LARIA - Local AI Assistant

LARIA - Premium Offline AI Companion

LARIA is a personal AI companion designed to feel warm, helpful and practical for everyday life.

Chat, think, write, brainstorm or generate ideas with an assistant that can run locally on your device — while letting you connect to powerful online models whenever you choose.

Now with smarter local model setup, Granite 4.1 support, Android home screen widgets, and experimental multilingual text-to-speech.

Offline, Online, or Private Cloud

Run LARIA locally on your device, connect to online providers, or plug into your own private-cloud AI endpoints.

Switch easily between offline, online, and self-hosted setups so each identity can use the model style that fits the moment.

Why You’ll Want LARIA

• Premium companion feel: warm, personal, and expressive, with clear boundaries.


• Multiple “moods”: create identities so your assistant matches your moment (serious, creative, comforting).


• Always close by: keep LARIA on your Android home screen with dynamic avatar widgets.


• More personal your way: tune greetings, presentation, and voice settings per identity.


• You stay in control: decide what runs offline and what uses the internet.


• Built for real life: writing, learning, planning, and creative work.

Key Features

• Offline models: including Granite 4.1/4.0, Mistral, Gemma, LFM, Phi-4, Dolphin, and Qwen.


• Device-aware setup: LARIA can suggest safer local defaults based on your phone hardware.


• Identity presets: different system prompt, avatar, and model per identity.


• Dynamic avatars: preset or custom images, with time- and weather-aware variations.


• Android widgets: dynamic avatar widgets bring your assistant right to the home screen.


• Per-identity voice settings: choose how each identity sounds, with experimental text-to-speech in English, French, Italian, and Spanish.


• Scheduled reminders: concise summaries on your schedule (news + weather).


• Advanced reminders: scheduled summaries with news sources, weather locations, and optional webhooks from your own HTTP endpoints.


• Optional web search: fetch web results only when a request needs them.


• Broader online and private-cloud provider support: connect OpenAI, Anthropic, Gemini, OpenRouter, xAI (Grok), Home Assistant, OpenClaw, or your own backend.


• OpenAI-compatible or Ollama endpoints: connect LM Studio, vLLM, LocalAI, llama.cpp server, Ollama, or your own compatible server.


• Attached-image chat: send text together with an image to supported multimodal models.


• Offline image generation: generate images on-device, with optional experimental hardware acceleration on supported devices.


• Storage controls: download/remove models and clean unused files.


• Chat history: resume and manage past chats.

Great For

• Writing: emails, posts, scripts, rewriting, tone.


• Learning: summaries, explanations, Q&A.


• Planning: lists, routines, schedules, travel ideas.


• Visual prompts: ask questions about images with multimodal-compatible models.


• Creativity: stories, prompts, worldbuilding, brainstorming.


• Tech: coding help, debugging notes, structured thinking.

Connectivity Notes

• Offline mode runs locally on your device.


• Online features require internet access and may send prompts to the selected provider, endpoint, or data source.