Back to Site

Deploying an Ollama AI server

Connect HelpChat to local large AI models. Common Chinese-language models such as DeepSeek and Qwen are supported.

1. Install Ollama

Download: https://ollama.com/download. Choose the version for your system (Windows / macOS / Linux).

2. Set the model storage location

AI model files are very large (from several GB to tens of GB). Before downloading, make sure the model storage location is on a drive with enough free space.

Tip

On Windows, set the environment variable OLLAMA_MODELS to specify the model storage path, for example D:\OllamaModels.

3. Download AI models

Model list: https://ollama.com/search. Well-known Chinese-language models are recommended, such as deepseek-r1 and qwen3 (Alibaba Qwen).

Download models from the command line:

bash
ollama pull deepseek-r1
ollama pull qwen3

4. List downloaded models

bash
ollama list

5. Configure Ollama in HelpChat Server

Sign in to the web admin console and set the Ollama API address. The default port is 11434, in this format:

text
http://127.0.0.1:11434

Info

Replace 127.0.0.1 in the address with the server's actual IP address. Ollama can also run on a separate computer; in that case enter that computer's IP.

6. Using it in the client

After configuration, sign in to the client again to see the AI button. In the AI chat window you can view and switch between the available AI models.