product
DBCV — Open AI Workflow Platform
An open multi-agent low-code platform: a visual scenario builder, runtime and AI assistant — one of the things we've built.
DBCV is one of the things we’ve built: an open AI workflow and multi-agent low-code platform where workflows are built and run visually. Not a model wrapper — the visual builder, runtime and AI assistant sit on an architecture grounded in a patented, entropy-resistant data-separation method. The workflow is assembled from blocks and links, and the LLM and AI agents are wired in as individual steps — only where they are actually needed.
The core engineering idea: the model is wired in only where it is needed. Logic, branching, data work and integrations can be built with no LLM at all, so token cost is controlled at the scenario-architecture level, not discovered after the fact on the invoice. Below — capabilities, integrations, video and the interface.
capabilities
Visual scenario builder
Design and debug UX scenarios in a visual editor: branching from message triggers, Python code, external requests and variable values.
No-LLM steps — token economy
Logic, branching, data work and integrations can be built without calling a language model at all. The model is wired in only where it is truly needed — direct cost control.
Multi-agent
Service-agents for chatbots, games, sites, automation and cross-system integrations. Services combine into a multi-channel environment, even without APIs.
AI assistant: Ask · Plan · Agent
Three modes — Q&A, planning and autonomous task execution — with transparency: you see when it thinks, calls tools, and the pipeline stage.
Flexibility and branching
A scenario is assembled from blocks and links with master-links and transition rules; behavior changes without rewriting the whole process.
Logs, history and sandbox
Detailed run history, before/after states per step, built-in logs inside agents, and a sandbox mode with diagnostics and no external tools.
integrations
Integration with the popular messenger for bot development
The largest news portal, available via API
GPT-4.1 for generating text, code and analytics
Online service converting Russian speech audio to text
Create, remind and share events
Calendar for planning and reminders
Cloud spreadsheets with formulas and scripts
Autofill and validation of addresses and phone numbers
Yandex's Russian LLM for text, code and analytics generation
The Claude family for reasoning, analysis and code
Messenger for business bots and conversation automation
Enterprise messenger for work communications and notifications
Knowledge base and collaborative documents and tables
Russian social network and messenger for bots and integrations
Communities and bots on voice and text channels
Repositories, issues and pull requests for DevOps scenarios
Sberbank's Russian LLM for text generation and knowledge work in Russian
Russian speech recognition and synthesis from Yandex for voice scenarios
Russian CRM and corporate portal for sales and process automation
Sales CRM with deal pipelines and communication automation
Russian messenger from VK Tech for business correspondence and bots
A universal HTTP connector to integrate with any external service via API
+ any external API
video
interface
patent
RU Patent No. 2 688 238 C1
Data separation method “Distributor — Block — Link — Presentation” for web services
The platform architecture is based on a patented, entropy-resistant data separation method. Patent holder — Philippe Delsaul.
deployment
Free cloud
Non-commercial cloud platform.eni7.ru: the visual builder with no separate hosting or DevOps.
Open source
Python source on Gitverse (carbonfay/DBCV): deploy, study, extend.
Portability
Export scenarios to JSON — move between cloud and local servers.
Materials, screenshots and video: dbcv.ru
related