Crates

Open-source Rust crates, published on crates.io.

Writing checks

ai-slop: Lints bug reports, commit messages, changelogs and other outbound text for AI slop before it ships.
unslop: Finds filler, hedging, puffery and stock transitions in essays, posts, email, reports and docs, and points to the exact span.
slop-detector: Scans received text for chatbot paste residue, prompt-injection phrasing and formulaic writing, citing the bytes behind every finding.
dale-chall: The New Dale-Chall list of easy American-English words, as data for readability scoring.

data-classification: Recommends a sensitivity tier for each text document from versioned policy packs, with JSON evidence of what matched and where. Only a recorded human decision lowers a tier.
text-mirror: Converts Office documents, text-layer PDFs, email and ZIP archives into text, with a manifest that records how every file was handled.
decomment: Removes comments from source text in 13 languages without touching strings, templates or regex literals.
sql-ddl-to-json-schema: Parses MySQL and MariaDB table definitions into a table model and JSON Schema documents.
subcue: Parses, composes and converts SRT, WebVTT and ASS/SSA subtitle files.

tokenx: Estimates how many tokens a language model reads in a string, with no vocabulary or BPE table.
token-sampling: Sampling stages over model logits, from temperature, top-k and top-p to Mirostat and greedy selection.
token-align: Maps overlaps between two tokenizations of the same text, and character positions between two strings.
sse-parse: Incremental parser for Server-Sent Events byte streams, fed chunk by chunk from any transport.

english-pos-tagger: Tags English parts of speech and adds base-form lemmas for nouns, verbs and adjectives.
pos-tagger: Brill-style English part-of-speech tagger with a built-in lexicon of about 92,000 entries.
english-tokenizer: Splits a sentence into tokens and tags each one by type, such as word, number, URL, hashtag or emoji.
snowstem: Snowball stemming for 21 languages, checked word for word against the Snowball test vocabularies.
vader-parity: VADER sentiment scoring with exact output parity against Python vaderSentiment 3.3.2.
thefuzz-parity: Fuzzy string scoring and ranked matching, with scores tested against a pinned Python release.
unicode-normalize-aligned: Unicode NFC, NFD, NFKC and NFKD normalization that aligns every output character to the input.
uax14-linebreak: Line break opportunities for Unicode 17.0.0 text, from the default rules of Unicode Standard Annex #14.

redos-detector: Scores the ReDoS risk of an ECMA-262 regular expression without running it.
xss-filters: Context-aware output encoding, so an untrusted string cannot break out of its HTML context and run script.
secure-json-parse: Parses untrusted JSON and blocks prototype-poisoning keys.
graphql-strip-sensitive-literals: Redacts sensitive literals from a GraphQL document before logging or usage reporting.

mp4-track: Reads tracks, sample tables and metadata from MP4 files under configurable limits, and writes progressive MP4.
wav-codec: Reads and writes PCM and IEEE-float WAV files.
pitch-estimate: Per-frame pitch detection with the McLeod pitch method and YIN.
tiny-webp: Lossy WebP encoder with a cwebp-shaped command line.
hqx: Pixel-art magnification by 2x, 3x and 4x that keeps sharp edges sharp.
mgrs-convert: Converts latitude and longitude to and from UTM, UPS and MGRS grid references.

aistackregistry.com: One catalog of current AI model ids, limits and provider settings, so a coding agent can update every repo from a single fetch.