About Eloquim
Eloquim transforms emoji into expressive, personality-driven texts through AI-powered translation, creating a universal language that bridges hearts and minds through modern hieroglyphics (aka, emojis). It uses Gemini 3 as an expressive reasoning agent that interprets emotion, tone, and conversational context, and serverpod as the backend.
At the core of Eloquim, Gemini 3 Flash translates emoji combinations into natural language by reasoning over user persona, selected tone, conversation history, and situational context. The same emoji can produce different outputs depending on emotional intent, enabling nuanced expressions.
Gemini 3 also powers the agentic UI layer through GenUI using Google's Agent 2 UI Protocol. Each message triggers Gemini-driven evaluation of whether the interface should adapt. Tone indicators, emoji recommendations, and animated emotional surfaces are dynamically emitted by the model itself, creating a non-deterministic interface that evolves with the conversation.
Ghost translation reveals the emoji logic behind any received message, increasing transparency and trust.
Flutter handles the frontend, and Serverpod handles the scalable backend for high-frequency and cost-effective database infrastructure.
The project was built end-to-end using Antigravity.
Eloquim demonstrates how Gemini 3 enables a shift from static chat to living, expressive agents in real-time human communication.
Built With
- agent2ui
- antigravity
- docker
- flutter
- gemini
- postgresql
- redis
- resend
- serverpod

Log in or sign up for Devpost to join the conversation.