简体中文

Adaptive Translation V3.0.1

Adaptive Translation is a privacy-minded browser extension for inline translation. It can translate selected text, translate full pages paragraph by paragraph, show dictionary bubbles for single words, customize translation bubble styles, and optionally translate captions, images, and PDFs with user-selected providers.

The extension is designed as a modular WebExtension project with separate Chrome and Firefox release manifests.

---

Features

• Selection translation: translate selected text from the context menu.
• Inline page translation: insert translations below original paragraphs, list items, headings, and other readable text blocks.
• Dictionary bubble: show a learner-friendly dictionary entry when a selection looks like a single word.
• Custom styles: configure translation font, font size, text color, bubble color, and border color.
• YouTube captions: optional bilingual caption overlay with built-in/API priority.
• Image and PDF translation: local OCR with bundled Tesseract.js, with optional vision-provider fallback.
• Prompt editor: customize LLM system/user prompts for translation and dictionary mode.
• Provider adapters: OpenAI, Gemini, Google Translate, Azure Translator, DeepL, and Chrome AI where available.
• Internationalized UI: English, Chinese, Japanese, Korean, French, German, and Spanish.
• Diagnostics: debug logging, background ping, and provider self-test.

---

Privacy Model

Adaptive Translation does not collect, record, sell, or upload user data to servers controlled by the extension developer.

Content is processed only when the user triggers a translation feature or enables a related feature such as caption translation. Depending on the selected provider, selected text, page text, OCR content, image/PDF content, or captions may be sent directly from the extension to the third-party translation provider configured by the user.

API keys and preferences are stored with browser extension storage and are used only to call the provider selected by the user. Users should review the privacy policies and terms of the third-party providers they choose.

See PRIVACY_POLICY_DRAFT.md for the current store-facing privacy policy draft.

---

Supported Providers

• OpenAI: chat completions and vision-capable translation flows.
• Gemini: text, dictionary JSON, and vision-capable translation flows.
• Google Translate: text translation.
• Azure Translator: text translation.
• DeepL: text translation.
• Chrome AI: local Chrome AI provider where the browser exposes the required APIs. This is Chrome-only and not advertised for Firefox builds.

Providers are implemented in providers/, with shared prompt helpers in prompts/.

---

Settings

General

• Translation provider
• Source language
• Target language
• Dictionary mode for single-word selections

Style

• Font family
• Font size
• Text color
• Bubble background color
• Border color
• Live preview

Advanced

• UI language
• Prompt editor
• OCR settings
• YouTube caption preferences
• Debug logging and diagnostics

---

Development Install

Chrome / Chromium

1. Open chrome://extensions.
2. Enable Developer mode.
3. Click Load unpacked.
4. Select this project folder.
5. Open the extension options page and configure a provider.

Firefox

1. Open about:debugging#/runtime/this-firefox.
2. Click Load Temporary Add-on.
3. Select manifest.firefox.json or a packaged Firefox build manifest.

For AMO packaging, use the build script described below.

---

Packaging

Generated files are written to dist/. Do not commit dist/ to source control.

Firefox AMO Package

powershell -ExecutionPolicy Bypass -File .\tools\build-firefox.ps1

This creates a timestamped .xpi under dist/.

AMO Source Package

powershell -ExecutionPolicy Bypass -File .\tools\build-amo-source.ps1

This creates a source archive for Mozilla review. It includes source files, bundled dependencies, manifests, build scripts, and SOURCE_BUILD_INSTRUCTIONS.md, but excludes generated dist/ artifacts.

Chrome Web Store Package

powershell -ExecutionPolicy Bypass -File .\tools\build-chrome.ps1

This creates a Chrome upload zip using manifest.chrome.json.

---

Store Manifests

• manifest.json: original cross-browser development manifest.
• manifest.chrome.json: Chrome Web Store release manifest.
• manifest.firefox.json: Firefox AMO release manifest.

Firefox packaging uses background.scripts. Chrome packaging uses background.service_worker.

---

Project Layout

background.js             Extension background entry
content/                  Page translation, image overlay, YouTube overlay
core/                     Browser API, settings, routing, i18n, logging
options/                  Options UI and prompt editor
providers/                Translation provider adapters
prompts/                  Shared prompt builders
ocr/                      OCR integration
pdf/                      PDF extraction helpers
vendor/                   Bundled PDF.js and Tesseract.js assets
_locales/                 Browser extension localization files
tools/                    Release packaging scripts
docs/                     Store listing, privacy, and review notes

---

Internationalization

Included locales:

en, zh, ja, ko, fr, de, es

To add a locale, create _locales/<lang>/messages.json with the same message keys.

---

Notes For Reviewers

PDF.js and Tesseract.js are bundled locally for PDF parsing and OCR. They are not downloaded from a remote host at runtime.

Dynamic import() is used to load extension-bundled modules via browser.runtime.getURL(...) / chrome.runtime.getURL(...) when the user triggers image, PDF, OCR, or YouTube-related features.

---

License

This project is licensed under the GNU General Public License v3.0. See LICENSE.