Dataset Search
Karini AI's Dataset Search provides powerful search capabilities over your vector embeddings, combining traditional keyword search with advanced AI-powered features like query rewriting, reranking, and hybrid search. The Search API enables programmatic access to these capabilities for integration into your applications.
Whether you're testing search quality interactively in the dataset UI or integrating search into production applications via the API, Karini AI provides comprehensive tools for accurate, relevant, and secure information retrieval.
Follow the steps to access the search panel and configuration options in dataset:
Navigate to your dataset: Begin by opening the dataset where you want to perform the search.
Open the Search Interface: Once you’re in the dataset, locate and click on the search icon . This icon will open the search panel, which will appear on the right side of the interface.
Configuration Panel: On the left side of the interface, you’ll find the configuration panel. This allows you to adjust settings and optimize your search queries and results.
The panel appears on the side of the dataset view, showing both search and configuration options.
The following features are available to enhance the search experience:
Query Input Field: Enter your search queries directly into the input field to initiate searches.
Multimodal Search via Image Upload: Upload images to perform multimodal searches that combine both text and visual inputs.
Real-Time Results Display: View search results immediately as queries are entered, facilitating instant feedback.
Structured Response Formatting: Results are presented in a structured, easy-to-read format for clarity and quick analysis.
Clear Chat History: Option to clear the chat history for a fresh search session.
Expand/Collapse Views: Toggle between expanded and collapsed views to manage the interface space according to your preferences.
Performing Searches
Karini AI provides powerful search capabilities for both text and images, allowing users to retrieve relevant results based on their input. Whether you're searching by text or combining text with images, the system provides flexible and efficient search methods.
Text Search
Text search allows you to find relevant content by entering a query in the search input field. You can configure settings for better precision and retrieve results based on semantic matching.
Steps:
Configure Search: Adjust settings in the left panel.
Enter Query: Type your query and press Enter or click Send.
Review Results: View the results as they load.
Example Queries:
"What is the company's revenue for 2023?"
"Explain the photosynthesis process"
Image Search (Multimodal Datasets)
For datasets containing images or multimodal data, you can perform searches based on images alone or in combination with text. This feature allows for more contextually relevant results, especially when working with visual data.
Prerequisites:
Dataset type: image or multimodal
Supported formats: PNG, JPG, WEBP, GIF (max size: 8MB)
Steps:
Click Image Icon: Attach an image by clicking the image icon.
Upload Image: Select the image file from your device.
Optional Text: Add a text query if needed.
Execute Search: Click Send to perform the search.
Search Modes:
Image Only: Find visually similar images.
Image + Text: Perform a combined search using both image and text.
Use Cases:
Image Search: Locate visually similar images within your dataset.
Multimodal Search: Retrieve results that combine both image and textual information for enhanced accuracy.
Last updated