Local LLMs
Run NeoStack AI with models hosted on your own machine.
Local LLMs let NeoStack AI talk to a model server running on your computer or LAN.
Best Options
| Server | Use when | Setup type |
|---|---|---|
| Ollama | You want the easiest local setup | Built-in provider |
| LM Studio | You want a desktop app for downloading and serving models | Custom provider |
| vLLM | You want fast GPU serving on a workstation or server | Custom provider |
| LocalAI | You want a self-hosted OpenAI-compatible stack | Custom provider |
| llama.cpp | You want direct GGUF model serving with llama-server |
Custom provider |
How NeoStack AI Connects
NeoStack AI needs an OpenAI-compatible chat endpoint:
/v1/chat/completionsFor custom local servers, add the server under Settings > Chat & Agents > Chat Providers > + Add Custom Provider.
Then:
- Set the provider name.
- Set the base URL.
- Disable Require an API key if the local server has no auth.
- Leave Auto-discover models from /models endpoint on when the server supports it.
- Start a chat with Local & BYOK Chat and pick the local model.
Important
Local models vary a lot. For Unreal editing, use a model/server that supports streaming and tool calls. If the model ignores tools, it can chat but it will not reliably edit assets.