Documentation

Everything you need to build and run TypeScript applications on the GPU — from your first sym server.ts to tuning the CUDA scheduler.

⌕

Searches every page, every API function, and every config option. Press / to focus.

Start here

Install Symmetra, verify your GPU, and run your first TypeScript-on-GPU server in minutes.

Which language features compile to GPU code — variables, loops, arrays, async, classes — explained in plain terms.

The honest list: patterns the compiler doesn't support yet, and the verified patterns to use instead.

All 8 packages — http, core, kv, db, fs, crypto, env, net. Every function, its parameters, and what it returns.

Every serve() option with complete working examples: ports, CUDA streams, batching, VRAM slabs, scheduler tuning.

How the pipeline works: esbuild → AST → CUDA codegen → NVRTC JIT → persistent GPU kernel.

7 publicly-known Turing machines, measured identically on an NVIDIA L4 — correctness notes, endpoint-by-endpoint RPS, and steps/sec for each one.

Browse the Turing machine engines — downloadable Symmetra applications with measured benchmarks.

Browse every @symmetra package — versions, exports, install commands, and links into the API reference.

Create topics, ask questions, and share what you're building — this is the community chat.

Tell us what's confusing, broken, or missing. Beta feedback goes straight to the team.