Artificial Intelligence on your Device.
CodEasy's assistant lives exclusively inside your smartphone. It never sends data across the internet, requires no subscriptions, and answers questions about your documents in total privacy.
How to choose the right profile
Artificial intelligence runs 100% inside your phone. Download the model best suited to your smartphone and habits just once: switch or remove it whenever you want.
How much RAM does your smartphone have?
RAM determines which neural models can be loaded into device memory without causing slowdowns.
Vision Profile (Base)
Based on your answers, this profile offers the best balance between accuracy and responsiveness.
Parameters required to run 100% local offline processing on the phone, without sending data to the cloud.
You indicated you do not wish to use extra storage: inside the CodEasy app, in the AI Management section, you can check whether your smartphone supports system-integrated Gemini Nano to process documents at zero storage cost.
Vision Profile
RecommendedDirect visual understanding: recognizes spatial layout, tables, receipts, and invoices.
- ✓Layout understanding: the AI inspects the entire image, interpreting columns, line items, logos, and headers.
- ✓Ideal for tabular data and receipts: connects line items, tax rates, and totals while preserving document structure.
- ✓Direct visual analysis: does not depend on an intermediate transcription step, reducing graphical misinterpretations.
- ✕Less suited for handwriting: for handwritten notes or difficult script, the Specialist profile ensures sharper character recognition.
- ✕Requires a phone with adequate RAM: needs at least 4 GB of RAM to keep the UI snappy.
Balanced intelligence for daily use. Answers smoothly and naturally about your archived documents, offering the ideal balance between reasoning power and speed.
The app supports Gemini Nano, the AI integrated directly into the Android OS. Take advantage of the dedicated chip to get faster responses with minimal battery drain and zero extra downloads.
Your personal archivist, always by your side.
Think of CodEasy as a trusted secretary safeguarding your papers in a protected wallet on your smartphone. Stop searching through thousands of photos: query documents, receipts, and records in natural language, or find answers to frequent questions.
Examples of what you can ask
Your driver's license (No. U12345678X) expires on September 18, 2028, in approximately 2 and a half years. We recommend booking your renewal medical visit before that date.
Frequently Asked Questions
🔒Do my documents ever end up on the internet or in the cloud?+
Absolutely not. Everything remains stored in your smartphone's private memory. No photos, transcribed text, or queries are ever sent outside: not to remote servers, not to Google, and not to third parties.
✈️Does it really work without an internet connection?+
Yes, 100%. The assistant resides entirely inside your phone's processor. You can scan, catalog, and query information in airplane mode, on trains, in the mountains, or in underground locations without signal.
💶Is the app or AI assistant paid?+
No, it is completely free. All document scanning, cataloging, and assistant consultation features are included with no hidden fees, no monthly subscriptions, and no query limits.
📑What kinds of documents can I save and ask the assistant about?+
Any type: national IDs, driver's licenses, passports, receipts and invoices with warranties, medical reports and lab tests, vehicle registrations, contracts, and loyalty cards with barcodes. The AI recognizes both structured cards and free-form documents.
🛡️What CAN'T the assistant do?+
The assistant queries only the documents you save in the app. It does not browse the web, does not know external weather or news, does not make payments or bank transfers, and does not replace medical or legal advice.
⚡How do I know if my smartphone is compatible?+
Inside the app, in the AI Management section, you can verify your phone's available RAM. If you have at least 3-4 GB of RAM, you can comfortably use the AI features. With less RAM, you can still use the app for document storage, but on-device AI models are not recommended.
📱Do I need a high-end or latest-generation smartphone?+
No, generally any smartphone from recent years can handle the app's AI features. If you have an older device or minimal RAM (under 3 GB), you can still use the app without AI models solely for document storage (native Android Text Recognition is also supported and requires minimal resources).
🔋Does artificial intelligence drain a lot of battery?+
It depends on the model and usage. The highest power consumption occurs during document scanning and chat reasoning, varying by model. For instance, using the Base Vision profile with casual scanning consumes minimal, virtually unnoticeable battery. If responses feel sluggish, switching to a lighter model will reduce power consumption.
🔄If I change or lose my phone, can I recover my documents?+
Yes! You can enable optional Google Drive backup: your data is saved in the app's isolated Drive storage. You can restore it on a new phone by signing in with the same Google account. Alternatively, you can export your documents to PDF and back them up locally or to your cloud of choice.
On-Device Models in CodEasy
In CodEasy, on-device models are available for Android 11 and later through the Flutter Gemma. Select a model for detailed specifications.
Gemma 4 E2B
RecommendedMultimodal VLMEnd-to-end Vision-Language Model: direct extraction from images, receipts, and tables without intermediate OCR.
Model Overview
Gemma 4 E2B (Effective 2 Billion parameters) is the compact Vision-Language Model developed by Google and adapted for local inference through the LiteRT-LM runtime with 4-bit AWQ quantization (weights distributed in .litertlm format). Unlike traditional text-only Large Language Models, the official Hugging Face model card highlights its native end-to-end multimodal architecture: the visual encoder and language transformer reside within the same neural network, enabling the model to interpret text and pixels simultaneously in a single compute pass.
Integration in CodEasy
In CodEasy, Gemma 4 E2B is the recommended engine for the Base Vision profile and local chat. Unlike Qwen3 0.6B, which does not support vision in any way and must rely on separate OCR, Gemma 4 visually understands spatial document layout: it perceives document geometry, correlating invoice columns, tax rates, and receipt line items even if arranged irregularly. Furthermore, compared to ultralight models like Qwen3, Gemma 4 E2B provides vastly superior contextual reasoning and drastically cuts hallucinations during automated categorization and field extraction. It requires approximately 2.4 GB of storage and a compatible mobile GPU (OpenCL/Vulkan), offering the best overall balance for phones with at least 4 GB of RAM.
Explore official documentation for Gemma 4 E2B
Decoupled Component Architecture in the App
CodEasy allows advanced users to freely decouple and combine the scanning engine, the language assistant, and hardware acceleration options:
- • Google ML Kit
- • PaddleOCR-VL
- • Gemma 4 Vision
- • Gemini Nano
- • Qwen3 0.6B
- • Gemma 4 E2B / E4B
- • GPU/NPU vs CPU Accelerator Toggle
- • Customizable System Prompt
- • Response Token Budget: 256 – 1536
From Capture to Vault: The AI Pipeline
Scroll down to discover how your smartphone's neural architecture processes, understands, and secures every document locally, without ever touching the cloud.
Model Specifications & Hardware Requirements
CodEasy is compatible with Android 11 and later versions.
Integrated RAM Monitoring:Inside the CodEasy AI app, under the AI Management section, you can check your smartphone's available RAM to select the optimal model.
| Model | Download | Key Capabilities | Recommended Accelerator | Minimum RAM | Recommended RAM | Target Phone Era* |
|---|---|---|---|---|---|---|
| Gemini Nano | 0 MBBuilt into OS | On-Device Chat (OS)Zero Download | NPU | — | — | ~2024+ |
| Qwen3 0.6B | ~586 MB | Thinking ✅Function Calling ✅Vision ❌ | CPU / GPU | 3 GB | 4 GB | ~2021+ |
| PaddleOCR-VL 1.6 | ~1.39 GB | Vision ✅ (Doc OCR)Thinking ❌Function Calling ❌ | GPU | 4 GB | 6 GB | ~2022+ |
| Gemma 4 E2B | ~2.4 GB | Vision ✅Thinking ✅Function Calling ✅ | GPU | 4 GB | 6 GB | ~2023+ |
| Gemma 4 E4B | ~4.3 GB | Vision ✅Thinking ✅Function Calling ✅ | GPU | 6 GB | 8+ GB | ~2024+ |
| Gecko 512 Embedder | ~116 MB | Embeddings (768-D) | CPU / NPU | — | — | ~2019+ |
The phone launch year is an indicative estimate based on typical chipset architectures (Snapdragon, MediaTek Dimensity, Google Tensor, Samsung Exynos) and standard RAM configurations of that generation. A flagship device from 2021 or 2022 can run complex models much faster than a budget 2024 phone. Actual compatibility depends on available RAM and the compute capability of the device GPU/NPU, rather than solely on production date.
Your documents, always with you.
Safe on your phone.
Download CodEasy AI now from Google Play. No mandatory account, no hidden subscriptions.