Ollama

What is the Ollama Server?

The Ollama server is a lightweight, open-source AI model runtime platform that allows users to easily deploy various large language models (LLMs) and AI models locally without complex configuration. It acts as an "AI model manager," simplifying tedious processes like model downloading, configuration, and runtime into just a few commands—making it accessible even for AI deployment beginners.

Ollama Local Deployment Tutorial

1. Download the Installer

You can download the installer from the official ollama.ai website here.

2. Run the Installer

Do not directly click the downloaded .exe file to install (as it will default to the C drive). Instead, open the Windows Command Prompt (CMD), drag the installer into the CMD window, and enter the following command:

OllamaSetup.exe /DIR="D:\Your\Path"  

This will install Ollama in the Your\Path folder on the D drive.

3. Configure Ollama's Cross-Origin Support

Set the following environment variables:

• API service listening address: OLLAMA_HOST=0.0.0.0
• Allow cross-origin access: OLLAMA_ORIGINS=*
• Model file download location: OLLAMA_MODELS=F:\ollama\models
  (This setting is unrelated to cross-origin access but is crucial because the default model download location is on the C drive, which may fill up your C drive. Manually specify the model storage path.)

Restart your computer after configuring the environment variables.

Common Ollama Commands

1. Check Ollama Version and Start the Server

ollama version  
ollama start  

Use these commands to check the currently installed Ollama version, confirm successful installation, view version details, and start the Ollama server.

The default address for the Ollama server is: http://127.0.0.1:11434. The API Key can be arbitrarily set in the plugin settings, but the model name must exactly match the names listed by ollama list.

2. Search for Available Models

ollama search [keyword]  

Example: Search for LLaMA-related models:

ollama search llama  

This command lists models matching the keyword, including names and descriptions, to help users select the required model.

3. Pull a Model

ollama pull [model name]  

Example: Pull the llama2 model:

ollama pull llama2  

After execution, Ollama downloads the specified model from the model repository to the local system for offline use.

4. Run a Model

ollama run [model name]  

Example: Run the llama2 model and interact with it:

ollama run llama2  

After launching, enter questions or commands in the command line, and the model will return corresponding responses. You can also add parameters during runtime to adjust model behavior, e.g.:

ollama run llama2 --system "You are a professional translator. Translate the user's input into English."  

Use the --system parameter to set the model's role or task description.

5. Stop a Running Model

ollama stop [model name]  

Example: Stop the running llama2 model:

ollama stop llama2  

This command closes the running model process and frees system resources.

6. List Downloaded Models

ollama list  

This command displays all locally downloaded models, including names, sizes, and creation times, to help users manage local model resources.

7. Delete a Model

ollama delete [model name]  

Example: Delete the unused llama2 model:

ollama delete llama2  

After execution, the model is removed from the local system, freeing disk space.

Ollama Usage Recommendations

Recommended model: gemma3:1b (815MB in size). Despite its small footprint, this model is fully functional, performs well in translation tasks, and has low system load (supports CPU operation). Avoid using inference models like deepseek, as they output reasoning processes, leading to poor translation quality and speed.