DEV Community

Cover image for ๐ŸŒ Get Started :Atlas Vector Search Use Cases, Infrastructure, and Product Catalogs
Danny Chan for MongoDB Builders

Posted on

๐ŸŒ Get Started :Atlas Vector Search Use Cases, Infrastructure, and Product Catalogs

Atlas Vector Search



Atlas Vector Search Use Cases:
๐Ÿ’ฌ Customer Chatbot
โ“ Question-Answering (Q-A)
๐Ÿ›’ Ecommerce Search
๐Ÿ‘ฅ User Recommendations
โœ๏ธ Content Generation
๐Ÿ“Š Analysis and Summary


Most Common Atlas Vector Search Use Cases:
๐Ÿ—„๏ธ Internal Knowledge Bases
๐Ÿ“„ Vectorized Documentation
๐Ÿ” JSON File to an Embedding Model
๐ŸŒณ LangChain or LlamaIndex


Atlas Triggers:
๐Ÿ‘€ Watch for any data changes in a single view


Atlas Vector Search:
๐Ÿ” Matching documents by similarity search on indexed embeddings data
๐Ÿ—“๏ธ Queries can use a vector's metadata (date created) to filter out older content


Atlas Search:
๐Ÿ” Matching keywords, chunked customer data
๐Ÿค– Fuzzy search to correct typos
๐Ÿ”ฎ Autocomplete (suggested search terms)
๐Ÿ”€ Index intersection (complex ad-hoc queries)



Infrastructure



Queryable Encryption:
๐Ÿ”’ Securing customer data
๐Ÿ”‘ Encrypt most sensitive data uniquely identifying an individual (e.g., SSN)


Multi-document ACID Transactions:
๐Ÿ”„ Integrity of data


Atlas Global Clusters:
๐ŸŒ Define single or multi-region Zones
๐ŸŒ Each zone supports write and read operations from geographically local shards
โš™๏ธ Configure zones to support global low-latency secondary reads


Atlas Online Archive:
๐Ÿ—„๏ธ Data lifecycle management
๐Ÿ”ฝ Automatically send outdated data from active databases into lower-cost cloud object storage
๐Ÿ’พ Keeping data accessible for querying
๐Ÿ’ฏ 9.995% uptime SLA


Distributed Architecture with Elastic Scale:
๐Ÿš€ Dynamically adjust database capacity
๐Ÿ›’ Based on application demand (e.g., shopping seasonality, sales promotions)



Product Catalogs



MongoDB Product Catalogs:
๐Ÿ“ฆ Diversity of different products
๐Ÿค— Benefit from flexible document data model


Challenges - Keyword Search:
๐Ÿค” Without extensive and laborious synonym mapping
๐Ÿšฒ e.g., mapping bikes to cycling or sneakers to trainers


Challenges - Recommendations:
๐Ÿง  Write complex rules-based engines to get specialized and scarce data


Solution - Product Catalog with Vector Embeddings:
๐Ÿ” Semantic meaning of products in the catalog
๐Ÿค Understand similarities and relationships between products


Benefits:
๐Ÿ” Search experience more intelligent & predictive
๐Ÿ“Š Track user click-through rates
๐Ÿ’ฐ Sales conversions from search results


More MongoDB Features:
๐Ÿ•ฐ๏ธ Time Series Collections: Ingest and store high-velocity, click-streams
๐Ÿ“Š Atlas Charts: Live visualizations of results, continuously tune and optimize business


Editor

Image description

Danny Chan, specialty of FSI and Serverless

Image description

Kenny Chan, specialty of FSI and Machine Learning

Top comments (0)