Onboarding guide

Updated at:

This guide explains the service registration flow and how to call APIs.

Onboarding flow

To register for retailbot during the public preview, follow these steps:

Step 1: Apply for the public preview

  • During the retailbot public preview, self-service registration is not available.

  • You can submit a registration request for the public preview on the Alibaba Cloud retailbot product page.

  • After the request is confirmed by both business teams, the retailbot team grants the developer access to the public preview.

Step 2: Confirm the customer solution and business requirements

  • Both teams work together to define a detailed collaboration plan based on the customer's business needs and existing capabilities. This plan includes the billing and settlement solution, a checklist of capabilities to integrate, target customers and their requirements, and the implementation and service flow.

Step 3: Integrate the system

  • You can follow the development documents to complete the service registration and debugging for the retailbot Platform as a Service (PaaS).

  • You can build and publish a Software as a Service (SaaS) product for retail businesses based on your company's existing capabilities and customer requirements.

Step 4: Deploy and configure

  • You can deploy the service based on the customer's business requirements.

OpenAPI onboarding guide

Endpoint information

Endpoint for the Chinese mainland: retailbot.cn-zhangjiakou.aliyuncs.com

Step 1: Initialize the bot instance

  • You must first create a bot for the customer. All subsequent operations require a ROBOTCODE.

  • Related API:

    • Create a bot (CreateRobot)

Step 2: Integrate business data

  • You can integrate the customer's business data, such as products, orders, users, and dialogue data, based on their multi-channel deployment needs.

  • Because the Minimum Viable Product (MVP) version supports only Taobao channels, you can skip this step.

Step 3: Subscribe the bot to built-in industry packages

  • You can subscribe the customer to the relevant industry packages as needed. By default, all customers are subscribed to the general package.

  • Related APIs:

    • Query all industry packages (ListIndustryPackages)

    • Subscribe to an industry package (InstallPackage)

    • Unsubscribe from an industry package (UninstallPackage)

Step 4: Configure merchant domain knowledge for the bot

  • The built-in industry packages may not cover all of a merchant's business questions, and a merchant's understanding of a scenario might differ from the definition in the built-in package. To improve the bot's coverage and business relevance, you can add merchant-specific domain knowledge to the bot.

  • The current version does not include merchant domain knowledge capabilities. During the public preview, we will add capabilities such as Frequently Asked Questions (FAQ), private merchant models, and platform model corrections. You can configure these capabilities using the relevant OpenAPI, visualization knowledge mining tools, and model training tools.

Step 5: Configure knowledge answers for the bot

  • You can call the knowledge-related APIs to add answers for the bot's subscribed industry packages and merchant domain knowledge.

  • Related APIs:

    • Query the list of official merchant scenarios (ListSubscribedPackageKnowledges)

    • Add a solution (AddSolution)

Step 6: Access the online Q&A service

  • You can connect to channels to obtain dialogue data. Then, you can call the retailbot real-time Q&A service to retrieve acknowledgement results.

  • Related APIs:

    • Real-time Q&A (RecognizeMessage)

    • Test window (RecognizeMessageForTest)

Step 7: Send back dialogue data

  • You can send dialogue logs and model labeling logs back to retailbot to help improve its Q&A performance.

  • Related APIs:

    • Man-machine dialogue data backflow (SyncRobotMessag)

    • Human-to-human dialogue data backflow (SyncMessage)

    • Annotated data backflow (UploadLabelData)