Orchestrator AI 🦕

Orchestrator is a powerful, agentic AI code editor for Linux, built with PyQt6 and powered by Ollama. It provides a seamless development experience with an integrated AI agent capable of understanding project context, editing files, and running terminal commands.

✨ Features

• Agentic AI Core: A ReAct-based agent that reasons, acts, and observes. It can create/edit files, list directories, and execute shell commands.
• Google Gemini Integration: Use Google's latest models as the primary brain. Configure your API key in settings for a supercharged experience.
• Provider Selection Toggle: Easily switch between local Ollama and cloud Gemini without manual configuration changes.
• Enhanced Chat Rendering: Support for LaTeX math formulas (Arithmatex) and Syntax Highlighting (CodeHilite) with Monokai styling.
• Brave Search Integration: Connected to the internet via Brave Search API. Use the agent to find latest docs, news, or technical solutions.
• Workspace Management: Open entire folders as projects. The file tree sidebar allows easy navigation and real-time filesystem monitoring via watchdog.
• Integrated Terminal: Run commands directly within the editor with real-time output.
• Model Memory Optimization: Automatically unloads models from GPU/RAM upon application closure, freeing up system resources.
• Robust JSON Parsing: Advanced agent response extraction with structural repair logic to handle truncated or malformed JSON from models.
• Session Persistence: Automatically restores your last workspace, open tabs, window layout, and preferred Ollama model.
• Agent Log Panel: Real-time activity log in the bottom right, monitoring "Neural Thoughts", Actions, and Sensor Data.
• Dynamic Agent Profiles & Settings: Create and edit multiple agent personalities via the new "Edit -> Settings" menu.
• Chat Aesthetic: Premium "Deep Space" visual identity with vibrant cyan/purple accents, responsive chat, and glowing capsule-style input.
• Model Selector: Switch between local Ollama models on the fly.

🚀 Getting Started

Prerequisites

• Python 3.10+
• Ollama: Install Ollama for local inference.
• Google AI SDK: google-genai for remote Gemini support.
• Required Models: Local (qwen3.5:4b) or Remote (gemini-2.0-flash).

Installation

1. Clone the repository:

   git clone <repo-url>
   cd orchestrator

2. Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

   Note: Ensure PyQt6, ollama, and watchdog are installed.

Running the Editor (Development)

./venv/bin/python3 main.py

📦 Building the Executable

You can generate a standalone executable for Linux using the provided build script:

chmod +x build_executable.sh
./build_executable.sh

The resulting binary will be located in the dist/ folder as Orchestrator.

🛠️ Architecture

• core/:
  • agent.py: The AI brain (ReAct loop).
  • config.py: Centralized configuration management and persistence.
  • tools.py: File and command execution tools.
  • workspace.py: Context snapshot and file management.
  • session.py: JSON-based state persistence.
  • ollama_client.py: Wrapper for local Ollama API.
  • gemini_client.py: Client for the latest Google GenAI SDK.
• ui/:
  • main_window.py: Layout integration and signal orchestration.
  • settings_dialog.py: Multi-tab interface for managing agents and tools.
  • editor_widget.py: Tabbed code editor with syntax highlighting.
  • file_tree.py: Sidebar for project navigation.
  • chat_widget.py: Specialized chat with thought/action/observation bubbles.
  • terminal_widget.py: Integrated shell access.

⌨️ Shortcuts

• Ctrl+S: Save current file.
• Ctrl+Shift+O: Open folder (Workspace).
• Ctrl+,: Open Settings Dialog.
• Ctrl+W: Close current tab.
• Ctrl+Tab: Switch between tabs.
• Enter: Send message in chat.
• Shift+Enter: New line in chat.

📄 License

GPL-3.0 License. See LICENSE for details.

---

We are Okan Team