Chatglot - Chat Translation

Chatglot

Chatglot is a Fabric client-side mod that translates Minecraft chat messages on the fly.
A translate button (✍ by default) is added to each chat message. Click it, or let auto-translate handle it, and the translated result is shown in chat either as a new line or by replacing the original message.

Client-side only. No server installation is required.

Features

• Translate button appended to every chat message (configurable label)
• Click-to-translate or automatic chat translation
• Works out of the box — the default provider uses Google Translate's free web API, with no API key, account, or setup
• Local translation cache — repeated messages are shown instantly without another request
• Result shown as a new line or replaces the original chat message

• Sign and hanging sign translation

  • Look at a sign and press G to translate its text

• Book translation

  • In a written book or on a lectern, press G or click Translate to translate the current page
  • Book and quill pages can be translated with the Translate button (page contents are never changed)

• Chat input translation

  • Choose an outgoing language in the chat input and preview a translation of your draft before sending
  • Apply the preview with Tab or a click
• Open the settings screen with F8 or /chatglot config
• Translate a specific chat message with `/chatglot translate `

Translation Providers

Chatglot supports multiple translation backends. Switch between them in the settings screen.

Recommended Providers

• Default: Best for getting started. Free, needs no setup, and supports auto-translate. Because it relies on an unofficial API, Google may rate-limit it during heavy use; switch to another provider if that happens.
• GAS: Good if you want a free setup with your own Google account. Google Apps Script documents up to 5,000 Translate calls per day for consumer accounts, though quotas can change. See Apps Script quotas.
• Codex: Recommended if you already subscribe to ChatGPT. OpenAI states that Codex is included with ChatGPT Plus, Pro, Business, Edu, and Enterprise plans, with plan-based limits. See Using Codex with your ChatGPT plan.
• Azure: A good option if you want an official API with a free tier. Azure Translator F0 includes 2 million characters free per month according to Microsoft's pricing page. See Azure Translator pricing.
• TranslateGemma: A strong option if your PC has enough headroom for local inference. After setup, it runs locally over localhost, can be fast, and does not require a paid API for each request. It does require downloading the llama.cpp runtime and a TranslateGemma GGUF model.

Configuration

The config file is located at config/chatglot/chatglot.json.

Key settings:

• enabled — Enable or disable the mod entirely
• autoTranslateEnabled — Toggle automatic translation
• outgoingTargetLanguage — Default language for translating your own chat draft (default: English)
• targetLanguage — e.g. JA, EN, EN-US, ZH-HANS (defaults to your Minecraft language)
• provider — Choose your translation backend
• overwriteOriginalWithTranslation — Replace original message instead of adding a new line
• appendTranslateButton — Show or hide the ✍ button
• translateButtonLabel — Customize the button label
• showTranslationPrefix — Show or hide the translated-message prefix
• requestTimeoutSeconds — Network timeout (5-240 seconds)
• maxConcurrentTranslations — Max parallel translation requests (1-16)
• useSharedAppDataSettings — Share API keys, model settings, and local backend settings through %LOCALAPPDATA%
• translationCacheEnabled — Enable the local translation cache (default: true)
• translationCacheMaxEntries — Maximum number of cached translations (default: 5000)

Translation Cache

Successful translations are saved to config/chatglot/translation-cache.json and reused for the same text, language, and provider settings. The oldest entries are removed once the limit is reached. Clear the cache with /chatglot cache clear or the button in the settings screen.

The cache stores the original text, including chat messages, in plain text on your PC. API keys and other credentials are never stored in it.

TranslateGemma Local Backend

translategemma_local is a Windows-only local translation option. It does not run inference inside the Minecraft JVM. Instead, it uses a separate local backend process over 127.0.0.1 / localhost.

When you use the TranslateGemma setup buttons in the config screen, Chatglot:

• Downloads a prebuilt llama.cpp Windows runtime
• Downloads a TranslateGemma GGUF model file
• Starts llama-server.exe locally and sends translation requests to that local server

The default model is mradermacher/translategemma-4b-it-GGUF using translategemma-4b-it.Q4_K_M.gguf.

By default, Chatglot stores the local backend runtime, model, logs, and state inside:

• config/chatglot/local-backend/ when shared AppData settings are disabled
• %LOCALAPPDATA%/ChatglotLocal/ when shared AppData settings are enabled

You can also override the runtime path, model path, download URL, or shared directory from the config screen.

Shared Settings Storage

If useSharedAppDataSettings is enabled, Chatglot stores shared settings in %LOCALAPPDATA% so multiple launch profiles can reuse them.

This may include:

• API keys for supported translation providers
• Codex token file location and related model settings
• TranslateGemma local backend settings such as runtime path, model path, port, and shared directory

Shared settings are stored locally on your PC. They are not uploaded anywhere just because this option is enabled.

Privacy Notice — External Data Transmission

This mod sends chat text to external translation services. The destination depends on the provider you select.

Translating a chat draft sends the text in the input field only when you press the translate button. Typing or opening the language picker does not send the draft. Results already in the local cache are shown without sending a request.

• default provider: Chat text is sent to Google's unofficial web translation API (translate.googleapis.com). It is free and needs no setup, but Google may limit or change access without notice. See Google Privacy Policy.
• gas provider: Chat text is sent to the Google Apps Script web app URL that you configure. Requests are processed through Google's infrastructure. See Google Apps Script Additional Terms and Google API Services User Data Policy.
• openai provider: Chat text is sent to the OpenAI API using your API key. See OpenAI data controls.
• codex provider: Uses browser-based OAuth, stores the token locally, and then uses that token to send chat text to OpenAI services. The mod also fetches the available model list from https://modelapi.yoima.com/api/codex-models/list. No chat content is sent to that model-list endpoint.
• gemini provider: Chat text is sent to the Gemini API. For Google AI Studio / Gemini API usage, review the provider terms here: Gemini API Terms.
• deepl provider: Chat text is sent to the DeepL API. See DeepL Privacy Policy.
• anthropic provider: Chat text is sent to Anthropic services using your API key. See Anthropic Privacy Policy.
• azure provider: Chat text is sent to Azure Translator using your API key. See Data, privacy, and security for Azure Translator.
• google provider: Chat text is sent to Google Cloud Translation using your API key. See Google Cloud Privacy Notice.
• translategemma_local provider: Chat text is sent only to a local backend running on 127.0.0.1 / localhost. During setup, this mod downloads the llama.cpp runtime and the TranslateGemma GGUF model from a third-party hosting source configured by the mod.

Source & License

Source code: GitHub — yoima-jp/Chatglot
License: MIT