Nebula is a new Network Institute initiative that aims to provide researchers from all departments with safe, seamless access to open-source large language models (LLMs). Nebula is one of the first fully-localized Large Language Model (LLM) deployments in the Netherlands—and likely across the EU.
Developed to support researchers at VU Amsterdam, Nebula is designed to foster a research ecosystem that is:
✅ Fully Replicable: Ensuring that AI-powered research can be reproduced without reliance on proprietary, black-box models.
✅ Independent from Private Companies: Unlike typical AI deployments that depend on private companies outside the EU, Nebula enables researchers to work with an entirely open and locally hosted system, guaranteeing full control over data and compliance with EU regulations.
✅ 1500+ models: Using Ollama and Open WebUI, Nebula will have the capability to access over 1500 different LLMs of different sizes and capabilities.
✅ An Open-Source Powerhouse: Demonstrating the strength of open-source AI models, Nebula will make its source code available via the NI GitHub organization, encouraging collaboration and continuous improvement.
✅ Create RAG-like models: Users can use the Knowledge collections to create RAG-like models that get information from the files they uploaded.
Key Features of Nebula
🙏 Advice and support: Not sure about what LLMs to use for your project, or need help running the experiment? We can help!
🌐 API Access: Use your favorite programming language to send automated requests to Nebula. Feature is on request!
✒️ Markdown & LaTeX Support: Ideal for academic writing, allowing seamless integration of formulas and structured formatting.
🎤 Hands-Free Communication: Integrated voice and video call features for enhanced collaboration.
🛠️ Custom Model Building: Easily create and fine-tune models via the Web UI, with the flexibility to import and customize AI agents.
🐍 Native Python Function Calling: Researchers can enhance AI interactions with custom Python functions, seamlessly integrated within the system.
🎨 Image Generation: Supports local and external APIs for visual content generation.
🌍 Multilingual & Localized: Nebula supports multiple languages for international research collaboration as one of the Netherlands’s first fully localized AI systems.
By leveraging Open WebUI, Nebula is not just an AI-powered tool—it is a statement on the future of open, independent, and replicable research in Europe. Researchers at VU will have the power to push the boundaries of AI while ensuring full transparency, reproducibility, and data sovereignty.
The image above is a screenshot of the current prototype of Nebula showcasing the chat function using DeepSeek’s R1 1.5B model from Ollama. Currently Nebula can switch between multiple LLMs, keeps a history of all chats, and allows for searching across chats (left side). Interaction with Nebula can be done by typing, or by speaking (bottom side). System prompts and advanced parameters can be controlled using Nebula, allowing for granular customizations (right side).
Contacts: Radu Apsan, Marco Otte
Nebula is built on the following open-source projects:
This text was partially generated by Nebula 🙂
Nebula Use Cases
After getting access to Nebula by contacting Radu Apsan and Marco Otte, you can use our service for a virtually unlimited number of applications.
Here are a few main example use cases that have been tried and tested by our first users.
Everyday use cases
- Text Editing: spellchecking and rewriting text
- Information Retrieval: asking for information on different topics
- Text Summarization: summarizing information on websites, and research papers
Research use cases
- Benchmarking: pinning a variety of LLMs against each other to determine which one is better for a specific task
- Text classification: determining what a body of text is talking about
- Image processing via API: object detection (using a requested model)
Generally, LLM tools are highly versatile systems that can be very powerful in the right hands. Below, you can find a few more ideas for use cases for Nebula. Interested to try it out? Contact us! (Radu Apsan, Marco Otte)
(source: https://arxiv.org/pdf/2411.05025)
