Multi-Chain/Multi-Dex DeFi trading terminal when Uniswap V2 was just launching and BNB/Polygon were emerging. We were the first to add trader categorizations, trends, token and trader profiles, and trading within the same application. But UST, FTX, and winter came, and retail, along with 90% of trading volume disappeared.
At DexGuru, we cut costs (link to infra) and re-evaluated the market. Over two years of developing the DexGuru Trading Terminal, our ETL pipeline technology matured enough to power a block explorer product, which we developed exactly a year ago. Our first Block Explorer deployment was for the Canto chain (https://canto.dex.guru) and has since expanded across 18 chains (https://b2b.dex.guru/explorer). These were business development sales. We organized a lead funnel to get there and improved our offering with Guru Data Warehouse.
We started as a data product MVP and developed from the same foundation as Dune Analytics (https://github.com/getredash/redash.git). As sales showed, we pivoted Data Analytics towards a dApp development toolset, allowing users to create endpoints out of feature-rich indexed datasets, create visualizations and dashboards, and embed them into applications using shared components and do data science research using our Jupyter Hub Data Warehouse component.
On this journey, we identified opportunities to embed new GPT integrations into block explorer charts as Web3 data-aware GPT agents and into the Query Composer as AI SQL Assist. To orchestrate these multi-step AI processes, we incorporated a Business Process Automation (BPA) Engine into our internal Development Platform, enabling quick modifications and allowing us to query a vector database of pre-indexed context embeddings, data requests, and researcher actions via the engine user’s task list. Initially tested during our AI POC implementation, this internal orchestration platform has increasingly become central to our architecture and processes, enabling the orchestration of On-Chain, Off-Chain, Compute, and User actions.
In our platform, the system acts as a state machine for frontends (Next.js, React, Aiogram (Telegram bot)) with processes defined as definitions storing the execution context of each user’s quest in process instances. This Project Development architecture, where all data with views is sourced from the Guru Data Warehouse and execution is orchestrated by the BPA Engine, has proven effective multiple times, allowing us to release five comprehensive B2B and retail products within a year, adapting to market trends.
As we evaluated using the BPA Engine (https://camunda.com) to orchestrate mainly Web2/AI Processor flows across our frontends, we decided to add more Web3 elements and participated in the ETHOnline 2023 Hackathon in November 2023, presenting Chainflow. Chainflow was introduced as a platform for crowdfunding campaigns, where participants are predefined in a diagram and can be locked out of collected funds if they don’t complete user tasks issued after the crowdfunding campaign reaches its goal.
Chainflow's Web3 stack was later integrated into the Data Warehouse to handle our points system, encouraging the builder community to publish queries and dashboards in block explorer charts and stats catalog. By February 2024, our team had a rich toolkit, allowing us to actively develop five products simultaneously within a lean team.
Starting with an architecture where each Guru Native orchestration had its own engine, we noticed that merging them under one roof made sense as they were all sourced from the same Data Warehouse. Instead of merging engines, we introduced a Middleware synchronization layer, working off an Event Bus shared through a Message Queue. Each engine posts and reads from the same queue, effectively implementing broadcasting mechanisms for state sync. We then explored utilizing Blockchain for this purpose, publishing events as inscriptions.
This mechanism has been recently implemented as we migrating dex.guru to the new architecture, releasing it as Dex Guru V2, focused on Web3 actions automation. Every action orchestrated by the engine leaves a trace as an inscription On-Chain, which can be re-evaluated and synced to the current state of the process instance using inscriptions indexing and ingestion into the engine’s context.
This setup allows us to use compute resources and definitions across engines, leading to the idea of a Micro Franchise Economy (MFE) powered by the Guru Network. MFE leverages a diverse ecosystem of contributors and users, each playing a crucial role in the network’s operation and growth.
These participants include:
Gurus: AI Processors, Compute Node Runners, and Individual Agents code contributors.
Ecosystem Projects: Retail and business projects combining multiple BBPA orchestrations and data scenarios to solve routine tasks or business functions, also acting as Gurus by contributing AI Processors and GURU AI Assistants to the network.
Individual Agents: AI/Non-Custodial Web3 Processors run by Compute nodes, used by BBPM defined processes.
Integrators: Provide B2B adoption of the Guru AI Flow Orchestrator and Framework for enterprise implementations.
Retail Users: Utilize the network as assistants from their wallets, with the option to participate in "Compute by Earn" by running Individual Agents when idle.
Phase 1 Testnet is crucial to syncing the state of dApp engine states, including native Guru ones. We are starting by spinning it up and deploying all infrastructure there.
Our first action is migrating the Guru applications stack:
Private Beta release of V2(Auto), resembling Dex Guru Bloomberg integration with a focus on DeFi actions automation and AI integrations.
Guru Data Warehouse becomes Guru Flow: It Includes Data Analytics actions, a BBPA definitions modeler, and a Catalog, along with a monitoring section displaying process stats synced through Guru Network.
Agents Release: As we migrate applications, all agents used internally will be published in Guru Routines Agents.
We also have ecosystem projects migrating to Guru Phase 1 Testnet, including https://pixelpact.ai and https://piccraft.xyz, sharing a BPA definition for producing images using AI models and syncing through Guru Network.
We are moving DexGuru products to Guru Network. Since early 2024, we have been developing DexGuru V2, featuring Web3 Automation functions. DexGuru V2(Auto) will launch as a closed beta for Guru NFT Season Pass Holders:
Traditional DexGuru analytics with a new expandable navigation menu will be fed from the Guru Data Warehouse.
Data reports are reorganized for cross-relation search and DYOR routines. Integrated with the BBPA Engine in Guru Phase 1 Testnet, new alerts, notifications, and a community-driven Automation Snippets catalog will be released. Integration with the Data Warehouse points system and tGURU (Phase 1 Testnet Native) provides a sandbox for early adopters.
We are working with a curated community of Builders, DeFi enthusiasts, and power DexGuru users. Learn about the architecture and use cases of Guru Network, Flow, and Framework:
Framework Github: https://github.com/dex-guru/guru-framework.git
Website: https://gurunetwork.ai/
Feel free to subscribe/DM me at https://x.com/evahteev.
We are actively building, and next steps are:
Guru Network Phase 1 Testnet Launch
DexGuru V2(Auto) Private Beta Release: https://v2.dex.guru
Flow Orchestrator GUI Release: https://flow.dex.guru
Guru NFT Season Pass V3: Invites mechanism open
Pixelpact and Picscraft Release on Guru Phase 1 Testnet
Keep yourself updated and subscribe to: