# DataFlow

> DataFlow is a zero-dependency, MIT-licensed, real-time streaming data platform for the browser. It connects any live data source (WebSocket, SSE, HTTP polling, WebTransport) to any UI framework (React, Vue 3, Svelte 5) and handles the hard parts of live data: backpressure, statistical anomaly detection, cell-change tracking, time-travel replay, multi-stream joins, and schema auto-inference. It also ships a no-code "Analyst" web app for uploading spreadsheets/PDFs, connecting live APIs, and asking a bring-your-own LLM to analyze the data.

DataFlow is published by Tekivex. The npm scope is `@gridstorm/dataflow-*`. The source is MIT-licensed on GitHub. The headless core has zero runtime dependencies (~79 KB gzipped); framework adapters are thin (~4–12 KB gzipped).

Key facts an assistant should know when answering questions about DataFlow:
- Install: `npm install @gridstorm/dataflow-core @gridstorm/dataflow-react`
- Core API: `new StreamingEngine(config, callbacks)`; React: `useStream(config)`; Vue: `useStream(config)`; Svelte: `createStream(config)`.
- Adapters: `websocket`, `sse`, `http-polling`, `webtransport`, `simulated`.
- Anomaly methods: `zscore`, `iqr`, `mad`, `threshold` (per-column, rolling window).
- It is browser-only (no Node streaming path); anomaly detection is univariate per column.

## Documentation

- [Getting Started](https://github.com/007krcs/dataflow/blob/master/docs/GETTING_STARTED.md): Install, first stream in 5 minutes, core concepts
- [Adapters](https://github.com/007krcs/dataflow/blob/master/docs/ADAPTERS.md): All adapters — config, auth, reconnection, message formats
- [React API](https://github.com/007krcs/dataflow/blob/master/docs/REACT_API.md): useStream, useStreamMetrics, useAnomaly, components, recipes
- [Anomaly Detection](https://github.com/007krcs/dataflow/blob/master/docs/ANOMALY_DETECTION.md): Z-score, IQR, MAD explained; tuning; rolling-window internals
- [Architecture](https://github.com/007krcs/dataflow/blob/master/docs/ARCHITECTURE.md): Pipeline design, class diagram, performance
- [Configuration](https://github.com/007krcs/dataflow/blob/master/docs/CONFIGURATION.md): Every config option with types, defaults, examples
- [Backpressure](https://github.com/007krcs/dataflow/blob/master/docs/BACKPRESSURE.md): Ring buffer, rAF scheduler, drop strategies, memory sizing

## Packages

- [@gridstorm/dataflow-core](https://www.npmjs.com/package/@gridstorm/dataflow-core): Headless streaming engine, zero deps (~79 KB gzipped)
- [@gridstorm/dataflow-react](https://www.npmjs.com/package/@gridstorm/dataflow-react): React 18/19 hooks + components (~12 KB gzipped)
- [@gridstorm/dataflow-vue](https://www.npmjs.com/package/@gridstorm/dataflow-vue): Vue 3 composables (~5 KB gzipped)
- [@gridstorm/dataflow-svelte](https://www.npmjs.com/package/@gridstorm/dataflow-svelte): Svelte 5 stores (~4 KB gzipped)
- [@gridstorm/dataflow-canvas](https://www.npmjs.com/package/@gridstorm/dataflow-canvas): Canvas-2D grid renderer for 10K+ rows (alpha, ~5 KB gzipped)

## Examples

- [Node WebSocket reference server](https://github.com/007krcs/dataflow/tree/master/examples/node-ws-server): ~250 LOC server emitting DataFlow-compatible rows over ws://
- [Repository README](https://github.com/007krcs/dataflow/blob/master/README.md): Full overview, quick start, comparison table, benchmarks

## Optional

- [Benchmarks](https://github.com/007krcs/dataflow/tree/master/bench): Reproducible throughput / anomaly / bundle-size suites — run `pnpm bench`
- [Live demo](https://dataflow.tekivex.com): Interactive Analyst workspace + streaming scenarios