Build Your First AI Agent with Coze for Free

Background Introduction

Practical Website: www.coze.cn

Specifically, we start from the starting node, then move to the large model node. In the large model node, we can choose general tools and add a Bing search plugin. During input, reference the variable from the starting node and set the parameter name asinput.The detailed settings for prompts and outputs will be provided at the end of the article.

Next is the text processing node, which has four parts. This step mainly involves formatting adjustments for the text input and the three values output by the model. Here, we need to referenceshuru, as well as the model’s output variable“output”, and customize the addition of the symbol“ – ”. Similarly, we also need to reference the model node’s outputs“output1” and “output2”.

Now, let’s look at the composition branches of the upper part of the image. The original word output generates drawing prompts for the large model input, referencing the output of the preceding text processing“startingshuru”. The default settings for prompts and outputs remain unchanged.

In the text processing node, the content input before modification references the output of the previous text processing node“modeloutput’s output”. The string concatenation input is:

Copy

1.{{String1[0]}}

2.{{String1[1]}}

3.{{String1[2]}}

In the image generation node, select the generalpro model, with a ratio of1:1. Input references the previous model node“original hot word output drawing word”. Detailed instructions for prompts will be at the bottom of the article.

Next is the cut-out plugin, which needs to be searched in the plugin store for smart cut-out. Reference the image generation_ output from the above image “data”.

Now, let’s look at the composition branches of the lower part of the image. The current hot word drawing word large model node is basically the same as the upper part, but note that it references the hot word generated after processing, namely the output of the text processing nodeoutput1.

In the text processing node, the modified content input simply references the output from the previous node’s text processing.

In the image generation_ node below, reference the output of the current hot word drawing word large model, with other configurations consistent with the above image.

The settings for the smart cut-out plugin are the same as in the upper part.

Once both branches are constructed, use the canvas node to combine the outputs. This step requires manual dragging and dropping. The elements needed come from the following nodes:

• Two images from the outputs of the two image generation nodes
• Two words, the top one comes from the output of the text processing node of“the starting nodeshuru”, and the bottom one comes from the output of the model nodeoutput’s”text processing node
• Two ordinary texts, the top referencing“text processing node: content before modification”, and the bottom referencing“text processing node: content after modification”

Other important notes:

• The black box in the image is hand-drawn using the rectangle tool
• The canvas size is width1050, height1290

Finally, for the ending node, choose the output of the canvas node“data”. The response content is:


If there are no issues, you can try running and publishing it.

The cute master has compiled a collection of commonly used AI tools. Reply with ‘666’ in the backend to get it for free.If you find it helpful, please consider giving a “follow, like, and view.” Also, feel free to give me a star⭐️ to show your support and appreciation!AI agents are the trend for ordinary people in 2025! For those who want to learn about AI agents, welcome to sign up for our specially designed AI agent training camp. Early bird tickets are available now. Please scan the code to sign up and leave your WeChat contact! We will get in touch with you!