Service test

Available in VPC

In the test interface, you can adjust each setting configured during service creation and perform search tests accordingly.

Service test interface

Component	Description
① Service name	Name of the service currently being tested.
② Settings tab	Configuration menu provided for creating the search service.
③ Detailed settings	Detailed setting menu area for each settings tab Service information tab: Displays RAG service information under test. Data settings tab: Checks configured data information and performs indexing according to the data integration method. Data integration information Data integration settings Retrieval setting tab: Data search settings used during answer generation. Chunking method Embedding model Vector DB Generation setting tab: Model settings used during answer generation Model settings Prompt settings Parameter configuration
④ Expand/collapse	[TEST collapse/open] button: Toggles visibility of the search test area.
⑤ Test area	Displays the question-and-answer format test content.
⑥ Input area	Input field for entering questions to execute searches.
⑦ Apply settings	[Revert previous settings] button: Reverts settings to before saving [Apply changes] button: Applies changed settings.

Accessing the service test interface

To access the service test interface:

In the VPC environment of the NAVER Cloud Platform console, navigate to > Services > AI Services > RAG.
Click the Services menu.
Click the [Test] button for any service in the list.
The test interface will open.

Note

The RAG service is provided as a closed beta test (CBT) and is available free of charge without limitations during the CBT period. After the CBT period ends, usage will be charged according to the pricing policy. SLA for the service during the CBT period is not guaranteed.

Service test settings

Enter your questions in the input area at the bottom right of the test interface to check search results. You can change some settings during testing and apply the final changed values.

Service information

You cannot modify service information on the test interface.

Data settings

You cannot modify data integration information or settings on the test interface. However, if you selected NCP Object Storage as the data integration method, you can change the following:

Object Storage buckets: Object Storage buckets: displays the number of buckets created
[Refresh] button: Refreshes the file list in the selected bucket.
* Object Storage buckets: Displays the list of buckets created in Object Storage.
File list: Displays the number of data files stored in the bucket.
[Refresh] button: Refreshes the file list.
- File name: Name of the data file in the bucket
- Document: To preview the file, click [View document].
- [Add to index] button: Select index data from the file list and click the [Add to index] button to add it to the indexing list.
Indexing target: Displays the number of files added to the indexing target.
- File name: Name of the data file added for indexing
- Document: To preview the file, click [View document].
- Indexing status: Shows the indexing progress status (failed, completed, needs review, indexing, pending).
- Delete: Click [Delete] to change the status for indexing canceled.
  - Cancel deletion: Click [Cancer] to restore the file for index target list.
  - Apply deletion: Click [Apply changes] to remove the data file from the index target list.

Note

For restful API integration, copy the issued API key and service ID and upload data files directly via API to perform indexing. For information on how to call the RAG service API, see the RAG service API guides.

Retrieval settings

You can modify some of the retrieval settings on the test interface.

Chunking method

Passage: Sets the reference span of text in search results.
- You can enter a value from 256 to 4096
- Default: 512
Overlap: A setting that controls the degree of overlap between individual areas in the search results to properly separate text
- You can enter a value from 0 to half of the Passage value (0 to Passage/2)
- Default: 128

Vector DB

K: The number of sources to be displayed in the search results
- You can enter a value from 1 to 10
- Default: 5

Generation settings

You can modify the following generation settings on the test interface:

Model settings

Model name: The name of the LLM model used for generating responses
- Check model: Go to the CLOVA Studio service interface.
  - Click [Edit] to change the model name, then [Apply]
- CLOVA Studio API Key: CLOVA Studio API key for service integration
  - Click [Edit] to change the API key name, then [Apply]

Note

For more information on the CLOVA Studio service, see the CLOVA Studio overview page.

Prompt settings

An area where you enter instructions on how the service should generate responses
- Default: Used when no prompt setting is configured by the user.

Parameter configuration

Top P: Calculates the probability of generating a specific token based on how often it is generated across all attempts by the language model. When calculating, set a threshold value p, tokens whose probabilities fall below the threshold are excluded from the result
Top K: Includes the top k tokens in order of relevance and excludes all others
Maximum Tokens: The maximum number of tokens used to generate a result; 300 to 500 is recommended
Temperature: The lower the temperature, the more structured the responses. Higher values increase diversity but may reduce response quality

Caution

Using CLOVA Studio LLM models during search tests incurs charges.

Applying final settings

After completing test settings, you can revert or apply changes.

[Revert previous settings] button: Reverts settings to before saving
[Apply changes] button: Applies changed settings.

Caution

Starting search tests incurs charges based on the number of questions.