Bhasini Indic Languages

io.github.Codegres-com由 GitHub 帳號 Codegres-com 發布v0.1.0更新於 Oct 26, 2026

Indic MCP: transliterate into all 22 Indian languages, convert scripts, keymaps. Hosted or local.

已驗證Streamable HTTP可網頁執行效率與工作流AI 與機器學習

概覽

AI 產生的概覽

針對印度語言的 MCP 伺服器，讓助理在印度 22 種法定語言之間進行音譯、文字轉換、翻譯與數字在地化。

可以做什麼
  • 依發音把拉丁字母羅馬化文字轉寫成印度文字，遵循 ITRANS/Baraha 鍵位慣例
  • 在婆羅米系文字與 Meetei Mayek 之間逐字轉換文字，不改變語言
  • 在英文與 22 種印度法定語言之間，或兩種印度語言之間翻譯語意
  • 列出支援的語言及其代碼，並回傳某語言的拼音鍵盤對應
  • 把 ASCII 數字替換為某語言的在地數字
適用情境
適合處理印度語言內容：把文字依發音輸入或羅馬化為天城文、泰米爾文、孟加拉文等文字，在不同文字間搬移文字，或快速取得英文與印度語言之間的翻譯。也適合建立正確的羅馬化輸入或教學某語言的鍵盤配置。
執行需求
可作為遠端託管端點使用，也可作為本機 .NET 套件（Bhasini.Mcp）透過 stdio 執行。未宣告任何身分驗證、環境變數或請求標頭。README 中的網頁編輯器若要在本機執行，需要 .NET 8 SDK。
安裝前請注意
翻譯僅在所述用量上限內免費，大量使用可能觸及上限或需要付費。印度語言之間的翻譯會經由英文中轉，文字轉換在目標文字缺少對應字母時會退回最接近的字母，結果可能不夠精確。送出翻譯的文字會傳送至託管服務。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 Bhasini Indic Languages，將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Web executable，透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。

交給代理程式安裝

把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式，或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南，先向你展示發布者、端點和登入方式，經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。

閱讀 https://sourceweft.com/mcp/io.github.Codegres-com%2Fbhasini/install.md，並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前，先告訴我它的發布者、端點和登入方式，等我確認。不要把 API 金鑰寫進命令或我會提交的檔案；問我金鑰應從哪裡取得。

檢視代理程式所依據的安裝指南

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。

{
  "mcpServers": {
    "bhasini-indic-languages": {
      "type": "http",
      "url": "https://mcp.bhasini.com/mcp"
    }
  }
}

README

Bhasini (भाषिणी)

[Sponsor via Razorpay] [Open Bhasini] [License: MIT]

A multilingual Indic typing editor for all 22 scheduled languages of India, built with Blazor WebAssembly. Everything runs in the browser; no server round-trips while you type.

It grew out of the Kannada Nudi web editor (KannadaNudiEditor/KannadaNudiWeb) and keeps its shape: a Quill rich-text editor whose keystrokes are transliterated by a C# engine.

Languages

LanguageScript
AssameseBengali-Assamese (ৰ, ৱ)
BengaliBengali
BodoDevanagari
DogriDevanagari
GujaratiGujarati
HindiDevanagari
KannadaKannada
KashmiriPerso-Arabic (RTL), also Devanagari
KonkaniDevanagari
MaithiliDevanagari
MalayalamMalayalam (chillus, Mozhi-style keys)
ManipuriMeetei Mayek (lonsum letters), also Bengali
MarathiDevanagari (ळ)
NepaliDevanagari
OdiaOdia
PunjabiGurmukhi (tippi, bindi, addak)
SanskritDevanagari
SantaliOl Chiki
SindhiPerso-Arabic (RTL), also Devanagari
TamilTamil (ந/ன by context, ற, ழ)
TeluguTelugu
UrduPerso-Arabic, Nastaliq (RTL)

Plus English (no transliteration).

Features

  • Phonetic typing: type how it sounds (namaste → नमस्ते, vaNakkam → வணக்கம், aadaab → آداب). ITRANS/Baraha-style keys: capitals for retroflex/long sounds (T D N L Sh A I U E O), M anusvara, H visarga, .h explicit virama, _ to stop letters joining, | / || danda.
  • InScript: the government-standard layout for every Brahmic script.
  • F9 switches between the selected language and English.
  • Key map panel for the current language and method. Tap a letter to insert it, which also helps on phones.
  • Native digits toggle (१२३, ௧௨௩, ۱۲۳, …).
  • Convert: rewrite text from one Indian script to another, or turn romanised text into script.
  • Dictation through the browser's speech recognition (Chrome and Edge, where the locale is supported).
  • Open .txt / .docx, save as .docx, .txt or .html. Settings are remembered per browser.

How it works

src/Bhasini.Core has no web dependencies:

  • Input/BrahmicPhoneticEngine turns Latin keys into Devanagari, with a per-language PhoneticProfile (Dravidian short e/o, word-final virama, Tamil nasals, local key conventions).
  • Scripts/OffsetScriptMapper maps Devanagari into each ISCII-derived Unicode block by offset. Letters a script lacks fall back to the nearest letter. Script-specific spelling rules (Malayalam chillus, Gurmukhi tippi/addak, Bengali য়) run as post-processing.
  • Scripts/MeeteiMayekMapper handles Meetei Mayek. Input/SequenceEngine handles the alphabetic scripts (Perso-Arabic with word-initial/final forms, and Ol Chiki).
  • Input/TypingSession keeps the Latin keys of the current word and re-renders the whole word on every key, returning a minimal { delete, insert } edit for the editor.

src/Bhasini.Web is the Blazor WebAssembly app. wwwroot/js/editor.js intercepts keys in Quill and calls the session synchronously.

Run

Requires the .NET 8 SDK.

bash
dotnet run --project src/Bhasini.Web
bash
dotnet test

Docker (nginx serving the published static files on port 5000):

bash
docker build -t bhasini .

Script fonts come from Google Fonts (Noto). When offline, the editor falls back to system fonts such as Nirmala UI on Windows.

Support Bhasini

Bhasini is free to use. If it helps you, please consider sponsoring its development:

[Sponsor via Razorpay]

Windows app

BhasiniWindows/ packages Bhasini as a Windows desktop app and Microsoft Store AppX. It adds Global mode: press F9 to type the selected Indian language in any application.

License

MIT © Codegres

來源：README.md，提交 1c29f12

工具

6

convert_script

Convert between Indian scripts

Rewrites text from one Indian script into another letter by letter, e.g. Devanagari 'नमस्ते' → Kannada 'ನಮಸ್ತೇ'. The language does not change, only the letters. Works between the Brahmic scripts (Devanagari, Bengali, Gurmukhi, Gujarati, Odia, Tamil, Telugu, Kannada, Malayalam) and Meetei Mayek; sounds the target script lacks use the nearest letter. Text in other scripts is kept as it is.

輸入參數text必填targetLanguage必填

get_keymap

Get phonetic keyboard map

Returns the phonetic keyboard map for a language: which Latin keys produce each letter. Useful for teaching someone to type a language or for building correct romanised input for transliterate.

輸入參數language必填

list_languages

List Indian languages

Lists the 22 scheduled languages of India (plus extra official scripts for Kashmiri, Manipuri and Sindhi) with the codes the other tools accept.

無需輸入參數。

native_digits

Write digits in a native script

Replaces ASCII digits 0-9 with a language's native digits, e.g. '2026' → '२०२६' (Hindi), '௨௦௨௬' (Tamil), '۲۰۲۶' (Urdu).

輸入參數text必填language必填

translate

Translate text

Translates the meaning of text between English and the 22 scheduled languages of India, or between two of them, e.g. 'Where is the station?' → 'स्टेशन कहाँ है?' (Hindi). English → Indic uses AI4Bharat IndicTrans2; Indic → English uses Gemma 4; Indic → Indic goes through English. Free use: up to 2,000 characters per call, 5 calls a minute and 10,000 characters a day. Unlike the other tools this sends the text to Bhasini's server (mcp.bhasini.com); the text is not stored. To write romanised words in a script by sound, use transliterate instead.

輸入參數text必填sourceLanguagetargetLanguage必填

transliterate

Transliterate romanised text

Writes romanised (Latin-letter) text in an Indian script by sound, e.g. 'namaste' → 'नमस्ते' for Hindi or 'vaNakkam' → 'வணக்கம்' for Tamil. This is transliteration, not translation: spell the words as they sound in the target language. Keys follow ITRANS/Baraha conventions: capitals for long vowels and retroflex consonants (A I U E O, T D N L Sh), M = anusvara, H = visarga. Use get_keymap for a language's exact keys.

輸入參數text必填language必填

版本歷史

1
  1. v0.1.0最新Oct 19, 2026

評分

已滿足適用檢查項的 100%

該評分衡量此收錄條目的完整程度，依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價，也不表示伺服器安全或目前可用：可連通性是帶日期的檢查結果，評分也不包含安裝量或使用者評價。

  • 近 30 天內可連通必需25 分
  • 已宣告認證方式必需10 分
  • 已列出工具必需15 分
  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分