What makes Graphext special?
A unique combination of accessibility, domain expertise, and scalable architecture.
Data Types
Smart Data Understanding
The UI adapts to your data schema. Each data type unlocks specific visualizations automatically.
Custom Libraries
World-Class Visualization
Render millions of nodes and hundreds of millions of edges. No other library comes close.
Frontend + Backend
Hybrid Architecture
Figma-like UX that scales. WebAssembly in browser, auto-scaling cloud for heavy ML.
Tech Stack
Full-Stack Platform
From intuitive UI to high-performance infrastructure, every layer designed for scale.
Recipe DSL
No Code / Low Code
Code when you want, click when you don't. Full reproducibility guaranteed.
Smart Data Understanding
The UI adapts to your data schema. Each data type unlocks specific visualizations and transformations.
Basic Types
Numeric
Integers, floats, percentages
Categorical
Discrete categories
Ordinal
Ordered categories
URLs
Web links, images
Dates
Timestamps, durations
Advanced Types
Texts
NLP, embeddings, sentiment
Lists
Multi-value fields
Links
Graph relationships
Adaptive Interface
When you import data, Graphext automatically detects column types and adjusts the available operations, visualizations, and models accordingly. Numeric columns get statistical tools, text columns get NLP features, and categorical columns get grouping operations.
World-Class Visualization Libraries
We built our own visualization engines because nothing else could handle the scale we needed.
Custom Graph Engine
Built from scratch using WebGL and GPU-accelerated rendering. No off-the-shelf library can match our performance at scale.
Enhanced Vega Charts
We've extended Vega with custom optimizations, smart defaults, and interactive features that make complex charts accessible to everyone.
Smart Plot Builder
Automatic chart recommendations based on your data types. From simple bar charts to complex multi-dimensional visualizations.
WebGL Acceleration
Every visualization leverages your GPU for buttery-smooth interactions, even with massive datasets that would crash other tools.
Create stunning visualizations from scatter plots to complex multi-dimensional charts. Customize every aspect and export publication-ready graphics.
Built on top of WebGL, Vega, and years of R&D
Hybrid Frontend / Backend
The best of both worlds: instant browser-side computation meets powerful cloud processing.
Light Steps
C++ / JavaScript
User's Browser
- Zero latency
- WebAssembly powered
- Infinite concurrent users
Heavy Steps
Docker Containers
Backend Pipeline
- Large datasets
- GPUs & Autoscaling
- Arrow format transport
Powered by
Full-Stack Data Platform
From intuitive UI to high-performance infrastructure, every layer is designed for scale.
Code When You Want, Click When You Don't
The power of code with the simplicity of visual tools.
Automatic Serialization
Every modification you make in the UI is automatically serialized to code. Full reproducibility and version control for your analytics workflows.
Reusable Recipes
Copy and paste recipes between projects. Share your analytics workflows with your team or use community-built templates.
Project Templates
Start similar projects instantly with templates. Churn analysis, lead scoring, customer segmentation - all ready to customize.
Simple & Constrained DSL
Our domain-specific language is intentionally simple and constrained. This means less room for errors and faster execution.
link_embeddings(ds.embedding, {
"metric": "euclidean",
"n_nearest": 15
}) -> (ds._gx_link_targets_, ds._gx_link_weights_)
cluster_network(ds._gx_link_targets_,
ds._gx_link_weights_,
{
"algorithm": "louvain",
"resolution": 1
}) => (ds.umap_cluster)
train_classification(ds[["Churn", "TotalCharges", ...]],
{
"target": "Churn",
"model": "CatboostClassifier",
"params": {
"depth": 6,
"nan_mode": "Min",
"iterations": 1000,
"one_hot_max_size": 10,
"max_ctr_complexity": 2,
"boosting_type": "Plain"
},
"validate": {
"n_splits": 5,
"metrics": ["accuracy"]
}
}) -> (ds.gx_prediction, "model_88808")