Communication Intelligence Engine

Function Overview

The Communication Intelligence Engine encapsulates call processes, detects and controls events, and integrates with your large language model (LLM) to create human-like interactions. The engine uses Voice Service to apply for phone numbers. It fully manages number resources and supports automatic number rotation to improve connection rates.

Preparations

Before you use the Communication Intelligence Engine, complete the following preparations:

• Account registration and service activation

  • Complete Alibaba Cloud account registration and enterprise identity verification.

  • Activate Artificial Intelligence Cloud Call Service and Voice Service. If you plan to use large language models (LLMs) from Alibaba Cloud Model Studio, log on to the Alibaba Cloud Model Studio console to activate the service.

• Large language model setup

  We recommend that you use Qwen series models through Alibaba Cloud Model Studio, or use open-source or commercially available models from the marketplace. If you host your own LLM, you must fine-tune it for your business scenario and configure prompts before you integrate it with the Communication Intelligence Engine. Your LLM must comply with the Communication Intelligence Engine gateway protocol. For more information, see LLM Gateway Integration Protocol.

• Phone number application

  Voice number resources are managed end-to-end by Voice Service. Therefore, you must perform the required operations in the Voice Service console. Before you apply for real numbers, you must complete credential filing and script filing. You can apply for fixed-line numbers and mobile phone numbers for specific locations based on your business needs. You can also contact the operations team for the fully managed service. The fully managed service is a Voice Service feature that automatically applies for and replaces numbers based on your business call patterns to maintain your answer rate.

Procedure

1. Configure the LLM gateway: Configure LLM gateway settings to connect the Communication Intelligence Engine with your LLM. For more information, see LLM Gateway Configuration.

2. Manage LLM applications: Create an LLM application, associate it with the LLM gateway, and configure call settings. For more information, see LLM Application Management.

3. Create a speech-to-text custom vocabulary (optional): Create a custom vocabulary and associate it with your LLM application. For more information, see Speech-to-Text Custom Vocabulary.

4. Create a custom voice (optional): Create a custom voice, associate it with your LLM application, and configure the voice for calls. For more information, see Custom Voice.

5. Configure callback messages: Subscribe to the MNS message queue consumption mode or the HTTP batch push mode to receive callback messages after a call ends. For more information, see Callback Message Configuration.

6. Initiate a call: After you complete the configuration, call the API to initiate a call. For more information, see LlmSmartCall – Intelligent Calling Based on Large Language Models.

7. Hang up a call manually: After a call is initiated, you can call the API to manually hang up the call. For more information, see HangupOperate – Manually Hang Up a Call Using the Communication Intelligence Engine.

Billing

The Communication Intelligence Engine is billed at CNY 0.30 per minute. Calls that last less than one minute are billed as one minute.

Experience the Results

The following recording demonstrates this feature using the Alibaba Cloud Qwen model. Play the audio file to listen to a sample call.