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.

Distribution

• Download on Modrinth
• Report issues on GitHub
• View source on GitHub
• Project roadmap

---

Why Chatglot Matters

Minecraft servers often bring together players who do not share a common language. Chatglot helps those communities communicate by making chat translation available as a client-side Fabric mod, so servers do not need to install or trust a server-side plugin.

The project focuses on practical choice and transparency:

• Multiple translation providers so users can choose by cost, privacy, quality, and availability
• A local TranslateGemma backend option for users who prefer local inference
• Clear privacy notes for every provider that can receive chat text
• Compatibility maintenance for current Minecraft and Fabric versions
• Public distribution through Modrinth with source, issue reporting, and MIT licensing

Features

• Translate button appended to every chat message (configurable label)
• Click-to-translate or automatic translation
• Result shown as a new line or replaces the original chat message
• Open the settings screen with F8 or /chatglot config
• Translate a specific message with /chatglot translate <id>

---

Translation Providers

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

Provider	Requires
default	Nothing. Uses an author-operated free GAS endpoint. Rate-limited, and auto-translate is unavailable.
gas	Your own Google Apps Script web app URL
deepl	DeepL API key
google	Google Cloud Translation API key
openai	OpenAI API key
custom_llm	Custom OpenResponses or Chat Completions API
gemini	Gemini API key
anthropic	Anthropic API key
azure	Azure Translator API key + region
codex	Browser OAuth (one-time, token stored locally)
translategemma_local	Windows only. Downloads llama.cpp + a TranslateGemma GGUF model, then runs a local backend over localhost

---

Recommended Providers

• 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.
• 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.
• 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.

---

Requirements

• Minecraft 26.1.2 (26.1.x compatible)
• Fabric Loader
• Fabric API
• ModDeck (required for in-game settings)
• ModMenu (optional, recommended)

---

Configuration

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

Key settings:

• enabled — Enable or disable the mod entirely
• autoTranslateEnabled — Toggle automatic translation
• 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) For translategemma_local, this also sets the managed llama-server --parallel value.
• useSharedAppDataSettings — Share API keys, model settings, and local backend settings through %LOCALAPPDATA%

---

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.

• default provider: Chat text is sent to an author-operated Google Apps Script endpoint. However, requests pass through Google's infrastructure. See Google Apps Script Additional Terms and Google API Services User Data 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.
• custom_llm provider: Chat text is sent to the custom Base URL that you configure. Select either OpenResponses (/responses) or Chat Completions (/chat/completions) format. The API key is optional for local servers.
• 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