OneBox AI Agent

<div align="center"> <img src=".github/readme/icon.webp" width="96" alt="OneBox icon" />

OneBox (万宝盒)

English | 简体中文

A free AI agent & toolbox for Android: say what you need and the built-in agent drives 90+ in-app tools to get it done — plus image & document tools, productivity and daily utilities, and 100+ file formats you can open right away.

</div>

Screenshots

<table> <tr> <td><img src=".github/readme/en/01.webp?v=3" width="180" alt="90+ tools in one app" /></td> <td><img src=".github/readme/en/02.webp?v=3" width="180" alt="Say the word, AI does it" /></td> <td><img src=".github/readme/en/03.webp?v=3" width="180" alt="Bookkeeping by voice" /></td> <td><a href="https://youtube.com/shorts/eGkwoe3puSw"><img src=".github/readme/intro-en.jpg?v=2" width="180" alt="Intro video (YouTube Shorts)" /></a></td> </tr> </table>

Download

• International: Google Play · Official site
• 中国: 万宝盒官网 · 小米 / 应用宝 / OPPO / vivo / 华为应用商店搜索「万宝盒」

AI Assistant: More Than Just Feature Count

Plenty of toolbox apps are long on features. OneBox's built-in AI assistant saves you from hunting for tools and memorizing steps — just say what you want:

"Turn these images into a PDF." "Log this bill screenshot into my ledger." "Make me a to-do list." "Look up today's stock prices online."

It understands, acts, and reports back — in one flow.

• Bounded by its own permissions: everything the agent does stays inside the app's own Android sandbox and declared permission set — it never touches contacts, SMS or call logs; local tools run entirely on-device, and only the conversation with your chosen model goes online.
• An agent that drives the tools: the AI agent can invoke 90+ in-app local tools (PDF processing, image editing, file management, bookkeeping…), chaining multi-step tasks across tools — executed locally with per-tool timeouts and iteration limits.
• Picture-in-picture robot, no more waiting around: long agent task? Just leave the app — with an active AI session it drops into system picture-in-picture, where the mini-window robot keeps you posted on progress and one tap brings you back to the result. Native PiP, zero permissions required.
• One chat, many models: talk to multiple mainstream LLMs from a single entry, with custom AI engines and prompt management.
• Bring your own key — use it for free: drop in any OpenAI-compatible API key (including free tiers like Gemini or OpenRouter's free models) and the full assistant + agent is yours at zero cost. Custom engines, endpoints and tokens are fully open in the Google Play build and in builds from this repo.
• Models that run on the phone — fully offline: download a local model (Gemma 3 1B or Qwen3 0.6B) inside the app, or import your own .litertlm file, and chat without network, account or token cost — inference runs entirely on-device via LiteRT-LM (Google Play and FOSS builds only). Translation and reading text from photos have on-device paths too — see On-Device AI.
• Transparent by source: every line is public — how data is processed locally and what never leaves the device is something you can verify, not a promise you have to trust.

On-Device AI: Chat, Translate and Read Photos Without a Network

Three things in OneBox run on the phone itself. Once the model is on the device, they work without a network, an account or points.

Task	Runtime	Model	Download
Chat	LiteRT-LM	Gemma 3 1B (int4), Qwen3 0.6B, or your own .litertlm	~585 MB / ~615 MB
Translate	LiteRT-LM	Tencent Hy-MT2 1.8B	~1.8 GB
Read text in photos	ONNX Runtime + OpenCV	PP-OCRv6 small	~26 MB

• Chat: download a recommended model or import any .litertlm from storage (it is copied in, so the original can go), then set it as the chat model. GPU first, CPU fallback, 4K context; memory pressure releases the engine instead of killing the app.
• Translate: Fast, Default and Local modes. Local runs Hy-MT2 on-device in 20 languages plus auto-detect; camera input goes through the on-device OCR, so the photo is not uploaded. No login or points for the translation itself — read-aloud is a separate cloud TTS call that needs both.
• Read text in photos: the 26 MB PP-OCRv6 pack is downloaded once and SHA-256 checked. Chinese, English, Japanese and common Latin scripts; copy by line, export Markdown, save as a note, history.
• Limits: long documents, tables and image understanding still go to cloud models — on-device OCR returns text only. LiteRT-LM ships in the Google Play and FOSS builds (feature/ai/src/litertlm/); domestic builds compile a stub, so local chat and translation are missing there, while OCR is in every build.

File Opening: 100+ Formats in One App

OneBox is also a file viewer. Tap an attachment in a chat app, a file in a file manager, or share one over from another app — it lands on the right screen, with no separate Office, PDF, media player or unzip app.

Type	Count	Common formats	Opens in
Images	24	jpg / png / gif / webp / heic / tiff / svg / avif / jxl / qoi …	Built-in viewer with swipe and zoom
PDF	1	pdf	Built-in PDF preview
Documents	11	doc / docx / xls / xlsx / ppt / pptx / odt / ods / odp / rtf / epub	Built-in offline preview (feature/file-viewer) with find; Word adds page numbers and night mode
Markdown	4	md / markdown / mkd / mdown	Built-in editor, preview + edit
HTML	3	html / htm / xhtml	Built-in browser
Text / code / config	42	txt / log / json / xml / yaml / csv / kt / java / py / go / rs / sql / sh …	Built-in code editor with syntax highlighting
Audio	12	mp3 / wav / flac / ogg / aac / m4a / opus …	Built-in player, background playback + notification controls
Video	13	mp4 / mkv / mov / webm / avi / 3gp …	Built-in player with picture-in-picture
Archives	14	zip / rar / 7z / tar / gz / xz …	ZIP browsed and extracted in-app; others handed to installed apps

124 unique extensions. Every single-file entry point — file browser, external intents, deep links, AI agent tool calls — goes through the same ContentRouter in core/ui, so a file opens the same way wherever it comes from. A few pro formats (camera RAW, uncommon audio codecs) depend on the system decoder; when they can't be decoded, OneBox says so or hands them off.

Features

OneBox ships 107 feature modules. Local tools run entirely on-device; only the conversation with your chosen AI model goes online. The tables below list every module — names refer to the feature/* directories in this repo.

AI & Assistant

Module	Feature	Description
feature/ai	AI Chat & Agent	Multi-model chat, dual-model duels, prompt execution, an agent that invokes in-app tools, and offline on-device models via LiteRT-LM (Google Play / FOSS builds), with token stats
feature/ai-image	AI Image	Text-to-image generation, with share / copy / edit
feature/ai-detect	AI Detector	Check whether a text or an image is AI-generated
feature/dsh-client	DSH Client	Drive a DeepSeek Harness agent on your computer — scan-to-pair, end-to-end encrypted, LAN or cloud relay (computer side: onebox-dsh-bridge)
feature/visual-automation	Visual Automation	Feed screenshots to a multimodal model for screenshot-based UI automation

Image Viewing & Editing

Module	Feature	Description
feature/image-viewer	Image Viewer	Immersive viewer with swipe browsing, double-tap / pinch zoom (up to 5×), online image download
feature/image-preview	Image Preview	Preview and manage images; the entry point to edit / crop / filter flows
feature/single-edit	Single Edit	All-in-one editor: crop, filters, draw, curves, resize, format & quality, before/after compare
feature/draw	Draw	Free-form canvas: pen / neon / highlighter / blur / pixelate / text / stickers / healing, shapes and flood fill
feature/markup-layers	Layers	Layer-based creation: add / order / merge / blend / opacity, brushes, text, shapes, export settings
feature/crop	Crop	Crop with rotation, free-corner crop, shaped outlines and aspect ratios
feature/image-stitch	Image Stitch	Horizontal / vertical / grid stitching with alignment, spacing and fading edges
feature/image-splitting	Image Split	Split one image into a rows × columns grid of tiles
feature/image-stacking	Image Stacking	Overlay multiple images into a single composite
feature/image-cutting	Image Cut	Precise cut by horizontal / vertical start–end percentages, with invert
feature/compare	Compare	Side-by-side or overlay diff of two images, slider and pixel-level highlight
feature/load-net-image	Net Image Loader	Paste a URL to parse, preview and download images from the page
feature/pick-color	Color Picker	Tap / drag to pick pixel colors, with pan and zoom modes
feature/collage-maker	Collage Maker	Collages of up to 10 photos with templates, spacing, rounded corners, backgrounds
feature/text-card	Text Cards	Turn quotes into beautiful cards with paper backgrounds, layers and AI illustrations

Filters & Effects

Module	Feature	Description
feature/filters	Filters	100+ filters (blur, color, artistic, edge, warp) with mask-based partial application
feature/gradient-maker	Gradient Maker	Pure / overlay / mesh gradients with configurable color stops
feature/mesh-gradients	Mesh Gradients	Browse and download mesh-gradient packs, jump into the editor
feature/noise-generation	Noise Textures	Procedural noise via FastNoiseLite (OpenSimplex2 / Perlin / Value…)
feature/ascii-art	ASCII Art	Convert images into character-based text art
feature/color-tools	Color Tools	Color info (HEX / RGB / HSL), harmonies, shades / gradients, histogram analysis
feature/palette-tools	Palette Extract	Extract palettes from images, Material You style, custom palette export

Conversion & Compression

Module	Feature	Description
feature/format-conversion	Format Convert	JPEG / PNG / WebP conversion with quality control, EXIF retention, batch processing
feature/resize-convert	Resize & Convert	Resize + convert in one pass: width / height / preset / scale modes, batch
feature/weight-resize	Compress by Size	Compress to a target file size, with before/after size comparison
feature/limits-resize	Limit Resize	Scale within max width / height; skip / re-encode / scale modes
feature/gif-tools	GIF Tools	Multi-image GIFs with frame rate, repeat count, size and per-frame preview
feature/apng-tools	APNG Tools	Multi-image APNGs with frame delay and repeat count
feature/webp-tools	WebP Tools	Multi-image animated WebP with frame delay and repeat count
feature/svg-maker	SVG Maker	Convert bitmaps into SVG vector graphics
feature/zip	ZIP Archive	Pack images into a ZIP; open a ZIP to browse its tree and extract one file or all

Metadata & Encoding

Module	Feature	Description
feature/edit-exif	Edit EXIF	View and modify camera model, capture time, GPS, copyright…
feature/delete-exif	Delete EXIF	Batch-strip EXIF for privacy, or add new EXIF data
feature/base64-tools	Base64 Tools	Two-way image ↔ Base64 conversion
feature/checksum-tools	Checksum Tools	Compute and compare file / text hashes with multiple algorithms

Documents & Text

Module	Feature	Description
feature/pdf-tools	PDF Toolbox	Preview, image → PDF, PDF → image, merge, split, compress, rotate, page management
feature/ocr-document	OCR	OCR recognition and format conversion with task tracking, preview and export
feature/on-device-ocr	On-Device OCR	Offline photo text recognition (PP-OCRv6): copy by line, Markdown export, notes, history
feature/translator	Translate	20 languages with auto-detect, camera OCR, voice input, read-aloud; offline with Hy-MT2 1.8B
feature/document-scanner	Document Scanner	Scan paper documents into images / PDF
feature/markdown-edit	Markdown Editor	WYSIWYG Markdown editing, export to PDF
feature/code-editor	Code Editor	Syntax highlighting, file I/O and history
feature/file-viewer	Document Preview	Offline preview for Word / Excel / PowerPoint / OpenDocument / RTF / EPUB with page jump, find and night mode

Watermark & ID Photo

Module	Feature	Description
feature/watermarking	Watermark	Text / image / steganographic / stamp watermarks with position, rotation, opacity, tiling
feature/camera-watermark	Camera Watermark	Stamp photos with time / location / device watermarks
feature/id-photo	ID Photo	Standard ID-photo sizes, background replacement, batch export
feature/wallpapers-export	Wallpaper Export	Read and save the system and lock-screen wallpapers

Scanning & QR

Module	Feature	Description
feature/scan-qr-code	QR Tools	Scan (camera / image) and generate QR & barcodes: text, URL, WiFi, contacts…

Measurement & Utilities

Module	Feature	Description
feature/unit-converter	Unit Converter	Length, weight, temperature, area, volume, speed, time and more
feature/compass	Compass	Compass & luopan dials with bearings, magnetic declination and calibration hints
feature/measurement	Ruler	On-screen ruler and protractor, cm / inch
feature/altitude	Altitude	GPS altitude with history
feature/speed-test	Speed Test	Download / upload speed with a gauge animation
feature/calendar	Calendar	Calendar with Chinese almanac, bazi chart and lunar conversion
feature/schedule	Schedule	Events with system-calendar sync and reminders
feature/teleprompter	Teleprompter	Script editing + fullscreen playback with speed / font / color controls
feature/dead-pixel-test	Dead Pixel Test	Fullscreen solid colors with color switching and grid overlay
feature/marquee	Marquee	Fullscreen scrolling text, LED style and subtitle modes
feature/audio-cover-extractor	Cover Extractor	Extract album covers from audio files
feature/device-status	Device Status	View and copy the local IP, network and device details

Security & Encryption

Module	Feature	Description
feature/cipher	Image Cipher	Encrypt / decrypt images with passwords
feature/password-vault	Password Vault	Local encrypted account & password storage with categories and search

Life & Productivity

Module	Feature	Description
feature/bookkeeping	Bookkeeping	Income / expense tracking with charts — the AI can log bills for you
feature/marktodo	To-do	Topic-based tasks with tags, priorities, due dates and filters
feature/habit-tracker	Habit Tracker	Daily check-ins with reminders, streaks and completion stats
feature/household-items	Item Tracker	Track where things are stored, expiry dates and notes; AI can log & query
feature/period	Period Tracker	Log periods, flow, symptoms and mood; predicts the next cycle
feature/record-center	Health Records	12 metric types (blood pressure, weight, glucose…), trends and AI insights
feature/loan-calculator	Loan Calculator	Equal-payment / equal-principal loans: monthly payment, total interest, schedule
feature/lifetime	Lifetime	A timeline of life events and milestones
feature/decision-wheel	Decision Wheel	Custom random spinner to help you choose
feature/blessing-wall	Blessing Wall	Greeting cards with background music and sharing
feature/poem	Chinese Poetry	Random poems, keyword search, AI interpretation, history & favorites
feature/iching-divination	I Ching	Shake to cast a hexagram, with changing lines and AI interpretation

Games

Module	Feature	Description
feature/game2048	2048	The classic sliding-tile merge game
feature/xiangqi	Chinese Chess	Vs AI, online play, game library, endgame analysis, whole-game or clip GIF export
feature/gomoku	Gomoku	15×15 board, vs LLM, AI battles, game library, review, GIF export
feature/chess	Chess	Full rules (castling / en passant / promotion), vs LLM, game library, review, GIF export
feature/minesweeper	Minesweeper	Multiple difficulties, timer, flags
feature/sudoku	Sudoku	Three difficulties, hints, undo, timer, stats
feature/survive30s	Survive 30s	Dodge obstacles for 30 seconds
feature/dice-roller	Dice Roller	D4 / D6 / D8 / D10 / D12 / D20 dice

Content & Community

Module	Feature	Description
feature/online	Discover	Content square aggregating recommended apps, notes and mini-programs
feature/blog	Blog & Feedback	Read articles and send feedback
feature/profile	Profile	VIP level, data backup & restore, about
feature/notification	Notifications	Message center with unread badges
feature/ad-watch	Watch & Earn	Watch rewarded ads to earn points, with a daily cap

Files & Media

Module	Feature	Description
feature/file-browser	File Browser	Directory navigation, search, sort, multi-select operations
feature/file-transfer	File Transfer	LAN file transfer with QR pairing and encryption
feature/cloud-storage	Cloud Storage	Browse, upload, download and delete cloud files
feature/media-player	Media Player	Unified local audio / video playback: video picture-in-picture, background audio with notification controls, media library, favorites and queue
feature/media-picker	Media Picker	Image / video picking shared by the other tools

App Shell & System

Module	Feature	Description
feature/app	App Shell	Entry navigation with adaptive layout: drawer + bottom bar in portrait, navigation rail in landscape
feature/settings	Settings	AI engines, themes, display, TTS, system prompts and more
feature/common	Common	Shared UI base, data sync and category management
feature/search	Global Search	Search features, content and messages, with quick navigation
feature/quick-tiles	Quick Tiles	Quick-settings tiles: screenshot, color picker, scan…
feature/webview	Web Browser	Built-in browser with tabs, bookmarks and history
feature/login	Sign In	Auth flows and user agreement
feature/wechat	WeChat Integration	WeChat SDK sign-in and payment callbacks
feature/demo	Demo	Developer component showcase grid
feature/libraries-info	Open-source Licenses	License list of all third-party libraries
feature/library-details	License Details	Per-library license details
feature/boardgame	Board Game Kit	Shared Xiangqi / Gomoku / Chess UI and 480 / 1440-pixel replay export: clip selection, speed, original-size preview, sharing and saving

Tech Stack

• UI: Jetpack Compose, Material 3
• Navigation / components: Decompose (StackNavigation, childStack, retained components)
• DI: Hilt
• Code generation: KSP
• Build conventions: custom Gradle convention plugins in build-logic/convention
• Persistence: SQLite / Room (AppDatabase), MMKV (per-module)
• Network & images: Retrofit / OkHttp, Coil
• On-device inference: LiteRT-LM for chat and translation (.litertlm models), ONNX Runtime + OpenCV for OCR

Repository Structure

Path	Description
app/	Application shell and packaging entry; AppActivity.kt is the main Activity
feature/app/	Top-level shell, root navigation, global CompositionLocals, screen wiring, adaptive layout
feature/*	Business / tool feature modules
feature/ai/	AI chat, agent, tool-calling chain, AI image features, on-device model runtime (src/litertlm)
feature/common/	Shared business logic, AI engine catalog/sync
feature/settings/	Settings, themes, AI engine configuration, local model management (download / import / set as chat model)
core/*	Foundation layers: UI, domain, data, database, theme, settings, etc.
libs/*	In-project reusable libraries
build-logic/convention/	Custom Gradle convention plugins
web/	Web frontend source of the file-transfer module (esbuild project)
fastlane/	Google Play listing copy & assets (fastlane supply format)

Architecture Entry Points

Package paths beginning with com.t8rin.* are kept for compatibility with the upstream ImageToolbox module layout; the package name alone does not indicate provenance. Some files in that namespace were newly written for OneBox, and some inherited files have since been moved or renamed upstream. The header templates in scripts/license_header_origin.py record provenance explicitly with an Origin: marker — unmodified upstream copy, modified derivative, or OneBox original work — and the script classifies files by git blob SHA rather than by package name.

• App entry: app/src/main/java/com/shifenmiao/app/AppActivity.kt
• Runtime navigation root: feature/app/src/main/java/com/t8rin/imagetoolbox/feature/root/presentation/screenLogic/RootComponent.kt
• Route definitions: core/ui/src/main/kotlin/com/t8rin/imagetoolbox/core/ui/utils/navigation/Screen.kt
• Child factory wiring: feature/app/.../navigation/ChildProvider.kt
• Child composable wrappers: feature/app/.../navigation/NavigationChild.kt
• Top-level shell selector: feature/app/src/main/java/com/wanbaohe/app/screen/ScreenSelector.kt
• Adaptive navigation layout: feature/app/src/main/java/com/wanbaohe/app/navigation/AdaptiveNavigationLayout.kt

Building from Source

Requirements

• Use the bundled Gradle Wrapper 9.8.2
• JDK 17 (build-logic and CI target it)
• Android targets: Compile SDK 37.2 / Target SDK 37 / Min SDK 24
• Toolchain: Kotlin 2.4.21 / AGP 9.4.1
• When in doubt, gradle/libs.versions.toml and gradle/wrapper/gradle-wrapper.properties are the source of truth

Clone

The repository carries a lot of binary assets, so a full clone is large. For building, a shallow clone is enough:

git clone --depth=1 https://github.com/wangzhishou/OneBox.git

Common Commands

./gradlew :app:assembleGoogleArm64Debug       # single-variant debug APK (google = default channel)
./gradlew tasks --group=assemble              # list all assemble tasks
./gradlew :app:tasks

Variants

The app module uses two flavor dimensions:

• app: xiaomi / yyb / oppo / vivo / huawei / onebox / google / foss (foss is the fully-FOSS build: no GMS, Firebase, WeChat or Alipay SDKs; this is the variant CI verifies)
• abi: arm64 (64-bit only)

Build and verify with the google (overseas / Play) flavor by default; pick a domestic flavor only when it is actually needed. Task names combine the two dimensions, e.g.:

• :app:assembleGoogleArm64Debug
• :app:assembleGoogleArm64Release
• :app:assembleHuaweiArm64Debug

See run.md for more command examples.

Signing & API Keys (Optional)

The project syncs and builds without any key files:

• release signing falls back to placeholder values (not for production use)
• debug builds fall back to AGP's default debug signing
• third-party service keys (AI providers, Google Places, etc.) are empty and the corresponding features stay disabled

To customize, copy the root templates and fill in real values:

• keystore.properties.template → keystore.properties: release signing + service keys
• keystore-google.properties.template → keystore-google.properties: dedicated signing for the google flavor

Both target files are git-ignored; never commit real values.

Multi-channel Release Builds

build_release.sh builds 7 channels × 1 ABI (arm64, 64-bit only) = 7 release APKs into release/ with a release/manifest.txt summary. Examples:

./build_release.sh                                # all 7
./build_release.sh --channel xiaomi huawei        # selected channels
./build_release.sh --abi arm64                    # 64-bit only
./build_release.sh --no-offline --clean           # fetch deps online + clean first
./build_release.sh --skip-build                   # re-collect artifacts only

A real keystore.properties is required for release packaging (the script exits early without it).

CI Releases

.github/workflows/android.yml defines the tag-based release pipeline: Ubuntu runner, JDK 17, assembleRelease, then APK signing and release upload.

Update reminders & domestic distribution

The "Open Source Project" row under Profile → About shows the latest release (plus a tab badge on Profile when a newer version exists). A startup dialog can be enabled from the remote config:

"appUpdate": { "dialogEnabled": true, "dialogForVersionBelow": 150, "dialogCooldownHours": 24 }

The version source depends on the flavor: overseas (google / foss) reads GitHub Releases, the six domestic channels read the GitCode OpenAPI. GitCode mirrors only branches/tags/commits — not releases — so domestic builds need an extra publish step:

export GITCODE_TOKEN=xxx          # https://gitcode.com/setting/token-classic
git push gitcode --tags           # the tag must exist: it is used as target_commitish
./scripts/publish_gitcode_release.py --tag 1.4.0 --dir release --abi arm64

Tag pushes do this automatically: CI additionally builds the six domestic channels (arm64, appended via matrix.include) and publishes them to GitCode — requires the GITCODE_TOKEN secret and the tag being pushed to GitCode (the script uses the tag as target_commitish). The GitHub Release keeps the overseas packages only. The GitCode repository has to be public, otherwise the anonymous API returns 403.

AI / Agent Capabilities

AI features are composed from several modules:

• feature/ai/: AI chat, agent loop, tool-call execution, interactive tool bridging
• feature/ai/src/litertlm/: on-device inference through LiteRT-LM (LiteRtLmRuntime, LocalModelDirectoryRegistry) — Google Play / FOSS builds, other channels compile a stub runtime
• feature/on-device-ocr/: PP-OCRv6 on ONNX Runtime + OpenCV (PpOcrV6Engine); model pack in filesDir/ocr_models/, SHA-256 verified
• feature/translator/: TranslationService with Local / Fast / Default modes; Local runs Hy-MT2 1.8B on LiteRT-LM and reuses the on-device OCR for camera input
• feature/common/: AI engine catalog, engine sync, shared managers
• feature/settings/: AI engine, model and working-mode configuration UI
• feature/settings/.../local/: local model management — download recommended models, import a .litertlm file, set one as the chat model; an installed Hy-MT2 shows up in the translator's Local mode

Local model files live in filesDir/local_models/; the one set as the chat model is loaded on first use, and inference falls back to the CPU backend when the GPU backend fails to initialize.

The tool-calling chain is built on ToolCallTaskManager, AgentLoopExecutor and AppDatabase.

Security & Third-party Services

• No production secrets in this repo: payment (WeChat Pay / Alipay), AI provider and Google Places keys are injected via the optional keystore.properties or kept server-side.
• Backend domains and guest tokens are not in source: API base URLs and read-only guest tokens are injected from keystore.properties at build time, so builds from this repository never contact the production servers and hosted content stays empty. To enable connected features, point the app to your own backend (core/r/**/UrlConstantsFlavor.kt).
• WeChat appId / WeCom corpId are public identifiers; the WeChat platform verifies callers by package name + signature, so third-party builds cannot impersonate the official app.
• Forks & redistribution: change the applicationId (com.shifenmiao.app is taken), replace app/src/google/assets/google-services.json with your own Firebase configuration, and point the app to your own backend (API domains in core/r/**/UrlConstantsFlavor.kt).

Documentation

• run.md: build / install / lint commands
• docs/modules.md: module catalog (Chinese)
• CONTRIBUTING.md: setup, conventions, and what will not be merged
• SECURITY.md: how to report a vulnerability privately, and what is deliberately public
• docs/database-migrations.en.md: database migration rules (中文)

Support the Project

OneBox is built and maintained by a one-person company (OPC), with a limited budget for servers and AI resources. If it helps you:

• In-app donation: in the app, go to Profile → Buy Me a Coffee (points-based)
• Partnership & sponsorship: AI providers, cloud vendors and partners willing to sponsor tokens or resources are very welcome — reach us at support@oneboxable.com

Every bit of support turns directly into better features and a more stable service.

Community & Feedback

The same links are in the app under Profile → Join Community (overseas builds show Discord / X / YouTube, domestic builds show the QQ and WeChat groups):

• Discord: discord.gg/7j4hEBYGWv — chat, show & tell, early builds
• X (Twitter): @OneBoxAndroid
• YouTube: @OneBoxAndroid — intro videos and feature demos
• Bug reports & feature requests: GitHub Issues
• Questions, ideas, show & tell: GitHub Discussions
• Security problems: report them privately, see SECURITY.md — please do not open a public issue
• 中文用户:QQ 群 / 微信群见中文版 README

Contributors Wanted

OneBox is a one-person project and its surface area is now much larger than one person can cover. Help here is genuinely welcome, not a polite formality.

The most useful contributions right now, roughly in order:

1. Translations. The app and the Play listing ship in 13 locales. If your language reads wrong, only you can tell us. See fastlane/metadata/android/<locale>/.
2. Device testing. Confirm a build actually works on your OEM ROM, and report what broke.
3. The open good first issue and help wanted list — start here.
4. The agent's tool choices. 90+ tools are exposed to the agent; telling us which tool it picks badly, and what it should have picked instead, is real feedback.

Start with CONTRIBUTING.md. No keystores, API keys or backend access are needed to build.

License

Apache-2.0, see LICENSE. The project is a fork of ImageToolbox (T8RIN, Apache-2.0): the image-processing core deliberately stays close to upstream so the diff stays reviewable, and upstream copyright notices are preserved in the files they came from. See NOTICE for the project-level attribution and the meaning of the per-file Origin: markers.

Feature Gallery

One box, everything AI. Fully open source, free to use, and your data stays on-device — 18 cards to see what OneBox can do:

<table> <tr> <td><img src=".github/readme/cards/en/01.webp" width="240" alt="AI bookkeeping" /></td> <td><img src=".github/readme/cards/en/02.webp" width="240" alt="Item tracker with dedupe" /></td> <td><img src=".github/readme/cards/en/03.webp" width="240" alt="Period tracker" /></td> </tr> <tr> <td><img src=".github/readme/cards/en/04.webp" width="240" alt="Health records" /></td> <td><img src=".github/readme/cards/en/05.webp" width="240" alt="Habit tracker" /></td> <td><img src=".github/readme/cards/en/06.webp" width="240" alt="Teleprompter" /></td> </tr> <tr> <td><img src=".github/readme/cards/en/07.webp" width="240" alt="Notes" /></td> <td><img src=".github/readme/cards/en/08.webp" width="240" alt="AI activity log" /></td> <td><img src=".github/readme/cards/en/09.webp" width="240" alt="Search and format conversion" /></td> </tr> <tr> <td><img src=".github/readme/cards/en/10.webp" width="240" alt="Customize your home" /></td> <td><img src=".github/readme/cards/en/11.webp" width="240" alt="Picture-in-picture agent" /></td> <td><img src=".github/readme/cards/en/12.webp" width="240" alt="Permission before execution" /></td> </tr> <tr> <td><img src=".github/readme/cards/en/13.webp" width="240" alt="Life tools" /></td> <td><img src=".github/readme/cards/en/14.webp" width="240" alt="Scrolling marquee" /></td> <td><img src=".github/readme/cards/en/15.webp" width="240" alt="Mini games" /></td> </tr> <tr> <td><img src=".github/readme/cards/en/16.webp" width="240" alt="Dev tools" /></td> <td><img src=".github/readme/cards/en/17.webp" width="240" alt="Handy tools" /></td> <td><img src=".github/readme/cards/en/18.webp" width="240" alt="Agents, prompts and sites" /></td> </tr> </table>