Use shared sample datasets to get started with DLF multimodal capabilities including image search, full-text search, and visual exploration, without preparing data or building retrieval services.
This quick start uses the search_samples dataset, which contains images, vectors, and documentation data for multimodal validation including image search and full-text search. For more datasets, see Shared sample datasets.
Preparation
-
Log on to the DLF console
In the left navigation pane, click Catalogs and go to , locate the data share named dlf_samples, and click Create Catalog.
The catalog created from a received share is read-only.
-
Click the Catalogs tab to view the newly created catalog. Its status appears as Running.
Image search
Use this mode to validate multimodal retrieval, test vector recall, and run similar scenarios. Upload a reference image, and the system automatically computes vectors based on the configured embedding column and returns the top N similar images, without the need to build your own vector retrieval service.
-
Open the Data Preview tab of search_samples.berkeley_deepdrive_100k_images, and select Image search mode.
-
Upload a reference image, such as a street view sample image.
-
Set the number of results (Top N) and compute resources, then click Run. The result set returns similar driving scene images from the Berkeley DeepDrive 100K dataset, sorted by vector similarity in descending order.
Visual exploration
Use this mode to query table data by field, verify table structure, and check data distribution.
-
In the left-side catalog tree, locate the target table, for example, search_samples.berkeley_deepdrive_100k_images, and go to the table details page.
-
Switch to the Data Preview tab, select Visual exploration mode, and configure the output fields, filter conditions, query enhancements (Group By, Order By, Limit), and compute resources as needed.
-
Click Run. The result set is displayed in a table, where you can directly preview fields such as images and tags.
Full-text search
The official documentation sample table is already configured with a full-text index. You can use the full_text_search function to search document content.
SELECT title, section, url, content
FROM full_text_search(
'search_samples.dlf_official_docs_zh',
'content',
'Expired snapshot cleanup',
3
);For detailed steps on using the data preview feature and exporting result sets, see Data discovery.