CLOVA Studio usage control policy
- Print
- PDF
CLOVA Studio usage control policy
- Print
- PDF
Article Summary
Share feedback
Thanks for sharing your feedback!
The latest service changes have not yet been reflected in this content. We will update the content as soon as possible. Please refer to the Korean version for information on the latest updates.
Available in Classic and VPC
ClOVA Studio has limits on the number of API request and tokens that users can use within a given time frame. This is called a usage control policy, which is necessary to maintain the availability of the service and ensure stability. You can use CLOVA Studio within the set limits. If your usage exceed the maximum limit, you receive an error message and an error.
By applying usage control policies, malicious mass API calls to launch an attack can be prevented, and services can be protected from excessive traffic. In addition, it prevents particular users from monopolizing resources, ensuring all users can have stable availability of resources. Because the maximum limit varies depending on the model, tool, and purpose, please carefully read this guide.
Note
- The maximum usage provided in this guide indicate the maximum limits that you can request during the use of CLOVA Studio. Thus, the request amount corresponding to the maximum usage is not guaranteed. Even though you haven't reached the maximum usage limit, the processing of your request may be delayed or failed depending on the infrastructure conditions and traffic.
- The maximum usage limit may be subject to change later on. If the maximum usage is changed, the users will be separately informed.
- When you need a guaranteed certain level of usage, please use a suitable plan. For more information about plans and policies that guarantee the maximum usage, please contact us.
Application criteria
The maximum usage is calculated based on the following items:
| Classification | Description |
|---|---|
| User account | The maximum usage is calculated based on the main account. |
| Model and tool | The maximum usage is varies depending on the model and tool you are using. |
| Purpose of use | The maximum usage varies depending on whether it is a test app or a service app. |
Maximum usage
In CLOVA Studio, the maximum usage is limited by the number of queries per minute (QPM) and the number of tokens processed per minute (TPM). When both QPM and TPM are applied, if either of the two values reaches the maximum value first, an error is returned.
The description of QPM and TPM are as follows:
| Classification | Description |
|---|---|
| QPM (Queries per Minute) | The number of tasks requested on models and tools for 1 minute |
| TPM (Tokens per Minute) | The number of tokens to be processed for 1 minutemaxTokens) |
Note
The maximum usage is supported as of March 21, 2024 and may be subject to change later on.
CLOVA Studio web and test app
This section describes the maximum QPM and TPM you can use in CLOVA Studio Playground and test apps.
Maximum usage by model
The maximum QPM and TPM available for each model are as follows:
| Classification | QPM | TPM |
|---|---|---|
| HCX-003 | 60 | 10,000 |
| LK-D2 | 60 | 10,000 |
| LK-B | 60 | 10,000 |
| Other models | 60 | 10,000 |
| Tuned model | Requests are considered for the model used during tuning | |
Note
Other models such as K-C, LK-D, LE-C and HCX-002 are included but may be subject to change later on.
Service app
This section describes the maximum QPM and TPM available for each model and tool.
When calculating the maximum QPM and TPM, the maximum usage of web and test apps are included as well. Therefore, to use the specified maximum values, operate both a service app and a test app concurrently. When you operate only a service app, the result value subtracting the maximum usage of test app is applied.
Maximum usage by model
The maximum QPM and TPM available for each model are as follows:
| Model | QPM | TPM |
|---|---|---|
| HCX-003 | 240 | 50,000 |
| LK-D2 | 420 | 60,000 |
| LK-B | 360 | 60,000 |
| Other models | 240 | 30,000 |
| Tuned model | Requests are considered for the model used during tuning | |
Note
Other models such as K-C, LK-D, LE-C and HCX-002 are included but may be subject to change later on.
Manage maximum usage
If the maximum value that can be requested during the use of CLOVA Studio is exceeded, HTTP 429 code is returned, or an error message appears. This section describes the actions you can take to avoid exceeding the maximum usage.
Note
CLOVA Studio is committed to provide stable and seamless service. However, even though you haven't reached the maximum usage limit, the processing of your request may be delayed or failed.
How to manage QPM
- Check the QPM value in advance, and request within the maximum value.
- Implement and add a separate API call control function (rate limit).
- Separately add a certain time delay function (time sleep) between requests.
- If an HTTP 429 code is returned with an error message, process an exception and request again after a certain time delay.
How to manage TPM
- Check the TPM value in advance to set the number of tokens you enter for tasks and the maximum number of tokens to be used when creating the result value to the value you actually need,
- To view the number of tokens in the entered strings when calling APIs, use the token calculator APS in the Explorer menu.
- To adjust the maximum number of tokens to be used for the result value when calling APIs, edit the value in the
maxTokensfield. - To view the number of tokens of the entered strings in the Playground menu, click the Calculate icon at the top of the Playground screen.
- To adjust the maximum number of tokens to be used when generating a result value in the Playground menu, edit the Maximum tokens field located on the left side in the Playground page.
Was this article helpful?