KognitiveKompanion

What is KognitiveKompanion?

KognitiveKompanion is a modern, versatile AI interface for KDE and other desktop environments, designed to provide seamless interaction with multiple AI backends including OpenAI, Ollama, and AMD Ryzen AI hardware acceleration.

How to use KognitiveKompanion?

Clone the repository, install Python dependencies with pip install -r requirements.txt, and run the appropriate launch script for your chosen backend: ./run_openai_gui.sh (OpenAI), ./run_kde_ai_interface.py (Ollama), or ./run_ryzen_ai_model.py (AMD Ryzen AI). See MULTI_BACKEND_GUIDE.md for detailed configuration.

Key features of KognitiveKompanion

• Multi-backend support: OpenAI API, Ollama, and AMD Ryzen AI
• Advanced UI with collapsible sections and conversation sidebar
• Screen capture for visual context and audio input support
• RAG (Retrieval-Augmented Generation) toggle and conversation management
• KDE theming integration, system tray icon, and floating window option
• Save, load, and export chat history

Use cases of KognitiveKompanion

• Interact with GPT-4o or other OpenAI models from a native desktop interface
• Run local LLMs privately via an Ollama backend
• Leverage AMD Ryzen AI hardware for accelerated inference
• Capture screen content to provide visual context to the AI
• Manage multiple conversations and export chat logs

FAQ from KognitiveKompanion

Which AI backends are supported?

KognitiveKompanion supports OpenAI API (GPT-4o, GPT-3.5-Turbo, etc.), Ollama for local models, and AMD Ryzen AI hardware acceleration.

Does KognitiveKompanion require KDE Plasma?

No, KDE Plasma is optional; the application works on other desktop environments as well.

How do I configure a backend?

Detailed setup instructions for each backend are provided in MULTI_BACKEND_GUIDE.md. For AMD Ryzen AI, see README_RYZEN_AI_SETUP.md.

What are the runtime dependencies?

Python 3.8+ (Python 3.12 recommended), PyQt5, and at least one backend (OpenAI API key, running Ollama instance, or compatible AMD hardware).

Does KognitiveKompanion integrate with MCP?

Yes, the application includes an MCP (Model Control Protocol) client in its app_root/mcp directory, as documented in MCP-INTEGRATION-GUIDE.md.