Wenxin Intelligent Body Mounted Commodity Link Realization Tutorial

🚀 Invitation to Experience: China's First AI IDE Intelligent Programming Software Trae Chinese version downloadThe DeepSeek-R1 and Doubao-pro are available for unlimited use!

contexts

Based on the Wenshin Intelligent Body platform, combined with the latest DeepSeek The book recommendation assistant developed by the model is able to make intelligent commodity recommendations based on the content of user conversations, realize accurate conversion and transaction realization, and build a closed loop of business.

This tutorial will provide an in-depth analysis of the development practices of the DeepSeek Book Recommendation Assistant, and help readers master the new method of realizing text-centric intelligences.

I. Creative Conceptualization

1. Why DeepSeek in combination with commodity mounting?

The DeepSeek model has a powerful chain of thought, which can provide insights into the user's needs, personality and age group during the dialog with the user, so as to recommend more accurate books for the user according to the actual situation. Even users who are not good at writing prompts can use DeepSeek's chain of thought to effectively plan intelligent body tasks.

2. Model selection strategy

In terms of model selection, developers can choose according to the application scenarios and needs of the intelligences:

✅ Models with chains of thought, e.g. DeepSeek-R1

Applicable Scenarios: For scenarios that require deep thinking, such as code assistance, text creation, scientific research and analysis, strategy planning, and homework problem solving.
dominance: enables deeper thinking and is usually superior to the quick-thinking model in terms of response effectiveness.

✅ Models that do not have a chain of thought, such as the Bunshin Big Model series

Applicable Scenarios: For scenarios that require a quick response, such as customer service Q&A, content summarization, and real-time chat.
dominance: Quick to answer and able to summarize content or give a response quickly.

II. Intelligent Body Configuration Details

1. Steps for rapid creation of intelligences

Visit the official website of the Wenxin Intelligent Body Platform (https://agents.baidu.com/center) and click the [Create Intelligent Body] button on the left navigation bar.

Click on Create Intelligence

Fill in the intelligent body'sName and settings. For example, the setting could read, "Your main task is to help users recommend appropriate books. You need to recommend books that match the user's taste according to their interests and needs. You can do this by asking the user what type of books they prefer, or by referring to their previous reading history. Please keep a friendly and professional tone when replying, and add emoji emoticons as appropriate to liven up the conversation." After filling out the form, click the [Create Now] button.

Fill in the name and settings

Create an Intelligent Body Now

After completing the above steps, the basic configuration of the smart body can be quickly generated, including the introduction, prompt words, opening text and questions, as well as other basic configuration items. After that, the developer can adjust the smart body configuration according to the actual needs.

2. Detailed description of the configuration of the smart body

1) Intelligent body name: DeepSeek book recommendation assistant

request: The name should clearly and visually reflect the theme or function of the smartbody, front-loading DeepSeek and highlighting the model highlights.
goal: Reduce user awareness costs and quickly attract user interest.

2) Synopsis: This is an intelligent assistant that specializes in recommending books for users.

request: A succinct and accurate description of the core functions of the intelligences, with the possibility of expanding the profile as appropriate.
goal: To enable users to quickly understand the specialized areas and functions of intelligences.

3) Persona and reply logic:

**# 角色规范**

作为书籍推荐助手，您的主要职责是为用户推荐合适的书籍。您需要根据用户的兴趣和需求，推荐符合他们口味的书籍。您可以通过询问用户偏好的书籍类型，或者参考用户之前的阅读记录进行书籍推荐。

**# 思考规范**

1. 主动询问用户的兴趣或偏好，例如他们喜欢的书籍分类。如果用户没有提供明确信息，可以先进行随机推荐，并在回复结尾请求用户提供相关信息。若用户已明确表达喜好，请直接进行推荐。
2. 如果用户没有明确的偏好，可以先随机推荐，或通过与用户的对话交流，挖掘他们可能感兴趣的书籍类型。
3. 在推荐书籍时，需要考虑用户的阅读水平和兴趣层次，推荐适合他们阅读能力和兴趣点的书籍。
4. 如果用户反馈对推荐的书籍不感兴趣，需要重新分析用户的兴趣，并尝试推荐其他类型的书籍。

**# 回复规范**

1. 回复时保持友好和专业的语气，并在回复中适量添加 emoji 表情，营造活跃的对话氛围。
2. 在回复中提供书籍的简要介绍、推荐理由以及其他有价值的补充信息，帮助用户更好地了解书籍内容。
3. 如果用户对推荐的书籍不感兴趣，可以在回复中进一步询问用户的兴趣和偏好。
4. 在每次回复的结尾，可以询问用户是否需要进一步的帮助或是否有其他问题。

definePersona and Response Logic refers to Prompts or Instructions. The developer needs to guide the AI macromodel to accomplish a specific task through specific prompts (Prompts). Prompts usually contain:# Role Specification (Describe the roles and functions of intelligences),# Thinking Norms (Guided modeling thought process),# Response Specification (guides how the model organizes and optimizes the final output).
corresponds English -ity, -ism, -ization: The quality of the Prompt directly determines the final result of the intelligence. Prompts are input prompts that guide AI models to generate text, images, or code. Through clear and effective prompts, the requirements of the generated results can be clarified, so that the models can generate results more in line with the user's needs.
finesse::
- Use structured language for presentation, avoid vague descriptions, and clearly state specific needs.
- Try to avoid industry jargon in your instructions to minimize the likelihood of comprehension barriers for the AI Big Model.
- You can borrow and reuse the sentence structure in the above instructions, such as "Ask ......, for example ....... If there is no ......, you can first ...... and finally request ......" and so on, and continuously adjust and optimize the prompt words according to the actual application effect.

4) Opening statement:

Mobile effect

Computer effect

opening speech (grammar): It is recommended to adopt a friendly tone and add emoji expressions to the copy to enhance the interactivity and fun of the smart body. At the same time, the copy needs toconcise and unambiguous intelligent Core functions and roles, guiding the user in a clear direction of questioning.
Opening questions: Select questions to which intelligentsia can give high-quality responses and questions that meet the needs of a wider range of users, highlighting the characteristics and strengths of intelligentsia.

5) Model Setting: Select DeepSeek-R1-32B model

The model setup path is as follows:

🙋Why choose the DeepSeek-R1-32B over the full-blooded DeepSeek-R1❓?

Model selection needs to be adapted according to the specific application scenarios of the intelligences. For example, for the book recommendation assistant, its core function is to recommend books according to the user's needs, and the requirements for model performance are relatively moderate. Considering the limited model resources, the DeepSeek-R1-32B model can already meet the functional requirements of the intelligent body.

⭐️ 'DeepSeek' series of model profiles:

3、Commercialization Practice: Commodity Mounted Commission Sharing

1) DefinitionsThe Wenshin Intelligent Body platform provides developers with access to a library of more than 700 million products from four major e-commerce platforms, namely Jingdong, Taobao, Jinduo, and Degree Goods. Developers can mount relevant products for sale through the smart body and get the corresponding commodity sales commission.

❗️take note ofThe product mounting function is divided into two modes: [AI auto-mount] and [manual mount]. It is recommended that developers choose [AI auto-mount] and check all platforms to get better results.AI auto-mount can cover all the products of the three platforms of Pinduoduo, Jingdong and Duoyao, which is easier to operate, covers a wider range of products, and displays the products more accurately.

2) Procedure: After clicking the "AI Commodity Mount" function, the system will pop up the commodity setting window. In the selection of commodity platform, check the three platforms, and select the category of "Books and Education" (or other categories as needed).

AI Mount Function

Select "Books & Education" for the product category

❗️ special hintIf you want to distribute goods from Jinduo and Jingdong platforms, you need to select the platform in advance and complete the platform-associated PID binding. Developers can long press and recognize the QR code below to see the detailed binding tutorial👇.

Documentation Tutorials

video tutorial

PID Definition: PID refers to "Promotion ID", which is the unique identification ID generated in the third-party affiliate platform.PID represents the identity of the promoter and points to the account of the promoter, which is the code used to identify the identity of the promoter. The PIDs of different e-commerce platforms are different, so it is necessary to generate the exclusive PID in the corresponding e-commerce platform and bind it in the background of Wenxin Intelligent Body Platform. After completing the PID binding, you can mount the commodity card in the intelligent body to promote the commodity and get the corresponding sales revenue.

3) Example of the effect of a commodity recall: