Build an AI‑Powered English Speaking Coach with Alibaba Cloud Function Compute

Solution Overview

The solution provides an AI‑driven English speaking practice platform that combines pronunciation training, dialogue simulation, instant feedback, and personalized scenario modeling. It uses Alibaba Cloud Function Compute to host a web app, the AI model platform (Bailian) for large‑model services, Intelligent Media Service for real‑time interaction, and ARTC for audio streaming.

Technical Architecture

The core cloud services involved are:

Function Compute (FC) : Deploys the web application.

Bailian Large‑Model Service : Provides one‑stop large‑model development and inference.

Intelligent Media Service (IMS) : Supplies audio processing, AI agent interaction, and speech synthesis.

Realtime Audio‑Video (ARTC) : Enables real‑time audio calls between the AI agent and the user.

These components are orchestrated to deliver a seamless AI‑assistant that can converse in English, assess user responses, and suggest improvements.

Creating the Bailian Workflow Application

Log in to the Bailian console and click My Applications → Add Application → Workflow Application → Create Dialogue Workflow .

In the workflow editor, rename the two default start‑node parameters to difficulty and scene to represent level and scenario.

Drag a large‑model node onto the canvas, connect it to the start node, and configure the model parameters (e.g., Model: Tongyi Qianwen‑Max, Temperature: default, Max Tokens: 1024, Search: off).

Set the System Prompt and User Prompt (e.g., ${sys.query}) and enable context as needed.

Connect the large‑model node to an End node, configure the output mode to Text, and map the model result to the next node.

Save the workflow.

Creating the Real‑Time Audio (ARTC) Application

Enable the Video Live service (billing by traffic usage) and open the Video Cloud console.

Navigate to Live + → Realtime Audio‑Video → Application Management and click Create Application .

Enter a custom instance name, accept the service agreement, and purchase the app.

After activation, note the generated AppId and AppKey for later configuration.

Creating the AI Agent

Open the AI Real‑Time Interaction console → Agent Management and click Create Agent .

Name the agent (e.g., Agent‑English) and bind it to the previously created workflow ( Workflow‑English) and ARTC application.

Submit the agent configuration.

Deploying the Web Application with Function Compute

Use the provided Function Compute template ( ai-english-coach) and configure the following parameters before creating the default environment:

Deployment Type : Direct deployment.

Application Name : Auto‑generated (default).

Role Name : Default role (grant additional permissions if prompted).

Realtime Audio‑Video AppId / AppKey : Values obtained from the ARTC console.

Agent Id : ID of the Agent‑English created earlier.

RAM Role ARN : Role used by Function Compute to access IMS.

Region : Default to China East 1 (Hangzhou).

Namespace : Default (can be customized for multiple instances).

Verification and Usage

After deployment, locate the example app’s domain name in the environment details and open it in a browser (HTTPS is required for microphone access). The app will prompt for microphone permission; once granted, you can select a scenario and difficulty level, start a conversation, and receive real‑time scoring and feedback from the AI assistant.

Note that the temporary domain provided by the Serverless Devs sandbox does not have a valid SSL/TLS certificate; for production use, bind a proper certificate to ensure secure HTTPS connections.

Summary

By following the steps above, you can build and deploy an AI‑enhanced English speaking coach that leverages Function Compute, Bailian large‑model services, Intelligent Media Service, and ARTC, offering personalized, real‑time language practice.