Inspired by Ender Wiggin’s realization that changed the rules of combat, ** Gravity UI** reimagines how users interact with complex data workflows. Rather than the traditional horizontal or form-based pipeline builder, this system visualizes data as something that flows downward—falling from raw input to structured insight.
This metaphor makes data flow natural, interpretable, and scalable. It also enables composable, modular workflows with real-time dependency propagation—unlocking arbitrarily complex chains of transformation, model application, and publishing in an intuitive visual interface.
+-----------------------------------------------------------+
| 🛠️ Top Toolbar |
| [New] [Save] [Load] [Export JSON] [Add Node ▼] |
+-----------------------------------------------------------+
| |
| ⬇️ Scrollable Gravity Canvas |
| |
| 🌐 Source: MySQL Table |
| ↓ |
| 💬 Node: Sentiment Classifier (LLM) |
| ↓ |
| 🔧 Node: Join with Survey CSV |
| ↓ |
| 🧠 Node: ANOVA |
| ↓ |
| ✅ Assertion: p < 0.05 |
| ↓ |
| 📤 Publish: Dashboard JSON Export |
| |
+-----------------------------------------------------------+
| 📊 Status Bar (staleness, warnings, system alerts) |
+-----------------------------------------------------------+
Each node is a modular unit that:
Type | Icon | Description |
---|---|---|
🌐 Source | CSV, SQL, IPFS, JSON URL, etc. | |
🔧 Transform | Join, filter, normalize, encode, derive | |
💬 Classifier | LLM-driven node (e.g., sentiment analysis) | |
🧠 Model | Regression, clustering, PCA, etc. | |
✅ Assert | Tests (e.g., “no nulls”, “R² > 0.7”) | |
📤 Output | CSV, JSON, IPFS, webhook, dashboard hook |
List<Node> inputs
List<Node> outputs
stale
run()
Visible when a node is selected.
Feature | Why It Matters |
---|---|
Vertical orientation | Reinforces intuitive sense of data flow |
Node previews | Helps users debug and validate logic in-place |
Assertion nodes | Prevents bad data from publishing |
Config sidebars | Reduces clutter on canvas |
Live reactivity | Keeps results in sync with upstream data |
Grouping | Manages complexity in large graphs |
“What if data flowed like water, and insight fell into place?”
Welcome to Critique.ai—the gravity-powered, drag-and-drop intelligence canvas. Inspired by Ender’s Game, ComfyUI, and the open-ended power of Factorio, Critique.ai lets you visualize, build, and publish machine learning pipelines with zero clutter and maximum control.
id
, type
, config
, position
, input_ids
income_level
)This document describes the full rationale and implementation plan for a gravity-based, vertical ML pipeline UI inspired by ComfyUI and Ender’s Game. It covers interaction design, execution models, node structures, visual metaphors, and marketing tone. The enemy’s gate is down—data doesn’t have to be trapped in flat spreadsheets or black-box clouds. Let it fall, flow, and flourish.