AI Dashboard Generation Platform

Enterprise-grade AI-powered BI platform. Transforms raw CSV / XLSX / JSON datasets into interactive, audience-segmented dashboards via a 5-node AI pipeline.

Architecture

Upload -> Backend API -> n8n Workflow -> AI Nodes (x5) -> Dashboard JSON Spec -> Frontend Renderer

AI generates specifications only. Frontend renders all charts, filters, and interactions.

Tech Stack

Layer	Technology
Frontend	Next.js 14, TailwindCSS, ECharts
Backend	Node.js, Express, TypeScript
Workflow	n8n
AI	OpenRouter (Claude-3.5-Sonnet)
Storage	SQLite (better-sqlite3)
File Parse	PapaParse (CSV), xlsx (Excel)

Folder Structure

BE/
├── backend/
│   └── src/
│       ├── index.ts              Express server entry
│       ├── routes/               upload, dashboard, export
│       ├── services/             fileParser, n8nOrchestrator, validator
│       ├── models/               SQLite database + queries
│       ├── middleware/           Multer file upload
│       └── types/                TypeScript interfaces
├── frontend/
│   └── src/
│       ├── app/                  Upload page + Dashboard page
│       ├── components/           KPIGrid, ChartRenderer, AudienceTabs,
│       │                         GlobalFilters, Sidebar, ExportButton
│       ├── lib/                  API client, ECharts adapters, types
│       └── store/                Zustand state management
├── n8n/
│   └── workflows/
│       └── dashboard-gen.json   Import this into n8n
├── ai-prompts/                   Optimized prompts per AI node
├── schemas/
│   └── dashboard.schema.json    Full JSON Schema for dashboard spec
├── docker-compose.yml
└── .env.example

Quick Start

1. Prerequisites

• Node.js 18+
• n8n (self-hosted or cloud)
• OpenRouter API key

2. Environment

cp .env.example .env
# Fill in OPENROUTER_API_KEY and N8N_WEBHOOK_URL

3. Backend

cd backend
npm install
npm run dev        # http://localhost:4000

4. Frontend

cd frontend
npm install
npm run dev        # http://localhost:3000

5. n8n Workflow

• Import n8n/workflows/dashboard-gen.json into n8n
• Activate the workflow
• Copy the Webhook URL -> paste into .env as N8N_WEBHOOK_URL

6. Docker All-in-One

docker-compose up --build

API Endpoints

Method	Endpoint	Description
POST	/api/upload	Upload CSV / XLSX / JSON
POST	/api/dashboard/generate	Trigger AI pipeline
GET	/api/dashboard/:id	Poll status + get spec
GET	/api/export/:id	Download spec as JSON

5-Node AI Pipeline

Node 1  Input Understanding   ->  dimensions, measures, time fields, quality score
Node 2  Insight Generator     ->  KPIs, trends, comparisons, anomalies (no hallucinations)
Node 3  Visualization Planner ->  chart type mapping, filter suggestions
Node 4  Audience Mapper       ->  4 tabs with hard visual count limits
Node 5  Layout Engine         ->  12-column responsive grid positions
Validator (backend)           ->  dedup KPIs, remove invalid fields, compute QA score

Audience Tabs

Tab	Focus	Max Charts
C-Suite	Strategic KPIs, growth	6
Top Management	Summary, department overview	8
Middle Management	Trends, team comparisons	10
End Users	Operational, granular detail	12

Future Extensions (designed for)

• Power BI / Tableau spec export
• AI narrative generation
• Forecasting node
• SQL / DAX generation
• Role-based auth + multi-user workspaces
• RAG business glossary