A TypeScript port of pandas, built from first principles using Autoloop — an automated research and experimentation platform that runs iterative optimization loops on GitHub Agentic Workflows.

🎮 Try the interactive playground →

TSB is named after the tsessebe (tseh-SEH-bee · IPA /tsɛˈsɛbi/) — a southern African antelope.

Project conventions

• Package name: tsb, e.g. import { DataFrame } from 'tsb'
• Runtime & tooling: Bun for everything — runtime, bundler, test runner, package manager
• Language: TypeScript in strictest mode — no any, no as casts, no @ts-ignore, no escape hatches
• Dependencies: Zero for core library. External deps only where absolutely required for non-core tooling (e.g. Playwright, WASM toolchains).
• Linting: Biome with all rules enabled, zero warnings tolerated
• Testing: 100% coverage — unit, property-based (fast-check), fuzz, and Playwright e2e for the web playground
• Build from scratch: Every pandas feature is implemented from first principles. No wrapping or porting existing JS/TS data libraries.

Goals

• Full feature parity with pandas — identical APIs adapted to TypeScript conventions and idiomatic structures. Every pandas feature is built from scratch, ground up, first principles. No ports of existing JS/TS data libraries.
• Interactive web playground — every feature ships with a rich, interactive tutorial, deployed to GitHub Pages. WASM where needed and useful.
• Performance — aggressive optimization throughout. Speed is a first-class concern.
• Exhaustive testing — pandas' own test suite as a baseline, extended with property-based testing, fuzzing, and e2e coverage. Target: 100%.