Features

Built as a layered cognitive runtime rather than a single conversational block.

Vaudryllis is not positioned as “just another AI product”. It is an attempt to build a software entity with continuity, memory, presence, and a virtual brain inspired by biological organization.

🗃

Multi-layer memory

Persistent memory designed around layered storage rather than a flat rolling transcript. The goal is continuity, selective retention, and long-term coherence.

Brain-inspired runtime

Internal regulation, salience, workspace, emotional state, and background processes designed to support the system rather than decorate it.

🧍

Avatar presence

Embodied UI with gaze, state display, and runtime-linked presence. The avatar is intended as part of the system, not a static cosmetic layer.

🔊

Voice & TTS settings

Voice output can be configured depending on the chosen runtime path and available providers.

🌐

Local or online model path

Vaudryllis can operate through local models or external APIs depending on hardware, preferences, and deployment mode.

🧠

Model is not the whole mind

The LLM is mainly used for interpretation and expression. Vaudryllis aims to move continuity, internal structure, and more of the cognitive burden into the runtime itself.

🧩

Configurable stack

Model selection, tokens, API settings, and TTS parameters can be adjusted from the interface depending on the chosen mode.

🛡

Built for extension

The system is structured to welcome new capabilities and deeper internal cognition over time, rather than freeze into a static assistant format.

What makes it rare

Rare by integration, not by marketing vocabulary.

What matters is not a single miracle module. It is the closed loops between memory, internal state, presence, voice, and runtime continuity — especially with comparatively smaller local models.

Continuity over raw window dependence

Vaudryllis aims to reduce the classic “conversation hits a wall” problem by structuring continuity outside the raw context window alone.

Relational truth memory

It preserves user-declared relational truth as a stable base, instead of trying to guess an external “objective truth” behind every interaction.

Continuous affective layer

The system is designed around emotions and sentiments that persist and evolve, instead of being treated as a momentary cosmetic roleplay effect.