Phonetik
Phonetic analysis engine that gives LLMs accurate rhyme detection, stress scanning, meter identification, and syllable counting backed by the full CMU Pronouncing Dictionary.
README
phonetik
A Rust library and HTTP server for English phonetic analysis. Rhyme detection, stress scanning, meter identification, and syllable counting — backed by the full CMU Pronouncing Dictionary compiled directly into the binary.
No files to download. No API keys. No runtime dependencies. Phonetik::new() and go.
What it does
- Rhyme finding — perfect, slant, and near rhymes with confidence scores
- Stress analysis — extract stress patterns from any English text
- Meter detection — identify iambic pentameter, trochaic tetrameter, etc.
- Phoneme lookup — ARPAbet transcriptions for 126K words
- Syllable counting — per-word and batch
- Rhyme map — detect phoneme repetition patterns across lines of verse
- Word comparison — phonetic similarity scoring between any two words
Install
[dependencies]
phonetik = "0.3"
Or install the binaries:
cargo install phonetik
This gives you phonetik-server (HTTP API) and phonetik-mcp (MCP tool server).
Usage
use phonetik::Phonetik;
let ph = Phonetik::new();
// Rhymes
for r in ph.rhymes("love", 10) {
println!("{} ({:?}, {:.0}%)", r.word, r.rhyme_type, r.confidence * 100.0);
}
// ABOVE (Perfect, 100%)
// DOVE (Perfect, 100%)
// GLOVE (Perfect, 100%)
// MOVE (Slant, 60%)
// GROOVE (Slant, 60%)
// Meter
let scan = ph.scan("uneasy lies the head that wears the crown");
println!("{} — {}", scan.visual, scan.meter.name);
// x / x / x / x / x / — iambic pentameter
// Comparison
let cmp = ph.compare("cat", "bat").unwrap();
println!("{:.0}% similar, {:?}", cmp.similarity * 100.0, cmp.rhyme_type);
// 67% similar, Perfect
// Lookup
let info = ph.lookup("extraordinary").unwrap();
println!("{}: {} syllables", info.word, info.syllable_count);
// EXTRAORDINARY: 6 syllables
How it works
The entire CMU Pronouncing Dictionary is compiled into the binary at build time. Every phoneme is encoded as a single byte — no strings in the hot path. Rhyme detection is algorithmic:
- Perfect rhymes are resolved by grouping words that share identical sound from their last stressed vowel onward. One hash lookup.
- Slant rhymes use a vowel distance matrix and coda suffix matching. Words sharing the same consonant ending but different vowels (love/move) are found by index traversal, not pairwise comparison.
- Near rhymes walk a precomputed graph of consonant clusters that differ by one edit. Same vowel, slightly different ending (night/nice).
Server
The crate includes an HTTP server binary behind the server feature (enabled by default):
cargo install phonetik
phonetik-server
# Phonetik starting on :1273 [126052 words]
Endpoints: /health, /syllables, /syllable-counts, /rhymes, /rhymes/perfect, /rhymes/slant, /rhymes/near, /compare, /scan, /rhymemap, /document
MCP
phonetik includes an MCP (Model Context Protocol) server so AI assistants can use it as a tool. Runs over stdio — no network, no auth.
cargo install phonetik
Add to your MCP client config (Claude Code, Cursor, etc.):
{
"phonetik": {
"command": "phonetik-mcp"
}
}
Tools: lookup, rhymes, scan, compare, analyze_document
Feature flags
Both server and mcp are enabled by default. For library-only use:
[dependencies]
phonetik = { version = "0.3", default-features = false }
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。