scitex-benchmark

Performance benchmarking, runtime monitoring, and profiling helpers.

Full Documentation · uv pip install scitex-benchmark[all]

Installation

pip install scitex-benchmark

Architecture

src/scitex_benchmark/
├── benchmark.py     # high-level benchmark() runner + statistical comparisons
├── profiler.py      # Profiler context manager (wall-clock + memory)
├── monitor.py       # long-running resource monitor (CPU/RAM/GPU)
└── _skills/         # SciTeX skills metadata

Demo

flowchart LR
    Code[user code] --> P[Profiler ctx]
    Code --> B[benchmark - repeated runs]
    Code --> M[Monitor - long runs]
    P --> Stats[wall-clock + RSS]
    B --> Stats
    M --> Stats
    Stats --> Report[summary - DataFrame / CSV]

Quick Start

import scitex_benchmark as bm

# Quick wall-clock + memory snapshot
with bm.Profiler() as p:
    work()
print(p.summary())

1 Interfaces

import scitex_benchmark as bm

# Benchmark suite — time/memory across input sizes
suite = bm.BenchmarkSuite("io")
suite.add_benchmark(my_func, gen_input, "name", sizes=["1MB", "10MB"])
results = suite.run()

# Runtime monitor — alerts when CPU/RAM/disk thresholds breached
monitor = bm.RuntimeMonitor(cpu_threshold=80, mem_threshold=90)
with monitor:
    long_running_job()

# Profiler — quick wall-clock + memory snapshot
with bm.Profiler() as p:
    work()
print(p.summary())

Status

Standalone fork of scitex.benchmark. Only dep is psutil. The umbrella package's scitex.benchmark import path is preserved via a sys.modules-alias bridge. Convenience builders inside benchmark.py import scitex.io/scitex.stats lazily if you opt in — without those installed they simply error out at the call site, but the core API works without them.

Part of SciTeX

scitex-benchmark is part of SciTeX. Install via the umbrella with pip install scitex[benchmark] to use as scitex.benchmark (Python) or scitex benchmark ... (CLI).

Four Freedoms for Research

0. The freedom to run your research anywhere — your machine, your terms.
1. The freedom to study how every step works — from raw data to final manuscript.
2. The freedom to redistribute your workflows, not just your papers.
3. The freedom to modify any module and share improvements with the community.

AGPL-3.0 — because we believe research infrastructure deserves the same freedoms as the software it runs on.

License

AGPL-3.0-only (see LICENSE).