Personal AI

A hyper-functional, completely free, sleek, web‑based personal AI interface powered by Ollama and Node.js. It leverages Turndown for HTML-to-Markdown conversion, supports Deep Research modes, and offers advanced settings to customize models and behaviors.

---

📚 Table of Contents

• Features
• Requirements
• Installation
• Usage
• Advanced Settings
• License

---

⚠️ Disclaimers

This AI may generate inaccurate or misleading information. Do not rely on it for critical tasks or decisions. While efforts have been made to reduce hallucinations, always verify output independently.

Additionally, a lot of memory, RAM, and CPU may be used due to the size of AI models, and you need a lot of RAM and CPU based on the models you are using. Using too big models or using many concurrent processes will take a lot of time, and it may even bug due to timeout errors, so beware.

🚀 Features

• HTML-to-Markdown: Uses Turndown (via Cheerio) to extract and convert web content to Markdown.
• Ollama Models: Integrates with Ollama to run local LLMs like qwen2.5vl:3b and deepseek-r1:1.5b.
• Deep Research Modes: Multiple research depths and breadths settings for fine-grained control.
• Customizable AI Settings:
  • System prompts
  • Primary and auxiliary models
  • Temperature sliders
  • WebSocket concurrency limits
• Responsive UI: Built with Tailwind CSS for a polished, mobile‑friendly experience.
• Different Modes:
  • Normal - Quick and easy responses
  • Deep Think - For complex tasks
  • Search - If up-to-date information is needed
  • Search + Deep Think - If complex thinking is required while searching
  • Deep Research Lite - Comprehensive searching using AI and COMPLETELY UNPAID APIs with NO API KEYS REQUIRED and ABSOLUTELY NO USAGE LIMITS.
  • Heavy Duty Deep Research - An even more comprehensive search just like Deep Research Lite, but requiring complex thinking and higher depth research.
  • Auto - Automatically chooses a mode for you based on the prompt. However, may be UNSTABLE.
• Location Support: For AI Searches regarding location, such as restaurants near you or how the weather is like.
• No Tracking: All of the AI's used here are completely local, meaning that there is absolutely no tracking (except for maybe google searches, as it is a network request; maybe your ISP can see what the AI searched up but doesn't know it was an AI).

---

📋 Requirements

• Node.js: Latest LTS version (>=18.x).
• Ollama: Installed and configured on your machine.
  • Pull the recommended models:

    ollama pull qwen2.5vl:3b
    ollama pull deepseek-r1:1.5b

---

🛠️ Installation

1. Clone the repository:

   git clone https://github.com/yourusername/personal-ai.git
   cd personal-ai

2. Install dependencies:

   npm install

3. Configure Ollama:
   • Ensure Ollama is running and models are available.
   • In the Advanced Settings modal (Settings ⚙️), verify that qwen2.5vl:3b and deepseek-r1:1.5b appear in the model lists.

---

🚀 Usage

1. Start the server:

   npm start

2. Open your browser and navigate to http://localhost:3000.
3. Interact with your Personal AI:
   • Select a mode (e.g., Normal, Deep Think, Search, Deep Research Lite, Heavy Duty Deep Research).
   • Type messages in the input area and hit Send.
   • View AI responses in the chat window.

---

⚙️ Advanced Settings

In the Settings modal, you can fine-tune:

• Custom System Prompt: Shape the AI’s behavior.
• Primary / Auxiliary Model:
  • Select from local Ollama models (e.g., qwen2.5vl:3b, deepseek-r1:1.5b, or pull new ones).
  • Adjust Temperature (0.0–2.0) for creativity control.
• Deep Research Lite:
  • Depth (1–5): How thoroughly to research.
  • Breadth (1–5): How many sources to consult.
• WebSocket Concurrency: Limit simultaneous model requests.
• Pull New Model:
  • Enter an Ollama model name (e.g., llama3:latest) and click Pull.

---

📄 License

MIT © 2025 Shayaan Hooda