https://github.com/luandro/wellness-bench
https://github.com/luandro/wellness-bench
Last synced: 5 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/luandro/wellness-bench
- Owner: luandro
- Created: 2025-12-20T10:39:18.000Z (6 months ago)
- Default Branch: main
- Last Pushed: 2025-12-29T14:13:54.000Z (6 months ago)
- Last Synced: 2026-01-01T17:16:46.993Z (6 months ago)
- Language: TypeScript
- Homepage: https://luandro.github.io/wellness-bench/
- Size: 881 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- Agents: AGENTS.md
Awesome Lists containing this project
README
# Wellness Bench
Evaluate how AI models reason about human and planetary well-being using a Buen Vivir (Sumak Kawsay) lens. The repo contains a React + Vite app for configuring questions, running evaluations, and reviewing results, plus a benchmark pipeline for batch runs.
## Quickstart
```sh
npm install
npm run dev
```
## Scripts
```sh
npm run dev # Start the Vite dev server
npm run build # Production build
npm run build:dev # Development-mode build
npm run test # Run tests with Vitest
npm run preview # Preview the production build
npm run lint # ESLint
npm run benchmark # Run benchmark pipeline
npm run benchmark:dry # Print benchmark plan
npm run benchmark:build # Type-check benchmark scripts
```
## App Features
- Manage benchmark questions, evaluation prompts, and providers
- Configure API keys (stored locally in the browser)
- Run evaluations and review results
- Import/export config and results bundles
## Benchmark Pipeline
Benchmark configs and providers live in `benchmarks/`. Outputs are written to `benchmarks/results/` (gitignored).
See [`benchmarks/README.md`](benchmarks/README.md) for detailed instructions on:
- Setting up the pipeline
- **Obtaining and configuring API keys** (OpenAI, Anthropic, Google, xAI, DeepSeek)
- Running the benchmark scripts
For a detailed comparison of the **Web UI** vs **CLI** workflows, see [**AGENTS.md**](AGENTS.md).
## Project Structure
- `src/` app entry and routes
- `src/pages/` route-level screens
- `src/components/` shared UI (shadcn primitives in `src/components/ui/`)
- `benchmarks/` pipeline scripts, configs, and providers
## Security Note
The UI stores API keys locally in browser storage (encrypted with Web Crypto when available). Treat this as a convenience for demos, not a production-grade secrets solution.