- Print
- PDF
Manage tasks on Playground
- Print
- PDF
Available in Classic and VPC
Managing tasks on Playground describes how to run and save tasks from the Playground menu, view a URL to share your saved tasks, and create a test app.
Run tasks in chat mode
To use chat mode in Playground, follow these steps:
From NAVER Cloud Platform console, click Services > AI Services > CLOVA Studio in order.
Click the My Product menu, and then click the [Go to CLOVA Studio] button.
Click the Playground menu.
- You can also click the [Start Playground] button in the function introduction area on CLOVA Studio home screen.
Set the parameter.
- For more information on the parameter, see Parameter.
- It is recommended to set the top P value among the parameters between 0.8 and 1.0 and the top K value at 0 except for special cases.
Enter a task title.
In the system area, enter specific directives or examples of what you want the model to perform.
On the right side of the screen, enter a question and answer.
- User: enter the message you expect the user to ask
- Assistant: enter an answer to the user's question
NoteTo break a line, press [Enter].
If necessary, click the [Add conversation turn] button to add more questions and answers.
Click [Run].
- If you are unsatisfied with the run result, clear the contents and run again to generate new result values or edit the example and run again.
Proceed with the following steps according to the desired function to perform:
Run tasks in general mode
To create a new task in general mode, follow these steps:
- From NAVER Cloud Platform console, click Services > AI Services > CLOVA Studio in order.
- Click the My Product menu, and then click the [Go to CLOVA Studio] button.
- Click the Playground menu.
- You can also click the [Start Playground] button in the function introduction area on CLOVA Studio home screen.
- Set the parameter.
- For more information on the parameter, see Parameter.
- It is recommended to set the top P value among the parameters between 0.8 and 1 and the top K value at 0 except for special cases.
- Enter a task title and directive.
- Directives help you contextualize the text you enter into a prompt.
- Enter an example for what the AI will perform.
- It is recommended to enter two or more pairs of samples into the prompt for better accuracy.
- Press [Enter] to create a line break.
- Enter an input statement to output the result value.
- Proceed with the following steps according to the desired function to perform:
Save Playground task
To save a task, follow these steps:
- From NAVER Cloud Platform console, click Services > AI Services > CLOVA Studio in order.
- Click the My Product menu, and then click the [Go to CLOVA Studio] button.
- Click the Playground menu in CLOVA Studio.
- Write the prompts and click the [Save] button on the top right corner of the screen.
- Enter the name and description of the task.
- Set tags if necessary.
- To add tags, select Click tags, click the desired tag to add, and select Click tags again.
- You can use tags when searching for tasks.
- Click the [Save as] button.
- If you want to save a saved task again by recalling it, the [Overwrite] button will be activated. You can overwrite and save the previously saved task by clicking the button.
- You can view your saved tasks in Profile > [My tasks].
Load Playground task
You can load a sample or My tasks you have saved to rerun or share the task, and you can also create a test app of the imported task.
To load a task to Playground, follow these steps:
- From NAVER Cloud Platform console, click Services > AI Services > CLOVA Studio in order.
- Click the My Product menu, and then click the [Go to CLOVA Studio] button.
- Click the Playground menu in CLOVA Studio.
- After clicking the [Load] button, click the task to load.
- You can only load samples or tasks that you have saved.
- Adjust the parameters, or edit the title or prompt in the loaded task, then run it again
To select a task to load from my task list.
- From NAVER Cloud Platform console, click Services > AI Services > CLOVA Studio in order.
- Click the My Product menu, followed by the [Go to CLOVA Studio] button
- Click the profile area in the top right corner of CLOVA Studio screen, and then click the My task menu.
- Click the [Explorer] tab > [Prompt].
- Search for tasks to view in Playground by selecting a tag or entering a search keyword if necessary.
- Click the task to open in Playground.
- When the preview pop-up window appears, click the [Open in Playground] button.
- Go to the Playground screen where the task is open.
- For more information on Playground, see Playground.
Share Playground task
You can share your tasks in the Playground with users in other groups. To share a task, first save the task and then click the [Share] button.
- You can view the shared URL in the share popup window.
- If you set the disclosure as Only me, the password to access the shared work is also displayed.
- You can click [Copy] to copy the URL and the password to the clipboard.
Create test app
To create the test app, follow these steps:
- Save the task you want to create a test app for.
- Click the [Test app] button.
- Enter a test app name, and then click the [Create] button.
- A test app is created, and a test app popup window appears.
- You can view the API information of the test app, and you can set whether to use an AI Filter (For more information on APIs, see CLOVA Studio API guide).
- As for the code type, curl and python are provided.
- You can click [Copy] to copy the API information to the clipboard.
- You can click [Reissue] to reissue the API Gateway key.
- Click View guide to view the AI Filter guide.
- A test app is created, and a test app popup window appears.
After creating a test app and verifying that it works properly, you can apply for a service app. To request a service app, fill out the Service app request form.
Request service app
The output that can be derived through CLOVA Studio fluctuates greatly depending on the user's input and it is difficult to make a perfect prediction. As a minimum device to prevent this, it goes through a screening process through an application before providing it to the service.
If you request a service app, you can use it after review and approval.
The application and issuance process is as follows:
- Make sure that you create a test app before requesting the service app review. Check that the AI Code of Ethics was not violated, and whether a stable service can be provided.
- Be sure to check CLOVA Studio AI Code of Ethics to understand and prevent various safety issues that may arise.
- The service app review can be requested only with a history of test app issuance.
- If there are no issues with using the test app, fill out the service app request form.
- You must create a "shared" URL for Playground and tuning tasks in advance so that the reviewer can directly check the task content during the service app review.
- To create a URL to share a task, see [Share Playground task].
- When approval is complete, you can view the service app API information in the same way as the test app by selecting the [Details] button in the app application status.
- Review and approval may take 1-3 weeks.
- Changes cannot be made when the service app is under review or after it has been approved.
See the following for a service app review:
- The service app request is approved or rejected by checking its compliance with CLOVA Studio AI Code of Ethics, whether it is a safe case, and if there are methods of response to risks or potential misuse.
- You must request the service app after checking that enough tests produced the planned results on the test app.
- The service app must not provide information that may pose direct or indirect risks to final users or is difficult to secure accuracy and fairness.
- The service app must have methods to monitor to avoid inappropriate results caused by abuse, and so on and respond to them.
- The inappropriate result is a typical example of the result of curses and other expressions with profanity being output, and also includes the results of the content that violates NAVER AI Code of Ethics and CLOVA Studio AI Code of Ethics.
- Changes cannot be made when the service app is under review or after it has been approved.
View application request status
To view the request status of test apps and service apps, follow these steps:
- From NAVER Cloud Platform console, click Services > AI Services > CLOVA Studio in order.
- Click the My Product menu, and then click the [Go to CLOVA Studio] button.
- Click the [Go to CLOVA Studio] button.
- Click the user name on the top right corner of the home screen, then click Application request status.
- Move to the app request status page.
- You can view the details of an app by clicking [Details].
- If you don't have a requested app, the [Request new task] button is displayed. Click the button to go to Playground.
AI Filter
From the point of view of complying with NAVER AI Code of Ethics, you can set the AI Filter function. AI Filter detects inappropriate output such as profanity from the test app (or service app included) created in Playground and informs the user.
Currently, AI Filter provides analysis results for inappropriate language, such as profanity, and assigns a label between 0 and 2 based on the risk level detected in the content. The descriptions for each label are as follows:
Label | Description |
---|---|
0 | High possibility of inappropriate expressions such as profanity in the text |
1 | Possibility of inappropriate expressions such as profanity in the text |
2 | Low possibility of inappropriate expressions such as profanity in the text |
Based on the results of the AI Filter analysis, you must prepare appropriate countermeasures to reduce the risk of content. For example, if the AI Filter response to an output is 0, it is safer to inform the end user that the output cannot be returned, and suggest a new input.
However, since the AI Filter is a model focused on detecting risks, false positives may occur and risk detection may be difficult depending on continuously changing environmental factors (laws, new words, contextual meaning, changes in meaning of words and sentences according to social changes, arbitrary meaning for specific words, etc.), so it cannot be a perfect safety device. If you are concerned about inappropriate output such as profanity, you should pay attention from the stage of designing the prompt to input, as well as preparing an appropriate countermeasure using the AI Filter. For cases to use a safe app with the recommendation of CLOVA Studio, see CLOVA Studio example guide.
The service limits of the AI Filter are as follows:
- The AI Filter limits the requested text to a maximum of 500 characters. Texts longer than 500 characters cannot be parsed normally.
- If there are too many unusual formats, emojis, and special characters in the requested text, it may not be analyzed properly.