langdetect-rs
RustPython
Language detection identical to Python's langdetect, bit for bit, 40–52× faster on one core and up to 315× with all cores.
pip install langdetect-rsOpen-source libraries · Rust · Python · Ruby
Tools for evaluation, scientific computing and data applications
We port slow reference implementations to fast, verified code, build gems for AI features in Rails apps, and publish the code and data behind our research.
New · 0.2.1rougers / rouge-score-rs
ROUGE scores identical to rouge-score, 30–80× faster on one core
A Rust port of Google's rouge-score, the implementation behind Hugging Face Evaluate. Every float matches, including the tokenizer, the Porter stemmer and rougeLsum. Change one import and keep your code.
pip install rouge-score-rs- rouge-score19.65 s
- rouge-score-rs, loop0.28 s
- score_batch, 8 threads0.056 s
Apple M1 Pro, Python 3.12, median of 3 runs, scores checked bit for bit. Speedups depend on text length, metrics and threads: 30–80× on one core and 83–353× on eight across CNN/DailyMail, XSum and PubMed. Full benchmarks
Libraries
Also maintained here
rust-lstm
Rust
LSTM, GRU and BiLSTM networks with full backpropagation through time, gradients checked against finite differences and PyTorch.
cargo add rust-lstmengram
Ruby
Long-term memory for AI agents in Ruby: recall the facts that matter and add them to the prompt, stored in your own Postgres.
gem install engramResearch
Code and data you can rerun
Each study ships with its code, prompts, raw outputs and a DOI on Zenodo.
Oct 2026 · O. Kholodniak, O. Prokhorov
Gradient reach in recurrent networks: code and training runs
How dropout, zoneout and gate initialisation change how far gradients reach back in time in LSTM and GRU networks, and whether that reach, measured early in training, predicts which runs learn long-range dependencies. Trained with rust-lstm.
Oct 2026 · O. Kholodniak
Comparing IRT models and simple baselines for budget-constrained LLM evaluation
Item response models against simple estimators on public model × item correctness matrices: how well each estimates item difficulty, a new model's capability and its full-benchmark accuracy when only 0.5–80% of the items are run.
Oct 2026 · O. Kholodniak
QoS-aware orchestration of local and cloud large language models in distributed information systems
Choosing per request between a small on-device model, a larger LAN server and a cloud API: latency model, routing policies, a netem testbed and the measurement data.
Updates
Latest releases
-
Porting
langdetectto Rust bit for bit, and why Python 3.12 changes its answers. -
langdetect-rs 0.1.0:
langdetectwith identical results, 40–52× faster, on PyPI and crates.io. -
rougers 0.2.1: a faster core, 30–80× faster than
rouge-scoreon one core and up to 350× withscore_batch, same scores. -
rougers 0.2.0: Unicode and character tokenizers for ROUGE in any script, also in parallel
score_batch. -
rouge-score-rson PyPI and theSyntaxSpirits/rougemetric for Hugging Face Evaluate. - Code and training runs of the gradient-reach study archived on Zenodo.
-
rust-lstm 0.10.0: Adam bias-correction fix, reproducible runs with
seed, cell-update dropout and GRU zoneout. - engram 0.8.0: long-term memory for Ruby AI agents, stored in your own Postgres.
Who is behind this
SyntaxSpirits is maintained by Oleksandr Kholodniak, a software engineer and PhD researcher at the National Aerospace University “Kharkiv Aviation Institute” in Ukraine. The libraries come out of production Ruby work and research on evaluating language models.
Found a bug or need a feature? Open an issue on the library's GitHub page, or write to hello@syntaxspirits.dev.