Integrate Alibaba Cloud Model Studio applications

Step 1: Create an Alibaba Cloud Model Studio application integration

1. Log on to the Mobi platform and go to the console.

2. Navigate to Resources -> Integration and click Create Integration.

3. Select Alibaba Cloud Model Studio Agent Application and enter a name and description for the integration.

4. Configure the connection information for the Alibaba Cloud Model Studio application for different environments, including the API-KEY and Workspace.

5. Click OK to create the integration.

Step 2: Create an integration operation within the application

1. Open the Code panel, click + in the corresponding scope, and select Integration Operation.

2. Select the Alibaba Cloud Model Studio agent application integration that you created as the resource. Enter the application ID and prompt. Then, select a trigger method and configure other settings as needed.

3. Click Run to preview the query results.

General configurations

• Parameters: Defines the parameters required for the query, including the type, default value, and description. In the search statement, you can reference parameters using {{ varName }}.

• Enable SSE response: Specifies whether to use streaming data transmission for each conversation turn.

• Application ID: The ID of the Alibaba Cloud Model Studio agent application to call.

• Prompt: The input prompt content for each conversation turn.

• Session ID: The unique identifier for the conversation history. If you pass a session ID, the conversation history is recorded in the cloud. Calls to the Large Language Model (LLM) will automatically include the stored conversation history.

• Display retrieval process: Specifies whether to output information about the retrieval process. If enabled, the process information for document retrieval and model inference is returned.

• Knowledge base retrieval scope:

  • Knowledge base retrieval scope: A list of knowledge base IDs to search when using a retrieval-augmented generation (RAG) application. Before using this parameter, ensure your application is a RAG application and does not have a knowledge base configured on the Alibaba Cloud Model Studio platform. For more information about how to obtain a knowledge base ID, see Create and use a knowledge base.

  • Temporary file ID list: A list of temporary file IDs to upload in the current request of the agent application. For more information about uploading temporary files, see Application calls.

• Image URL list: An array of image URLs to pass to the Model Studio agent application. Before using this parameter, ensure the model used by your Alibaba Cloud Model Studio agent application supports image input, such as the Qwen-VL model.

• Custom parameters: Custom parameters to pass to Alibaba Cloud Model Studio workflow applications and agent orchestration applications. If you use an agent application to pass custom user parameters, use user_prompt_params to identify the custom prompt variable parameters.

• Workflow mode: For a Model Studio workflow application, you can use this parameter to control the different output modes of the application. For more information, see the workflow application output modes in the streaming output section of Application calls:

  • full_thought: The streaming results of all nodes are output in the thoughts field. Set has_thoughts to True.

  • agent_format: Uses the same output mode as the agent application. In the console application, enable the Return Result switch for a specified node. The streaming result of that node is then output in the text field of the output object.

• Optional parameters: Defines additional optional parameters.

  • Model name: Replaces the original model in the agent. This parameter is supported only by agent applications. For a list of supported model names, see Supported models.

  • Enable web search: If enabled, the model uses search results as reference information when generating replies. However, the model decides whether to use the web search results based on its internal logic. Enabling web search might increase token consumption. This parameter is supported only by agent applications.

  • Automatically obtain current time: If enabled, the model can directly respond to real-time requests, such as "What is today's date?". This parameter is supported only by agent applications. It is enabled by default.

  • Enable thinking mode: This parameter switches between thinking mode and non-thinking mode in the Qwen3 model. It is effective only for the Qwen3 model and has no effect on QwQ or DeepSeek-R1 models. This parameter is supported only by agent applications. It is disabled by default.

  • Number of context turns: Sets the maximum number of historical conversation turns for the input model. A higher number of turns results in stronger conversation relevance. This parameter is supported only by agent applications.

• Trigger method: You can select a manual trigger or automatic run. A manual trigger executes only when an integration operation event is triggered or when the action.trigger method is called through an API. An automatic run executes automatically on the initial load and when dependencies change. For example, if the integration configuration references data from the page input component `input1`, the integration automatically triggers and updates data when {{input1.value}} changes.

• Timeout: The maximum running time for the operation.

• Description: A detailed description for the integration operation.

Callback configuration

• Data transformation: Processes the data returned by the interface. The default is return data, which means no processing is performed.

• Streaming receive callback: Configures the event triggered when chunk data is retrieved from a streaming API call.

• Success callback: Configures the event triggered when the API call is successful.

• Failure callback: Configures the event triggered when the API call fails.

Configuration with the chat component

For more information about the chat component, see Chat.

If the integration operation for the Alibaba Cloud Model Studio application is named bailianapp and the chat component is named chat1, select bailianapp in the AI integration operation for chat1.

In the integration operation for bailianapp, configure the following parameters to accept input from chat1:

• Prompt: Use {{chat1.currentMessage.content}} to accept the current user text input from chat1.

• Image URL list: Use {{chat1.currentMessage.image_list}} to accept the current user image URL list from chat1.

Advanced configuration

• Debounce: You can set a debounce expression to prevent frequent API calls.

• Do not run if: Determines whether to execute this integration operation based on a conditional expression.

Event trigger

1. In the component that needs to trigger the integration operation, configure an event handler.

2. Set the action type to Integration Operation and select the corresponding integration.

3. Enter the configured parameters as an object.

API trigger

If the integration operation is named action1, you can use the action.trigger() method to trigger the operation from any location where scripts can be executed. If the integration operation has parameters configured, you must enter them as an object.