Llanite guide
Local AI agent with Ollama
Ollama makes it easy to run large language models locally, but a model server on its own is not an agent. An agent needs a prompt loop that decides when to call tools, tools it can actually use, and permissions that control what those tools are allowed to do. Llanite adds that layer on top of Ollama in one install command — you get the full stack without wiring anything together manually.
What Ollama provides
Ollama downloads and runs large language models locally. It manages model files on disk, handles GPU acceleration (Metal on Apple Silicon, CUDA on Nvidia), and exposes a local HTTP server on port 11434 with an OpenAI-compatible API. Ollama is excellent at the model serving layer — fast downloads, good hardware support, and a simple CLI. Most Llanite stacks use Ollama as their runtime.
What's missing without an agent layer
Ollama gives you a local model you can send prompts to. It does not provide a loop that decides when to call tools, a set of tools the model can use (like reading files or running shell commands), permission controls for those tools, or a terminal UI for agent sessions. If you've used Ollama directly and found yourself wanting the model to actually do things — edit a file, run a test, navigate a repo — that's the agent layer Llanite adds.
How Llanite adds the agent layer
A Llanite stack declares the full runnable setup: the Ollama model, the agent loop that drives the session, the tools available to the agent, the permission policy for each tool, and the hardware requirements. You install the whole thing with one command and can inspect every component before it runs.
$ llanite inspect local-coderSetting up Ollama and Llanite together
Install Ollama first, start it, then install Llanite via npm. The llanite doctor command confirms both are working before you install a stack.
$ npm install -g @llanite/cli$ llanite doctor- 1Install Ollama from ollama.com
- 2Start Ollama: ollama serve
- 3Install Llanite: npm install -g @llanite/cli
- 4Confirm the setup: llanite doctor
- 5Browse available stacks: llanite stacks
Running your first agent session
Once a stack is installed, llanite run opens a terminal UI agent session. The status line shows the active Ollama model, the agent, and the workspace. Type your first prompt and the agent will use its tools to work through the task.
$ llanite install standard$ llanite run standardSwapping the Ollama model in your stack
Every component of a Llanite stack can be swapped with a single command. If you want to try a different Ollama model in your stack, set it and run the doctor check to confirm the new model is available.
$ llanite set local-coder model qwen3:14b$ llanite doctor local-coderFAQ
Do I need to configure Ollama separately before using Llanite?
Just install Ollama and start it with ollama serve. Llanite handles the model pull automatically when you run llanite install. You don't need to manually run ollama pull for stacks that Llanite manages.
Which Ollama models work best with Llanite?
Llanite's pre-configured stacks use Qwen 2.5 Coder for coding tasks and Gemma 3 for chat. These models are chosen because they produce reliable tool calls with the agent layers Llanite ships. You can swap to any model available in Ollama with llanite set.
Can I use a model I already have in Ollama?
Yes. If the model is already pulled in Ollama, Llanite uses the cached version and skips the download. You can also set any stack to use a model you already have with llanite set <stack> model <model-id>.
What if Ollama is running on a different port?
Llanite targets Ollama's default port (11434). If you run Ollama on a custom port, llanite doctor will report a connection failure. The runtime URL can be configured per stack.
Does Llanite use Ollama's OpenAI-compatible API?
Yes. Llanite communicates with Ollama through its OpenAI-compatible local endpoint, which means any model Ollama serves through that API is usable with Llanite.
How do I update the Ollama model in my stack?
Use llanite set <stack> model <new-model-id> to swap the model, then llanite doctor <stack> to confirm the new model is available in Ollama. If it isn't cached yet, run ollama pull <model-id> first or let llanite install pull it automatically.