FAST DATA — AI-Native Data Analysis Platform

FastData is an AI-native data analysis platform with a ChatGPT-style conversational interface. Upload a CSV or Excel file and chat with your data — get instant profiling, visualizations, ML-powered predictions, domain-specific analytics, and comprehensive export packages, all within a single conversation.

---

Features

Category	Highlights
🤖 Chat Interface	Conversational UX with rich messages (text, charts, insight cards). History saved to localStorage.
📊 Data Profiling	Auto column typing, descriptive stats, quality scoring, outlier & duplicate detection.
🔗 Correlation Engine	Pairwise Pearson correlations, matrix views, human-readable interpretations.
💡 Insight Engine	7 insight categories scored by importance — overview, quality, risk, trends, and more.
📈 Chart Recommender	Automatic chart selection (bar, line, scatter, pie, histogram, box plot) with relevance scoring.
🧠 Client-Side ML	Pure TypeScript models (Ridge, kNN, Naive Bayes, Decision Stump) for regression & classification. Train, predict, and export Python loaders — no server required.
🏢 Domain Analytics	Auto-detect domain, then unlock specialized analysis for Finance, Crypto, Ecommerce, Medical, and Marketing.
🤝 AI Copilot	Ask natural-language questions. Uses OpenRouter (free models) via the backend, with a full rule-based fallback when offline.
🌙 Executive Narratives	Generate summaries in McKinsey, Bloomberg, or Consulting style through the AI Copilot.
📦 Export Suite	ZIP with reports, slides, charts (PNG), CSVs, JSON, Python ML loader, and PPTX presentations (night/day themes).
📋 Reports & Slides	Structured multi-section reports (Markdown + HTML) and auto-generated slide decks.
🔌 Offline Mode	All core analysis runs client-side. The backend is optional and used only for AI features.

---

Quick Start

# Install frontend dependencies
npm install
# (Optional) Install server dependenciescd server && npm install &&cd ..
# Start development (frontend + server)
npm run dev
# Frontend only
npm run dev:client
# Server only
npm run dev:server

Environment Variables (Server)

Copy server/.env and set:

Variable	Description	Default
PORT	Server port	3001
OPENROUTER_API_KEY	API key for AI Copilot	—
CORS_ORIGIN	Allowed origin	http://localhost:5173
MAX_FILE_SIZE	Upload limit	52428800 (50 MB)

No API key? No problem. The client-side analysis engine works fully offline. The AI Copilot fallback uses rule-based answers when no key is configured.

---

Usage

1. Start a new analysis — Click "New Analysis" in the sidebar.
2. Upload your data — Drop a .csv or .xlsx file, or use the built-in sample dataset.
3. Chat with your data — Ask questions like:
   • "Summarize this dataset"
   • "What's the data quality?"
   • "Show me correlations"
   • "Detect anomalies"
   • "What are the key insights?"
   • "Predict the target variable"
   • "Export everything"
4. Explore results — Charts, insights, and tables appear directly in the conversation.
5. Export — Download a comprehensive ZIP package with everything.

---

Domain-Specific Analytics

When you upload a file, FastData automatically detects the domain and unlocks specialized analytics:

Domain	Capabilities
💰 Financial	RSI, MACD, Bollinger Bands, drawdowns, Sharpe/Sortino/Calmar ratios, trend pattern detection
₿ Crypto	Wallet profiling, gas optimization, token metrics, staking rewards, portfolio P&L, rug-pull risk scoring
🛒 Ecommerce	RFM segmentation (7 customer segments), cohort retention, market basket analysis (association rules), CLV prediction
🏥 Medical	Kaplan-Meier survival curves, clinical distribution analysis, trial efficacy/safety metrics, diagnostic accuracy (sensitivity, specificity, PPV, NPV), patient clustering, biomarker analysis
📢 Marketing	Funnel analysis, attribution models, customer journey mapping, channel performance (ROAS, CPA, CTR), marketing mix modeling, CLV calculations

---

Client-Side Machine Learning

FastData includes a full ML pipeline implemented in pure TypeScript — no external ML libraries or server required.

• Regression models: MeanBaseline, Ridge Linear (gradient descent), kNN Regressor, Decision Stump
• Classification models: Majority Baseline, kNN Classifier, Gaussian Naive Bayes, Decision Stump
• Auto feature engineering: Z-score normalization, one-hot encoding, date features
• Auto feature selection: Drops text, high-null, and high-cardinality columns
• Train/test split with deterministic sampling
• Ensemble comparison — best model is selected automatically
• Prediction interface — input values and get predictions
• Export: Python .pkl-compatible model artifact, standalone Python loader script, performance reports

---

Architecture

FastData/
├── src/ # Frontend (React + TypeScript)
│ ├── components/ # 38 UI components (Chat, Charts, Dashboard, etc.)
│ ├── engine/ # Core analysis engine
│ │ ├── parser.ts # CSV / XLSX parsing
│ │ ├── profiler.ts # Column profiling & data quality
│ │ ├── correlations.ts # Pearson correlation engine
│ │ ├── insightGenerator.ts # Automatic insight generation
│ │ ├── chartRecommender.ts # Chart type & config recommender
│ │ ├── mlEngine.ts # Client-side ML (no deps)
│ │ ├── copilot.ts # Rule-based AI fallback
│ │ ├── aiEngine.ts # OpenRouter API integration
│ │ ├── reportGenerator.ts # Markdown & HTML reports
│ │ ├── slideGenerator.ts # Slide deck generation
│ │ ├── exportEngine.ts # ZIP export with all artifacts
│ │ ├── chartCapture.ts # ECharts PNG capture (DOM + offscreen)
│ │ ├── clientPptx.ts # Client-side PPTX generation
│ │ ├── premiumExport.ts # Premium HTML/JSON reports
│ │ ├── domainDetector.ts # Automatic domain detection
│ │ └── domains/ # Domain-specific analytics
│ │ ├── financial.ts
│ │ ├── crypto.ts
│ │ ├── ecommerce.ts
│ │ ├── medical.ts
│ │ └── marketing.ts
│ ├── services/ # API client & server adapter
│ ├── stores/ # Zustand state (app, conversations, interactions, theme)
│ ├── types/ # TypeScript type definitions
│ └── utils/ # Utility functions
├── server/ # Backend (Express + TypeScript)
│ ├── src/
│ │ ├── routes/ # upload, copilot, reports, ai
│ │ ├── services/ # analyzer, insight engine, AI orchestrator, etc.
│ │ └── config/ # Server configuration
│ └── package.json
└── scripts/ # Dev server launcher

---

Tech Stack

Layer	Technology
Framework	React 19 + TypeScript 6
Build	Vite 8
Styling	Tailwind CSS v4
Charts	Apache ECharts 6
Animation	Framer Motion 12
State	Zustand 5
Icons	Lucide React
Data Parsing	PapaParse, SheetJS (xlsx)
Export	JSZip, FileSaver, PptxGenJS
Server	Express 4, Multer, Bull (Redis)
AI	OpenRouter (multi-model)

---

API Endpoints

Base URL: http://localhost:3001/api

Upload

Method	Path	Description
POST	/upload	Upload CSV/XLSX
POST	/upload/:id/analyze	Run full analysis
GET	/upload/:id/insights	Get insights
GET	/upload/:id/anomalies	Get anomalies
GET	/upload/:id/correlations	Get correlations
GET	/upload/:id/charts	Get chart configs
GET	/upload/:id/report	Generate report
POST	/upload/:id/export	Export package
DELETE	/upload/:id	Delete dataset

AI Copilot

Method	Path	Description
POST	/copilot/query	Ask a question
POST	/copilot/explain/:insightId	Explain an insight
GET	/copilot/intents	List available intents

Reports

Method	Path	Description
GET	/reports/styles	List report styles
POST	/reports/generate	Generate report
POST	/reports/generate-pptx	Generate PowerPoint

Health

Method	Path	Description
GET	/health	Server status & AI config

---

License

MIT License. Built for the future of data analytics.