| name | macos-computer |
| description | Technical specifications and dev stack for this MacBook Air running the Vidext monorepo. Use when OS, hardware, or toolchain details are relevant to a task or needed to fix something. |
Host Environment: macOS 15.2 Sequoia
- Model: MacBook Air (Mac15,12)
- Chip: Apple M3 (ARM64) — 4 performance + 4 efficiency cores
- GPU: Apple M3 integrated — 10-core GPU, Metal 3
- RAM: 16 GB unified memory (shared CPU/GPU)
- Storage: 251 GB NVMe SSD (APFS) — ~117 GB free
- Kernel: Darwin 24.2.0 (xnu-11215.61.5, ARM64)
Critical Runtime Context
- Shell:
/bin/zsh
- Runtime: Bun 1.3.2 (primary), Node.js v25.9.0
- Package Manager: Bun 1.3.14 (pinned in
packageManager field) — never npm or pnpm
- System tools: Homebrew 5.1.9 (
brew) for system-level installs
- Default Python: 3.9.6 (
/usr/bin/python3, Xcode CLT) in the Vidext monorepo; standalone template projects may use conda through python-project-setup
- No CUDA: M3 uses Metal/MPS — use
torch.backends.mps for any ML tasks
Vidext Monorepo Stack
- Monorepo: Turborepo 2.9.12 + Bun workspaces
- Language: TypeScript 6.0.3 (strict mode)
- Framework: Next.js 16.2.6 (App Router) + React 19.2.6
- API: tRPC v11.17 + TanStack Query 5.100
- Validation: Effect Schema + Standard Schema adapters
- Database: PostgreSQL + Drizzle ORM 1.0 rc
- Auth: better-auth 1.6
- Workers: Cloudflare Workers + Wrangler 4.90
- AI/LLM: Vercel AI SDK 7, Cloudflare AI
- Styling: Tailwind CSS 4.3.0
- Linter/Formatter: Biome 2.4
- Storage: Cloudflare R2
- Cache: Upstash Redis + Ratelimit
- Analytics: PostHog
- Payments: Polar SDK
Key Commands
bun install
bun dev
bun build
bun typecheck
bun run typecheck:safety
bun format
bun format:fix
Instructions
- Always use
bun — never npm, npx, or pnpm.
- Shell scripts must target zsh on macOS (Darwin). No
apt — use brew.
- ARM64: verify package ARM compatibility before suggesting installs.
- In the Vidext monorepo, do not use conda; use
pip3 or python3 -m venv. For standalone Python template projects, load and apply python-project-setup instead.
- For ML/GPU: Apple MPS only (
torch.backends.mps), not CUDA.
- Run
bun typecheck, bun run typecheck:safety, and bun format as separate calls before claiming any change is complete.