Application scenarios

Vector databases enable similarity-based product search and recommendation for e-commerce. An e-commerce platform stores product images and descriptions. Users can search by image or text and receive personalized product recommendations.

Use embedding models to convert product images and descriptions into vectors stored in a vector database. When a user queries, DashVector converts the query into a vector and returns the most similar products. Based on user behavior and preferences,DashVector converts browsing and purchase history into vectors to find and recommend similar products, delivering personalized shopping experiences.

Q&A systems are a common NLP application, including Qwen, ChatGPT, online customer service systems, and Q&A chatbots. A Q&A system stores predefined questions and answers. When a user asks a question, the system matches it to the most similar predefined question and returns the answer. To build this, useDashVector to convert predefined Q&A pairs into vectors stored in a vector database. When a user asks a question,DashVector converts it into a vector and retrieves the most similar question from the database. Combined with model training, Q&A inference, and post-optimization, you can build intelligent conversational systems like Qwen and ChatGPT.

Image stock websites and social platforms host hundreds of millions or even tens of billions of images but typically offer only basic text or single-image search. With DashVector, you can represent images and text descriptions as vectors in a vector database. The service supports multimodal search patterns: text-to-image, image-to-image, and combined text-image search for precise filtering. Query inputs are vectorized for similarity matching, helping users find target images quickly.

Video monitoring systems, film and television websites, and short video apps generate massive video data. Use DashVector to convert video data into vectors. When users have a movie clip or screenshot, a video similarity search system retrieves the most similar videos by comparing content vectors. Clustering-based retrieval further improves efficiency and accuracy by grouping similar videos for faster in-cluster search.

Molecular fingerprints such as ECFP and MACCS keys convert molecular structures into vectors for storage in a vector database. Query molecules are converted the same way, and the database returns the most structurally similar molecules. This similarity-based retrieval enables molecular retrieval and screening, accelerating molecular discovery and drug design.