Ollama & LM Studio + AionUi
Use local models inside AionUi by pointing the app at an OpenAI-compatible endpoint—typical for Ollama (http://localhost:11434/v1) or LM Studio.
Why local models + AionUi?
- Offline-friendly workflows – Run models on your hardware when supported.
- Same cowork UI – Sessions, preview, and file tools as with cloud models.
- Hybrid setups – Switch between local and cloud models in LLM configuration.
Configuration steps
- Start Ollama or LM Studio and note the API base URL.
- In AionUi, choose a custom / OpenAI-compatible provider.
- Set base URL (example:
http://localhost:11434/v1for Ollama). - Select a model ID exposed by your server.
Details: LLM configuration guide.
Performance & hardware
Latency and quality depend on your CPU, RAM, and whether a GPU is available to the runtime. Smaller quantized models feel snappy on laptops; large uncensored models may need a desktop GPU and patience.
When a local model stalls, verify the server log in Ollama/LM Studio, ensure nothing else is saturating VRAM, and try a lower context window for interactive chat.
Privacy posture
Prompts never leave your machine if the model and data stay local—ideal for confidential code or regulated data. Remember that downloading models still hits the internet, and any future cloud fallback must be disabled explicitly in settings.