Vizier

21st Century Digital Steward; Right-hand agent for you majesty

Vizier is a Rust-based AI agent framework that provides a unified interface for AI assistants across multiple communication channels (Discord, Telegram, HTTP, etc.) with memory, tool usage, and extensible architecture.

Features

• Multi-Channel Support: Connect to Discord, Telegram, HTTP (REST API & WebSocket), and WebUI
• AI Model Integration: Support for multiple AI providers (DeepSeek, OpenRouter, Ollama, etc.)
• Memory System: Session-based short-term memory, configurable recall depth, and vector-based long-term memory
• Tool System: Extensible tool framework including CLI access, web search (Brave Search), scheduler (cron & one-time tasks), vector memory, and workspace document management
• Scheduler: Built-in task scheduler for automated agent execution
• WebUI: Modern React-based web interface for interaction and management
• TUI Interface: Built-in terminal user interface for local interaction (WIP)
• Configuration Driven: Flexible configuration via YAML files with environment-specific overrides

Installation and Configuration

Prerequisites

No prerequisites required for standard installation. The install script handles everything automatically.

For Custom Installation (Building from Source)

• Rust and Cargo installed

Quick Start

1. Install Vizier (Recommended):

   curl -fsSL https://get.vizier.rs | sh
   

   Or install via cargo (requires Rust):

   cargo install vizier
   # Or using cargo-binstall (faster)
   cargo binstall vizier
   

2. Generate configuration and workspace:

   vizier init
   

   This will create a minimal config and sample agent to run in your directory.

3. Run the agent:

   vizier run
   

Quick Start with Python

If you need the Python interpreter tool:

# Install with Python feature
cargo install vizier --features python

# Or from source
cargo build --release --features python

Development Setup

For development, clone the repository and use the provided just commands:

# Install dependencies (Rust crates and webui npm packages)
just install

# Run in development mode with hot-reload
just dev

# Build the webui
just build

See the Justfile for all available commands.

WebUI

The web interface is built with React and served automatically when the HTTP channel is enabled. After building (just build), it will be available at http://localhost:9999 (or the port configured in your config.yaml).

Update Installed Version

Using Install Script

Simply re-run the install script to get the latest version:

curl -fFSL https://get.vizier.rs | sh

Using Cargo (if installed via cargo)

1. Install cargo-update if you haven't already:

   cargo install cargo-update
   

2. Update the binary:

   cargo install-update vizier
   

Planned Features (V1.0.0)

• Web UI (React-based interface)
• Scheduler and task system (cron & one-time tasks)
• Vector memory for long-term retention
• Brave Search integration
• Local embedding model support
• Docker Sandbox
• Simple TUI (terminal user interface)
• Additional AI providers (Google Gemini, OpenAI, Anthropic, etc.)
• Sub-agent spawning for parallel task execution
• Model Context Protocol (MCP) integration
• Skill system for reusable agent behaviors
• WASM-based plugin system
• Built-in HTTP client tool

Development

Project Structure

• src/: Rust source code (agents, channels, tools, scheduler, database, etc.)
• webui/: React-based web interface (built with Vite + React Router)
• templates/: Template files for agent configuration and identity
• .vizier/: Workspace directory for runtime data (config, database, agent workspaces)
• migrations/: Database migrations (SurrealDB schemas)

Available Commands

See the Justfile for available commands:

CLI Commands

The vizier binary provides these subcommands:

• vizier run --config <path>: Start the agent with given config
• vizier tui: Launch the TUI client (requires running agent)
• vizier init: Initialize a new vizier workspace
• vizier configure: Generate a new config non-interactively

Adding New Features

1. New Tools: Add to src/agent/tools/ and register in src/agent/tools/mod.rs
2. New Channels: Add to src/channels/ and implement the Channel trait
3. New Models: Extend the provider system in src/agent/agent_impl/provider.rs
4. New Schedules: Add to src/scheduler/ and integrate with task database

License

MIT License