OpenParable
Skip to content
← Back to skills

Turbo Benchmark

Testing & Quality
turbo-benchmark

Turbo Benchmark is an Agent Skill for testing & quality published in the justrach/turboAPI repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Run performance benchmarks for TurboAPI. Use when testing performance, checking for regressions, or comparing against FastAPI.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/justrach/turboAPI/blob/HEAD/.claude/skills/turbo-benchmark/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/turbo-benchmark, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Run TurboAPI Benchmarks

Steps

  1. Determine benchmark type: $ARGUMENTS[0] — http (default), db, full, or all
  2. Build turbonet: Run uv run --python 3.14t python zig/build_turbonet.py --install --release
  3. Run the benchmark

HTTP benchmark (TurboAPI vs FastAPI)

uv run --python 3.14t python benchmarks/turboapi_vs_fastapi.py --duration 10 --threads 4 --connections 100

DB benchmark (pg.zig vs SQLAlchemy)

docker run -d --name bench-pg -p 5432:5432 \
  -e POSTGRES_USER=turbo -e POSTGRES_PASSWORD=turbo -e POSTGRES_DB=turbotest \
  postgres:18-alpine
sleep 3
# Seed, then:
uv run --python 3.14t python benchmarks/db_bench_ci.py
docker stop bench-pg && docker rm bench-pg

Full comparison (all query patterns)

# Needs Postgres running on :5432
uv run --python 3.14t python benchmarks/full_comparison.py

Tests: baseline, SELECT by PK, paginated list, full-text search, GROUP BY, and more.

Varying IDs benchmark (cache spread test)

wrk -t2 -c50 -d5s -s benchmarks/varying_ids.lua http://127.0.0.1:8000

Expected numbers (Apple Silicon M3 Pro)

EndpointTurboAPIFastAPI+SQLAlchemySpeedup
Baseline (no DB)~133k req/s~9.8k req/s14x
SELECT by PK (cached)~133k req/s~2.6k req/s48x
Full-text search~110k req/s~2.5k req/s44x
GROUP BY sum/avg~129k req/s~1.4k req/s96x
Auth + Depends~116k req/s——
Varying IDs (100 keys)~126k req/s——

Regression threshold

CI fails if cached PK drops below 10k req/s. See .github/workflows/db-benchmark.yml.