How to connect Local Ollama to Crescelo.
Run Llama 3.2, Mistral, or Qwen entirely on your own hardware. Follow this step-by-step guide to download Ollama, configure cross-origin permissions, and connect your assistant.
Download & Install Ollama
Download the free Ollama installer for your operating system:
Ollama runs locally on macOS (Apple Silicon M1/M2/M3/M4 is exceptionally fast), Windows, and Linux.
Pull Your Desired Open-Source Model
Open your terminal (Terminal on Mac/Linux, PowerShell or Command Prompt on Windows) and run:
Ollama will download Llama 3.2 (~2.0 GB) and boot the local server. You can also pull larger models such as ollama pull llama3.1:8b or ollama pull mistral.
Configure CORS Allowed Origins (Crucial Step)
By default, Ollama only accepts requests originating from localhost applications. To allow your Crescelo web dashboard to send prompts to your local Ollama server, you must set the OLLAMA_ORIGINS environment variable:
launchctl setenv OLLAMA_ORIGINS "*"OLLAMA_ORIGINS → Variable value: * → Click OK → Restart Ollama.sudo systemctl edit ollama.serviceEnvironment="OLLAMA_ORIGINS=*", save, and run sudo systemctl restart ollama.Connect to Crescelo
- In your Crescelo Dashboard, go to Settings → AI Settings.
- Select Local Server (Ollama).
- Verify the host URL is set to
http://localhost:11434(or your custom server IP). - In the Model Name box, enter the model tag you pulled (e.g.
llama3.2). - Click Save AI Settings.