Transparency Report: The Zero-Cloud Biometric Engine
The core of VeraTongue's architecture is its unwavering commitment to Data Sovereignty. Unlike traditional applications that upload raw biometric images to remote servers for processing, VeraTongue utilizes an edge-based NPU (Neural Processing Unit) approach.
When you initialize a scan, your device's camera stream is fed directly into a local <canvas> context. The Science Fraud Engine performs real-time hardware attestation to ensure the feed originates from a physical camera (preventing virtual camera deepfake injections).
Once the local NPU extracts the topographical data from your tongue (mapping variables like Heat, Moisture, Stagnation, Vitality, Coating Density, and Structural Integrity), these metrics are mathematically translated into a semantic vector.
If you choose to engage the LLM Oracle for deeper insights, the Semantic Shield ensures that only abstract, cryptographically hashed descriptors (e.g., "Archetype: Fire, Axis: Heat +85") are transmitted to our sovereign LLM infrastructure. The server remains mathematically blind to your physical appearance.
VeraTongue's diagnostic heuristic is built upon the Zunge Corpus, a strictly curated knowledge base of 86 classical Traditional Chinese Medicine (TCM) tongue observation charts.
Our algorithm maps your 6-axis biometric signature against this corpus to identify your primary constitutional archetype (Fire, Water, Earth, Metal, or Wood). The process uses a deterministic, transparent rule engine rather than a black-box neural network, ensuring that every archetype assignment can be logically traced back to classical TCM principles.
Transparency requires absolute clarity about what VeraTongue is—and what it is not.
© 2026 Offline Human Mode (OHM). Sovereign Technology in Your Pocket.