Zileo-Chat
assistance-micro-design/Zileo-Chat
Multi-agent AI chat desktop app built with Svelte 5, Rust/Tauri 2 and SurrealDB. Agent orchestration, Kanban automation, MCP tools, vision and voice dictation. Works with Mistral, Ollama and any OpenAI-compatible provider.
Instalação
npx convez-ai assistance-micro-design/Zileo-Chat1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
README
A tradução para o português deste README ainda está na fila.
Zileo Chat
Application desktop multi-agent avec interface conversationnelle intelligente.
Developed by Assistance Micro Design
Built with Claude Code by Anthropic
Beta Warning
This software is currently in beta (v0.28.0).
Before using Zileo Chat, please be aware of the following risks:
| Risk | Description |
|---|---|
| Data Loss | Database schema may change between versions, potentially requiring data migration or reset |
| API Costs | LLM API calls incur costs based on token usage - monitor your provider usage |
| Instability | Features may be incomplete, contain bugs, or change without notice |
| Security | While security measures are implemented, the software has not undergone formal security audit |
| Breaking Changes | Updates may introduce breaking changes to configurations or workflows |
Recommendation: Back up your data regularly and avoid using for critical production tasks until v1.0 release.
Description
Zileo Chat is a desktop application for orchestrating AI agents through a conversational interface. It supports multi-agent workflows with tool execution, memory persistence, and human-in-the-loop validation.
Key Features
- Guided Onboarding - 8-step first-launch setup (language, theme, API key, optional config import, getting-started deep links) honoring reduced-motion preferences
- Multi-Agent System - Create and orchestrate multiple AI agents with specialized tools and sub-agent delegation
- Kanban Supervisor Board - 4-column board (
todo / doing / review / done) with async card auto-compose into a "tasks to validate" zone, recurrence scheduler, and "Improve prompt" feedback loop driven by a dedicated supervisor agent kind - Per-Card Review Chat - On a card in
reviewordone, chat in place with the card's supervisor agent to re-run the worker, move the card (validate / send back), or attach a recurrence -- backed by a dedicated hidden workflow that persists across navigation - Prompt & Skill Versioning - Every prompt/skill update auto-snapshots a version row; restore or compare past versions from the settings UI
- Real-time Streaming - Token-by-token response display with usage metrics and prompt caching
- Tool Execution - Memory, Todo, Calculator, FileManager (with
read_imagefor vision), Skills, and sub-agent tools - Multimodal Vision - Attach images to messages (paste, picker, drag&drop) for vision-capable models, cross-provider (OpenAI / Mistral / Ollama / Custom)
- Voice Dictation - Push-to-talk transcription via Mistral Voxtral (hold mic button or
Ctrl+Shift+Space, release to insert at cursor) - Human-in-the-Loop - Approve critical operations before execution with risk-based validation
- Custom Providers - Add any OpenAI-compatible LLM endpoint alongside built-in Mistral and Ollama
- Reasoning Effort - Granular thinking control (low/medium/high, plus
xhigh"Think Max" for DeepSeek / GPT-5.x / Grok / Claude Opus families) on supported models - Background Workflows - Run multiple workflows concurrently with progress tracking
- MCP Protocol - Extend agents with Model Context Protocol servers, hardened with a Docker spawn guard, HTTP SSRF protection (opt-in private-network access), and a per-agent tool allowlist gating unattended runs
- Bilingual UI - English and French interface
Supported LLM Providers
Zileo Chat supports built-in and custom LLM providers:
| Provider | Type | Link |
|---|---|---|
| Mistral AI | Cloud API (built-in) | https://mistral.ai |
| Ollama | Local / Cloud (built-in) | https://ollama.com |
| Custom Providers | OpenAI-compatible API | Any OpenAI-compatible endpoint |
Mistral AI (Cloud)
- Requires API key from console.mistral.ai
- Supports all Mistral models (Mistral Large, Codestral, Magistral, etc.)
- Pay-per-use pricing based on token consumption
Ollama (Local / Cloud)
- Free and runs locally on your machine or via cloud endpoints
- Requires Ollama installed and running
- Supports open-source models (Llama, Mistral, Qwen, Kimi, etc.)
Custom Providers (OpenAI-compatible)
- Add any OpenAI-compatible API endpoint (Routerlab, OpenRouter, Together AI, vLLM, LM Studio, etc.)
- Full streaming, tool calling, and reasoning support
- Configure via Settings > Providers
Prerequisites
Required for First Launch
| Dependency | Purpose | Installation |
|---|---|---|
| Docker Desktop | MCP servers execution | docker.com/products/docker-desktop |
| At least one LLM provider | AI model access | See Supported LLM Providers |
Mistral API vs Le Chat Pro: The Le Chat subscription ($14.99/month) is for the web chat interface only. Zileo Chat requires a separate API key from La Plateforme with pay-per-token billing.
MCP Servers Configuration
Use Docker MCP Toolkit for MCP server management:
- Enable MCP Toolkit: Docker Desktop > Settings > Beta features > Enable Docker MCP Toolkit
- Open MCP Toolkit tab in Docker Desktop
- Browse the MCP Catalog (200+ servers available)
- Select and install desired MCP servers
- In server Overview > "Use this MCP Server" > copy the Docker configuration
Recommended: Always prefer Docker configurations over NPX/UVX for better isolation and zero dependency management.
Optional: Ollama (Local + Cloud Models)
| Step | Command |
|---|---|
| Install Ollama | ollama.com/download |
| Run cloud model | ollama run kimi-k2-thinking:cloud |
| List available models | ollama list |
Ollama Cloud: For large models like Kimi K2 (1T params), use
ollama run <model>:cloud. No local GPU required.
Build Requirements
- Node.js >=20.19.0, 22.12+, or 24+
- Rust 1.80.1+ with Cargo
node --version # >= 20.19.0
rustc --version # >= 1.80.1
Installation
# Clone repository
git clone https://github.com/assistance-micro-design/zileo-chat.git
cd zileo-chat
# Install dependencies
npm install
# Development
npm run tauri:dev
# Production build
npm run tauri:build
Tech Stack
| Layer | Technology |
|---|---|
| Frontend | SvelteKit 2.63 + Svelte 5.55 |
| Backend | Rust + Tauri 2.11 |
| Database | SurrealDB 2.6 (embedded) |
| LLM | Direct HTTP (Mistral/Ollama/OpenAI-compat) |
Documentation
Full documentation is available in the docs/ directory:
Contributing
- Fork the repository
- Create a feature branch (
git checkout -b feature/your-feature) - Commit changes (
git commit -m 'Add feature') - Push to branch (
git push origin feature/your-feature) - Open a Pull Request
Security
To report a vulnerability, please open a private issue on GitHub Security.
License
This project is licensed under the Apache License 2.0. See LICENSE for details.
Third-party licenses are documented in THIRD_PARTY_LICENSES.md.
Copyright 2025-2026 Assistance Micro Design
Licensed under the Apache License, Version 2.0
Acknowledgments
- Built with Claude Code by Anthropic
- Powered by Tauri, SvelteKit, SurrealDB
- Direct HTTP integration with Mistral, Ollama, and OpenAI-compatible providers