LLM Model
Overview
The LLM Model tab configures the platform's primary AI model connection. Once configured, the platform's AI assistant (Collie) can use the large language model you specify to provide AI capabilities such as natural-language interaction, intelligent Q&A, and automated operations.
The platform connects to the LLM using the OpenAI-compatible API protocol — any model service compatible with the OpenAI Chat Completions interface (including official OpenAI, Azure OpenAI, and various domestic / self-deployed compatible services) can be connected.
Configuration Items
This tab provides three configuration items that together define the large model connection used by the AI assistant.
Primary Model Name
| Property | Description |
|---|---|
| Business meaning | The name of the large language model used by the AI assistant |
| Example | gpt-4o, deepseek-chat, qwen-max |
The model name must exactly match the model ID supported by your model service. If unsure, refer to your model service's documentation.
Primary Model Base URL
| Property | Description |
|---|---|
| Business meaning | The OpenAI-compatible API root URL of the model service |
| Format | A valid URL |
| Example | https://api.openai.com/v1, https://api.deepseek.com/v1 |
OpenAI-compatible means the model service provides the same HTTP interface specification as OpenAI Chat Completions. Most mainstream model services (DeepSeek, Tongyi Qianwen, Moonshot, Azure OpenAI, etc.) offer a compatible mode — simply point the Base URL to the corresponding service's address.
Primary Model API Key
| Property | Description |
|---|---|
| Business meaning | The API key for accessing the model service |
| Display | Shows ******** mask when configured |
The API Key is issued by your model service provider. Keep it secure and do not disclose it to unauthorized personnel.
The API Key is a password-type field — it shows a ******** mask when already configured. When saving:
- If the masked field is unchanged, the existing API Key is not overwritten
- To change the API Key, type the new value directly into the field and save
- To clear the API Key, empty the field and save
Configuration Validation
After saving, how do you confirm the AI model connection is working?
- Check whether the AI assistant is available: After a successful configuration, the AI assistant (Collie) in the platform UI should be able to converse normally
- Check error messages: If the API Key is invalid, the Base URL is wrong, or the model name does not exist, the AI assistant will return a connection error during conversation
- Review backend logs: If you have access to backend logs, check whether the model invocation succeeded
Unlike the Email tab, the LLM Model tab currently has no "Test connection" button. After configuration, you need to verify the connection by actually using the AI assistant.