Ollama
Ollama is optional. Install it on a Device to run local models, then use BarkVisor to manage those models and connect inference clients.
1. Install Ollama
Section titled “1. Install Ollama”Open Ollama in BarkVisor. If the runtime is unavailable, the page shows installation instructions and a Recheck button.
On macOS:
brew install ollamabrew services start ollamaFor other platforms, follow Ollama’s download instructions. After it starts, click Recheck.
2. Download a model
Section titled “2. Download a model”Choose the Device where the model should be stored.
Use Search the Ollama library to find models and click Download, or use Pull by name if you already know the model name. Filter catalog searches models already downloaded in your Home.
Model files stay on the Device that downloads them. Choose a model that fits that Device’s available memory.
3. Start or stop a model
Section titled “3. Start or stop a model”Click Start beside a downloaded model. If several reachable Devices have it, choose one. With only one eligible Device, BarkVisor uses it directly.
A model cannot start on a Device that does not have its files. Stop unloads it from the Device running it and asks for confirmation.
4. Connect a client
Section titled “4. Connect a client”Copy the Completions URL from the Ollama page. BarkVisor exposes an OpenAI-compatible endpoint through the Home console:
http://<device-address>:7777/v1/chat/completionsUse the displayed HTTPS address if you configured HTTPS access. Clients that ask for a base URL use the address ending in /v1.
Create an inference key under Settings → API Keys and enter it in your client. API requests send it as an Authorization: Bearer <key> header.
Use BarkVisor’s endpoint to route requests across the Home. Connecting directly to Ollama’s port 11434 bypasses that routing.