// the find
justrach/turboAPI
FastAPI-compatible Python framework with Zig HTTP core; 7x faster, free-threading native
A FastAPI-compatible web framework where the HTTP layer is written in Zig instead of relying on uvicorn/ASGI, aimed at people who want to keep a Python/FastAPI codebase but need more raw throughput. It targets Python 3.14's free-threaded build specifically, so it's for teams already committed to (or experimenting with) no-GIL Python, not a general FastAPI swap-in yet.
The dispatch classification (native_ffi / simple_sync_noargs / model_sync / enhanced) is a real design decision, not just marketing — routes get the cheapest possible path based on their actual signature, and JSON validation happens in Zig before the GIL is even touched for POST bodies. The zero-copy response path (writing directly from PyUnicode's internal buffer) and the extraction of the router/HTTP core into a separate turboapi-core library shared with a sister project (merjs) both suggest real engineering discipline rather than a one-off perf demo. 380+ tests plus a fuzzed HTTP parser and router is more test coverage than most alpha frameworks bother with.
It's alpha and it shows: no native TLS/HTTP2/HTTP3, a hardcoded 16MB body cap with no config knob, and WebSocket support that's exact-route-match only. Three consecutive patch releases (v1.0.25–v1.0.27) exist purely to fix broken artifacts from the previous release, which is a bad sign for whoever pins a version in production. The benchmark story is also doing a lot of work for you — the headline HTTP-only numbers are real, but the end-to-end DB benchmarks compare a Zig-native Postgres driver (pg.zig) against Python's asyncpg/SQLAlchemy, which is really a driver benchmark wearing a framework benchmark's clothes. And requiring the free-threaded Python 3.14t build means most of the ecosystem's C extensions are an unknown quantity for thread-safety, which the README itself admits.