olace

[ local ai ]

Ollama

Ollama is a local model runtime with a large catalog of open models. Olace installs it for you, starts it on demand, tunes it for your hardware, and makes it reachable from every device you pair.

Install

Easiest: Settings › Local AI in the app, or olace setup --provider ollama. Manual install works too:

# Linux
curl -fsSL https://ollama.com/install.sh | sh

# macOS
brew install ollama

# Windows
winget install Ollama.Ollama

Olace talks to Ollama at http://localhost:11434 (configurable via ollama_url in daemon.json).

Models

Ollama model tags use a colon, like granite4.1:3b or qwen3.5:4b. Browse the full catalog at ollama.com/library, or use the Model hub in the app, which curates it and labels what fits your hardware. From the terminal:

olace pull qwen3.5:4b
olace list
olace rm qwen3.5:4b
olace pull qwen3.5:14b --device home-pc   # install on a paired host, remotely

Managed for you

  • Starts on demand. The daemon starts Ollama when a request needs it. You do not have to keep anything running.
  • Keep-alive. Models stay loaded in GPU memory between messages. Tune how long with olace settings keep-alive <minutes|always>, or free memory instantly with olace unload.
  • Auto-optimize. Olace applies Flash Attention and KV cache quantization where safe, which roughly doubles the context that fits in the same memory. On Linux, if Ollama runs as a root-owned system service, this needs a one-time approval: olace settings auto-optimize on --approve-system-service.

Context sizing

Every model gets its own context window, sized from what the model costs and what your GPU can spend, then tuned as Olace learns what actually runs on your machine. See Hardware & context.

Running more than one runtime

Ollama, LM Studio, and llama.cpp coexist cleanly. Models from all of them show up in one library, and each runtime's context sizing is tuned independently.