This topic answers frequently asked questions about the Communication Intelligence Engine to help you quickly resolve issues.
Question overview
High end-to-end perceived latency?
Perceived latency is a subjective metric. Generally, excluding network latency, the end-to-end time to first audio should be about 1000 ms to 1200 ms longer than the Large Language Model (LLM) time to first token. This range is considered normal. If the latency is significantly higher, contact us for troubleshooting. If everything is normal but you are still unsatisfied with the performance, optimize the LLM latency.
Are Alibaba Cloud Model Studio models and applications supported?
Alibaba Cloud Model Studio models can be integrated with the Communication Intelligence Engine. Currently, Alibaba Cloud Model Studio applications only support agent applications and conversational workflow applications.
Configuration errors after setting up the LLM gateway?
"code: InvalidParameter": This is a version compatibility issue. The issue occurs if you create an Alibaba Cloud Model Studio model and then change it to an application. Delete the LLM gateway and create a new one. This issue will be resolved in a future version.
"code: Required parameter(AppId) missing or invalid": This error can occur if the AppId or the API-KEY is incorrect.
The AICCS API returns a partnerId not found error
This error is usually caused by an incompatible API version. If you call the AICCS API and receive a partnerId not found error, this typically indicates that the API version you are using for development is incompatible with your business.
To resolve this issue, change the API version that you use for development from aiccs20230516 to aiccs20191015, and then call the API again.