Model Adding

As the underlying intelligent engine of an agent, an AI large language model (LLM) must be bound to the agent during runtime. The model is responsible for core capabilities, such as natural language understanding, logical reasoning, SQL statement generation, and report generation. Through the Model module, LLMs available for agents to invoke within the platform can be configured and managed in a unified way.

Adding Entry

As a super administrator or a developer with model management permissions, choose Admin Console > Model, click the Add Model button in the upper right corner, and enter the basic model information in the Add Model pop-up box.

Basic Information

Taking how to add the deepseek-v4-flash model as an example, the following figure shows the effect.

Connection Testing and Saving

After entering the information, click Test Connection to check whether the model's chat, Tool Call, image recognition, and thinking mode functions can work properly. During the connection test, the system will check and display the thinking modes supported by the model:

• Autonomous Thinking: The model autonomously decides whether to enable the thinking mode. Users cannot manually enable or disable the thinking mode in the dialog box.

• Controllable Thinking: Users can manually enable or disable the thinking mode in the dialog box.

After the test passes, click Save to complete the model adding.

Note:

(1) A newly created model must pass the connection test before its thinking mode can be properly recognized and used.

(2) Any existing model with a new version must pass the connection test again.

(3) Before the connection test is completed, the system will uniformly use the model in the Autonomous Thinking mode. Even if the model actually allows users to manually control the thinking mode, no option to enable or disable the thinking mode will be provided in the dialog box.

(4) You can add the controllable thinking mode and multimodal LLMs.

Model Classification

Models can be classified into the following three categories based on permissions:

• My models: Models added by developers or super administrators to the My tab page, which are visible and available only to individuals

• Public models: All models added by developers or super administrators to the Public tab page, which are visible and available to all members

• User models: Models added by each user, which can be viewed and edited only by super administrators on the User tab page

Note:

Developers can create or delete public models only when they are assigned the public management permissions on models.

Model Operations

The model list displays all added LLMs and supports the following operations on existing models:

• Search: Quickly locate the target model by model name through the top search box.

• Test Connection: Hover your cursor over the target model, click ... on the right, and select Test Connection from the drop-down list to check whether the model's API connection is normal. If so, the agent can be invoked normally.

• Delete: Hover your cursor over the target model, click ... on the right, and select Delete from the drop-down list to delete this model that is no longer in use.

(1) Disclaimer: FanRuan only provides recommended LLMs and connection methods for reference and takes no liability for any issues with the LLMs themselves.

(2) A model deleted cannot be restored. Check whether the model is referenced by an agent before the deletion.

(3) The API key and endpoint are sensitive information and should be kept properly to avoid leakage.

(4) Ensure that the network environment can access the API URL of the corresponding model. Otherwise, the connection test will fail.

(5) The model name must be exactly the same as the one provided by the service provider. Otherwise, the agent invocation will be abnormal.