Viking AI Search is built upon the core philosophy of "Integrated Search, Recommendation and Q&A." By seamlessly weaving these three pillars into the end\-to\-end user journey from initial discovery to final conversion, we deliver an intuitive and intelligent interaction experience. This tutorial provides a **comprehensive, step\-by\-step roadmap** to building your own AI\-powered application from scratch. Follow our instructions to quickly master the core configuration and deploy your solution. ![图片](https://p9-arcosite.byteimg.com/obj/tos-cn-i-goo7wpa0wc/2a0aa50c30a141f8a1abaa2167d81a8b =2432x) :::tip \*说明 **Note**: To enable recommendation features, you must upload a user behavior dataset alongside your item dataset. For detailed specifications on behavior data, please refer to the "Step 4: Configure behavior Dataset Fields" guide below or the "[Behavior Dataset Introduction](https://www.volcengine.com/docs/85296/1963593?lang=zh)" section in the help documentation. ::: # Quick Application Creation and Dataset Configuration ## Process Overview Following the activation of the Viking AI Search Engine, you will navigate through a structured workflow: **creating an application, importing item datasets, and defining field configurations**. To enable and evaluate recommendation capabilities, simply upload and configure the corresponding behavior datasets. Once these foundational steps are completed, you can begin exploring the platform's core functionalities. ### Step 1: Create Application The first step is to create your AI Search application. This serves as the ** ** foundation for all subsequent data upload and configurations.

### Step 2: Import Item Dataset Once your application is created, the next step is to upload your Item Dataset to activate the engine’s capabilities. We will demonstrate how to create a new item dataset and import your item dataset to familiarise you with the end\-to\-end workflow.

### Step 3: Configure Item Dataset Fields In this stage, you will review the system\-generated data configuration, define field purposes, assign unique identifiers, and validate configuration before finalising the setup. > **Note:** Processing may take several minutes depending on the data volume.

:::tip \*说明 **Data Upload Instructions:** * ++Unique Identifier++ \- Each record must include a unique **String ID** to serve as the primary key for indexing and retrieval. * ++Schema Consistency++ \- Ensure consistent data types for **fields with the same name** across multiple datasets to prevent ingestion errors. * ++Nested Data++ \- The system supports hierarchical structures up to **3\-layer depth**, enabling efficient indexing and retrieval of complex nested attributes. * ++Image Data Upload++ \- You can upload multiple image fields of **String or Array ** type; the content should be image URLs (or URL lists). Ensure URLs are publicly accessible. * ++Image Processing++ \- Images must be passed via URLs within a dedicated JSON property. To ensure optimal performance, keep file sizes **under 5 MB**. ::: ### Step 4: Configure behavior Dataset Fields Upload your user behavior dataset and define the fields. (You may choose to link to an existing behavior dataset as well) This is a pre\-requisite to enable Recommendation features.

### Step 5: Test Functions After successfully uploading your data, click **"Experience"** to enter the interactive staging environment. The experience portal offers a comprehensive suite of features to test your application: * **Core Capabilities:** Toggle among Search, Recommendation, Q&A modules to see the engine in action. * **Development Tools:** Access API code snippets, manage datasets, and review real\-time call logs. * **Simulation & Tuning:** Preview results in a simulated UI and apply various strategies to fine\-tune your ranking logic.

This concludes the quick\-start guide for application creation. For deep dives into specific feature configuration or API integration, please refer to the respective configuration sections in the help documentation. :::tip \*说明 * To learn more about Viking AI search product capabilities, please see: [Viking AI Search Engine: Integrated Enterprise Search, Recommendation, and Q&A Intelligent Engine](https://www.volcengine.com/product/AI-Search-Rec) :::