DOCSEARCHbyAlgolia
Submit Docs

Dinoki Documentation

website·Indexed 32 days ago

https://docs.dinoki.ai

Documentation for Dinoki, a native AI-powered desktop companion. Includes guides on Chat Mode, Character Mode 2.0, AI provider setup, tool integration, and a local-first architecture for offline capability.

Tokens
7K
Snippets
20
Records
70
Agent score
61%

What's inside Dinoki

  1. Overview of Dinoki AI Desktop Companion

    Dinoki is a native desktop application (SwiftUI for macOS, WPF for Windows) that provides an AI assistant interface. Key features include lightning-fast performance due to its small footprint (6MB on macOS, 69MB on Windows), persistent conversation memory, and the ability to run AI models locally for complete privacy.

  2. Overview of Dinoki AI Desktop Companion

    Dinoki is a lightweight, AI-powered desktop companion built with native frameworks (SwiftUI for macOS and WPF for Windows) to avoid the resource overhead of Electron apps. It provides an intelligent chat interface with persistent memory, support for 300+ AI models via OpenRouter, and the ability to run AI locally for complete privacy.

  3. Overview of Character Mode 2.0 features

    Character Mode 2.0 (targeting v1.4 release) transforms Dinoki characters into personality-rich desktop companions. Key capabilities include:

    • Context-aware perception: Ability to read screens, track windows, and hear audio, sharing only the minimum data required for tasks.
    • Goal tracking: Characters can create and track goals based on conversation history, with 'usefulness' defined via the system prompt.
    • Interactive feedback: Use of multiple-choice prompts to unblock tasks.
    • Autonomous tool use: Ability to run a controlled set of tools independently in parallel with the user.
    • Transparency: Users have visibility into exactly what data is sent to inference providers.

  4. Understand Dinoki's native performance and system requirements

    Dinoki is built using 100% native frameworks rather than Electron, resulting in smaller download sizes, faster startup times, and lower resource consumption.

    Performance Metrics:

    • macOS: Built with SwiftUI; ~6MB download size.
    • Windows: Built with WPF (.NET 8); ~69MB download size.
    • Startup Time: Less than 1 second.
    • RAM Usage: Approximately 100MB.
    • Idle CPU: Approximately 0%.

  5. Tools and MCP Ecosystem Integration

    Dinoki is expanding its toolbox by acting as both an MCP (Model Context Protocol) Client and MCP Server. This allows cloud-based AI tools to run in local-first environments for improved security. Planned growth includes:

    • Adding more built-in Dinoki tools.
    • Integrating with open MCP tool directories.
    • Open-sourcing selected tools for community contributions.

  6. General troubleshooting steps for Dinoki

    If you encounter issues with Dinoki, try these initial steps:

    1. Restart Dinoki (resolves most issues).
    2. Check for updates via Settings → About → Check for Updates.
    3. Restart your computer to clear system-level issues.
    4. Join the community Discord at discord.gg/dinoki for immediate support.

  7. Local-First Architecture and Offline Capabilities

    Dinoki's local-first architecture aims to provide offline power equivalent to online capabilities. This includes:

    • Running AI models directly inside the Dinoki application.
    • Support for Apple's on-device LLM.
    • Performance optimizations for smaller local models such as Qwen3.

  8. Agent Mode 2.0 Desktop Automation

    Agent Mode 2.0 is a long-term vision for a complete desktop automation suite (launching after the privacy-first architecture is complete). It is designed to act as a digital copy of the user with the following capabilities:

    • Full desktop control: Reading inputs (keyboard, mouse, screen, audio) and generating outputs (clicks, keystrokes, file edits, app actions).
    • Goal-oriented execution: Breaking objectives into actionable steps and executing them without constant supervision.
    • Context-aware decision making: Adapting actions based on app states and environment.
    • Complex workflows: Automating multi-step tasks from data gathering to document creation.

  9. Install Dinoki on Windows or macOS

    Dinoki is available as a native application for Windows and macOS. Choose the appropriate installer based on your operating system and hardware architecture.
    Windows: MSIX Installer or Portable ZIP
macOS (Apple Silicon): DMG Installer
macOS (Intel): DMG Installer

  10. Configure Osaurus for private, offline AI

    Osaurus is recommended for users prioritizing complete privacy and offline use. To set it up:

    1. Install Osaurus from GitHub Releases.
    2. Launch Osaurus and press Start (it uses port 8080 by default).
    3. In Dinoki, select "Osaurus" as the provider.
    4. Choose a model within Osaurus. Recommended models include llama-3.2-3b-instruct-4bit (fast), llama-3.2-7b-instruct (more capable, requires more RAM), or mistral-small3.2.

  11. Set up autonomous tasks in Agent Mode (Pro)

    Agent Mode allows Dinoki to work on goals independently in the background. Users must define a clear goal, select the necessary tools, and set an execution interval.
    # Switch to Agent Mode
Cmd+2 (macOS)
Ctrl+2 (Windows/Linux)

  12. Enable and set up Dinoki Tools

    To start using tools in Dinoki, follow these steps:

    1. Enable Tools: Navigate to Dinoki menu → Preferences → Tools.
    2. Connect Services: Authorize external services (e.g., Slack) via OAuth.
    3. Test: Verify functionality with a simple request like "What's the weather?"

    Note: All tools require a Dinoki Pro lifetime purchase.