Stop uploading your personal life to the cloud. VITA is a fully offline personal health companion that tracks your nutrition, manages daily budgets, and answers your voice queries entirely on your phone. No subscriptions, no trackers, and no server logging—just a single $5 one-time purchase for total food sovereignty.
One-Time Purchase · Offline SQLite Database · Local Speech-to-Text & LLM
Health decisions are personal. We believe your daily choices shouldn't be packaged and sold to advertisers. VITA is designed from the ground up to protect your privacy by running 100% on-device.
How we run high-end AI capabilities directly on your mobile hardware, without remote servers.
Whisper Tiny runs locally on your phone. In the future, we will transition to native on-device OS speech APIs, removing Whisper entirely to ensure 100% free local execution.
Currently using an optimized Qwen 2.5 1.5B model. Our roadmap includes retraining and distilling a custom **Mistral 3B** model to fit within tight mobile RAM constraints.
The C++ query engine (libciqual_api) manages static SQLite databases with batch lookup capabilities, returning results in under 0.2 milliseconds.
We lift the mental load of nutrition, grocery shopping, and fitness tracking by doing all the planning for you.
Speak in a natural, conversational tone. VITA understands your commands, queries, and logs, acting as an interactive assistant rather than a static form.
Track 27 essential nutrients, calorie budgets, and weight logs in real-time. Watch progress bars update instantly inside local databases.
Log workouts or meals hands-free directly from your smartwatch or wearables at the gym. No screen-tapping or phone-unlocking needed.
Automatically identifies gaps in your micro-nutrients and recommends specific vitamins or supplements from our trusted partners.
Once partner market integrations launch, VITA coordinates with grocery delivery systems to drop ingredients at your door automatically.
By automating logging, meal recommendations, and grocery delivery, VITA handles the planning and logistics so you can focus on living.
No subscriptions. No advertisements. Just clean code and complete privacy.
One-time payment for lifetime access.
Previously, we planned to use the LLM as the main orchestrator to read inputs and decide intents. Now, a lightweight CamemBERT/BERT intent classifier running as a native ONNX model runs first. It determines the intent instantly (under 50ms), saving battery. The local Mistral LLM is used solely to convert text to structured JSON (entity extraction).
VITA's key innovation is training a Joint Embedding Predictive Architecture (JEPA) model locally on users' devices. Every month, the aggregated, anonymous weights are uploaded to our secure cloud to improve the global predictive model. By training this JEPA model on top of sovereign European foundations (Mistral, CamemBERT, and CIQUAL), VITA establishes a private AI architecture fully aligned with strict EU regulations (GDPR), paving the way for public funding eligibility and scaling.
VITA is built on a fully sovereign European technology stack: leveraging the French CamemBERT intent classifier, the French national CIQUAL database, and local Mistral models. By processing data 100% locally and matching strict European regulatory frameworks (GDPR), VITA secures user privacy, decouples from foreign cloud dependencies, and positions itself for institutional European funding and market access.
Our main struggle with hardware chips (Raspberry Pi/Radxa/Orange Pi) was the rising cost of RAM. Building dedicated hardware with sufficient memory was prohibitively expensive. Shifting to mobile leverages the high-performance RAM that users already own in their smartphones, allowing us to offer the app for a $5 one-time fee.
All database lookups and biometric calculations run inside native C++ libraries (compiled for arm64-v8a architectures). The Kotlin app links to these native libraries via the Java Native Interface (JNI), maintaining desktop-level execution speeds directly on the phone.
Using custom algorithms, VITA normalizes nutritional data across food items. The static database stores nutrients per 100g. The app automatically intercepts query outputs and scales values dynamically based on portion sizes (e.g., cups, tablespoons, or raw grams) before logging.
We started VITA because we believe health technology should respect your privacy, adapt to your life, and run on code you can trust. As developers, we were frustrated by the hardware cost inflation and invasive cloud practices that plague modern health tools.
VITA is our answer: we pivoted from a $250 custom kitchen hardware box to a fully local mobile app, bringing Whisper speech processing and lightweight LLMs directly to your phone. 100% offline, subscription-free, and open to anyone who values their data sovereignty.
— The VITA Team
Designed in Europe