Skip to content

DashForge

A JSON-first dashboard framework that starts simple with static hosting and grows into a full Metabase-like analytics platform with an AI-powered visual builder.

Why DashForge?

  • Visual Dashboard Builder - Drag-and-drop editor optimized for both humans and LLM agents
  • Saved Questions - Metabase-style GrokifyQL questions that can be run, formatted, exported, and reused across dashboards
  • JSON IR First - Dashboards are defined in a non-polymorphic, AI-friendly JSON format that's easy to generate, validate, and version control
  • Cube.js Semantic Layer - Business-friendly queries with pre-built relationships and AI context
  • Static to Dynamic - Start with static file hosting (GitHub Pages), graduate to a full PostgreSQL-backed server when you need it
  • Multi-tenant Ready - Built-in Row Level Security for SaaS deployments
  • SSO Authentication - GitHub and Google OAuth out of the box
  • ChartIR Integration - Uses echartify for declarative chart definitions

Three Modes of Operation

Builder Mode

Visual drag-and-drop dashboard creation:

  • 12-column responsive grid layout
  • Widget palette with charts, metrics, tables, and text
  • Real-time chart preview with ECharts
  • Cube.js query builder for semantic queries
  • AI-powered widget generation from natural language
  • Question builder with formatted SQL display, edit mode, results preview, CSV/XLSX export, and field value browsing
cd builder && npm install && npm run build && cd ..
./dashforge-server serve --port 8080
open http://localhost:8080/builder/

Static Mode

Perfect for:

  • Documentation sites
  • GitHub Pages hosting
  • Simple data visualization
  • No backend required
# Just open in a browser
open viewer/index.html?dashboard=./examples/dashboard.json

The viewer includes a light/dark theme toggle in the header (persisted via localStorage, defaulting to the OS's prefers-color-scheme) and formats numeric values — including a "currency" metric format — to at most 2 decimal places in tooltips, axis labels, and metric tiles.

Server Mode

Full-featured analytics platform with:

  • PostgreSQL data sources
  • User authentication (OAuth)
  • Multi-tenant isolation
  • Dashboard CRUD API
  • Saved Questions
  • Optional OmniAgent marketplace integration for reusable analytics agents
./dashforge-server serve --database-url postgres://... --auto-migrate

Architecture Overview

┌──────────────────────────────────────────────────────────────────┐
│                        DashForge                                   │
├──────────────────────────────────────────────────────────────────┤
│  Builder Mode            │  Static Mode     │  Server Mode       │
│  ────────────            │  ───────────     │  ───────────       │
│  /builder/               │  viewer/         │  dashforge-server    │
│  React + TypeScript      │  index.html      │  PostgreSQL + Ent  │
│  Questions + dashboards  │  JSON files      │  OAuth             │
│  AI generation           │  GitHub Pages    │  Multi-tenant RLS  │
├──────────────────────────────────────────────────────────────────┤
│                    Dashboard IR (JSON)                           │
│                    ChartIR (echartify)                           │
│                    Cube.js Semantic Layer                        │
└──────────────────────────────────────────────────────────────────┘

AI-First Design

DashForge is designed from the ground up to work seamlessly with LLM agents:

  • Non-polymorphic JSON - Dashboard IR uses flat structures without oneOf/anyOf, making it easy for AI to generate valid configurations
  • Semantic context - Cube.js provides business-friendly names like "revenue" and "customer_count" instead of raw column names
  • Validated generation - JSON Schema validation ensures AI-generated output is always valid
  • Dual rendering - Same ChartIR spec works in the dashboard builder and chatbot responses
User: "Add a line chart showing revenue trends"
  ↓
AI generates ChartIR JSON
  ↓
Validation passes
  ↓
Widget added to dashboard