We're excited to share what we are building behind the scenes — a powerful, visual flow chart system that brings automation, AI, and tool integration together like never before. Whether you're a streamer, gamer, content creator, or just a curious power user — this is a tool that will change the way you work and interact with your PC.

Click to enlarge

VortiDeck Flow Builder — Visual, AI-Powered Automation

Flow Builder is VortiDeck's visual workflow designer. It turns complex automation into a drag-and-drop canvas where you connect nodes — triggers, AI decisions, tool executions, and control logic — into workflows that actually do things.

Think of it as building with intelligent LEGO blocks. You wire together what should happen, when it should happen, and let AI handle the parts that need to think.

What Flow Builder Does

At its core, Flow Builder lets you:

• Trigger workflows from button presses, webhooks, file changes, timers, or MQTT events

• Think using AI nodes that reason through decisions and adapt to context

• Execute tools — OBS controls, system scripts, API calls, cloud integrations, and more

• Chain everything together visually on a canvas with typed connections

Unlike traditional automation that follows rigid if/then rules, Flow Builder lets you combine the reliability of direct tool execution with the intelligence of AI decision-making — in the same workflow.

What Makes It Different

Dual execution architecture

Most automation tools give you one mode: either scripted steps or AI agents. VortiDeck gives you both, and lets you mix them:

• Direct tool execution — fast, predictable, zero latency. The node runs the tool exactly as configured.

• AI-coordinated execution — the node calls an AI model that decides which tools to use and in what order.

• Hybrid workflows — use direct execution for the critical path, AI for the decision points. Same flow, same canvas.

Per-node AI model selection

Every AI node in your flow can use a different model. Put GPT-3.5 on the quick classification step. Put GPT-4 on the complex reasoning step. Optimize each node independently for cost, speed, and intelligence.

Supported models:

• GPT-3.5 Turbo — fast and affordable

• GPT-4 / GPT-4 Turbo — for complex reasoning

• Any OpenAI-compatible API

• Self-hosted models via Ollama and LocalAI (coming soon)

• Anthropic Claude, Google Gemini (planned)

[callout variant="tip" title="Cost optimization"] Most workflows don't need GPT-4 on every step. By mixing models per node, you can cut API costs dramatically while keeping intelligence where it matters.

AI augments, doesn't replace

Flow Builder follows a simple philosophy: you design the structure, AI handles the dynamic decisions, tools execute the critical actions, and you stay in control. AI is a tool in your workflow, not the workflow itself.

The Building Blocks

Trigger nodes

Every flow starts with a trigger — the event that kicks things off:

• Webhook — receive data from external services

• Button — press a deck button to start the flow

• Timer — run on a schedule (cron or interval)

• File System — react to file changes, creations, or deletions

• MQTT — subscribe to IoT message topics

AI nodes

These are where intelligence lives in your flow:

• AI Agent — fully autonomous; give it a goal and tools, it figures out the rest

• AI Orchestrator — coordinates multiple tools in sequence based on context

• AI Text Generator — pure text generation for content, summaries, responses

• AI Strict Executor (coming soon) — constrained execution that follows exact tool schemas

Tool nodes

The things that actually get done:

System tools — audio control, file operations, HTTP requests, system commands, data processing

MCP tools (extensible via Model Context Protocol) — OBS Studio, Discord, Twitch, Microsoft To Do, cloud services (AWS, Azure, GCP), databases, and custom integrations

Control flow nodes

Logic that shapes how your workflow runs:

• Delay — wait before continuing

• Branch — conditional paths based on data

• Variables — store and pass data between nodes

• Loops — repeat operations over collections

How It Works Under the Hood

The execution pipeline runs through three layers:

graph TD
    A[Frontend - Vue.js] --> B[Tauri Commands]
    B --> C[Flow Executor - Rust]
    C --> D[Direct Tool Execution]
    C --> E[AI-Coordinated Execution]
    D --> F[Unified Tool Executor]
    E --> F
    F --> G[System Tools]
    F --> H[MCP Tools]
    F --> I[Custom Integrations]

The Rust-based executor handles scheduling, parallel execution, error recovery, and real-time status updates. Both execution modes feed into the same unified tool system — so an AI node and a direct node can call the exact same tools.

Real-World Examples

Streamer go-live automation

Press "Go Live" → AI Orchestrator checks your setup → switches OBS to starting scene → posts in Discord → waits 30 seconds → transitions to main scene → starts intro music.

One button. Everything happens.

Smart content pipeline

A new file appears in your uploads folder → AI Agent analyzes the content → Text Generator writes a description → API node uploads to your platform → Discord notification goes out to your team.

System monitoring

Timer fires every 5 minutes → AI Orchestrator runs system diagnostics → Branch node checks if anything's wrong → if yes: cleanup script + email alert → if no: log entry and move on.

Business process

Webhook receives a customer request → AI classifies the type → Branch routes to the right handler → task gets created in your project tool → confirmation email sent → team gets notified.

[callout variant="info" title="Start simple, add AI later"] You don't need AI in every flow. Start with direct tool execution — triggers, actions, branches. Add AI nodes only where you need dynamic decision-making. The best workflows use AI surgically, not everywhere.

Current Status

Shipped

• Visual flow editor with drag-and-drop canvas

• Dual execution system (direct + AI-coordinated)

• Per-node AI model selection

• Unified tool system with MCP integration

• Real-time execution monitoring

• Type-safe port connections

• Multi-node selection and bulk operations

In development

• Right-click context menus and snap-to-grid

• Visual debugger and execution replay

• Minimap and canvas navigation

• Node grouping and sub-flows

• Flow templates and import/export

• Parallel execution logic

[release_summary title="Roadmap"]

• Cloud hosting & public webhooks | Coming next

• Self-hosted AI (Ollama, LocalAI) | Coming next

• Flow versioning & visual diffs | Coming next

• Custom Node SDK | Coming next

• Multi-instance sync | Coming next

• Enterprise SSO & RBAC | Coming next

• Flow marketplace | Coming next

• IoT & voice triggers | Coming next

• A2A protocol support | Coming next

• CI/CD integration | Coming next

Who Is This For

• Streamers and creators — automate your entire broadcast workflow

• Developers — build complex integrations without writing glue code

• Business teams — connect your tools with intelligent routing

• Smart home users — orchestrate devices with context-aware logic

• Anyone who automates — if you've outgrown simple if/then tools, this is the next step

Getting Started

Start with a trigger and an action. Connect them. Run the flow. Then add complexity as you need it — a branch here, an AI node there, a loop for batch processing. The canvas grows with your ideas.

The future of automation isn't choosing between smart and reliable. It's both, on the same canvas, in the same workflow.

That's Flow Builder.