Objectives

• Differentiate between vector database and traditional database based on their functionality and use cases

• Operating in Chroma DB

• Understand similarity search, to retrieve relevant information from real-world datasets

• Key internal mechanisms within RAG

Introduction to Vector DB

Purpose

• Organise and support highly complex data for analysis and retrieval

Data types supported

• Social engagement signals such as likes and interactions

• Geospatial information

• Genomic information

• Images, audio and text

• Pattern and mapping data

Core capabilities

• Store data as vectors, mathematical objects defined by magnitude and direction

• Handle complex data operations including integration, transformation, reduction and cleaning

• Provide easy storage, fast retrieval and robust analysis

Similarity search and recommendations

• Find similar images, sounds and text

• Power recommendation engines

• Enable nearest neighbour and semantic search

Performance and scalability

• Distributed computing for scale and resilience

• Parallel processing to manage large datasets and speed up queries

• High-performance execution for low-latency workloads

Indexing and query efficiency

• Efficient vector indexing to accelerate similarity and range queries

• Support for hybrid indexing when combining vector and attribute filters

Machine learning and AI support

• Integrates with ML pipelines for feature storage and model serving

• Enables tasks that group, classify and suggest relationships among items

• Facilitates embedding management and continual model updates

Typical use cases

• Content search and personalisation

• Fraud detection and anomaly spotting

• Geospatial proximity analysis

• Genomic similarity and pattern discovery

• Multimodal analytics across text, image and audio

Vector Databases vs Traditional Databases

A vector database stores vectors in a multiple-dimension space, but conventional database store data in a table(2-dim);

Vector database can store text, vedio, and speech.

Vector libraries : in memory vector databases. besides libraries, using pre-configured algorithms to store and update data, finding similarity, and CRUD operations.

Vector database good at similirity search, rectrieving vecotrized data, facilitating tasking image retrieval, recommendation systems, and anomaly detection.

Vector database handling a large big data set, using distributed architecture for horizontal scaling

Vector database strong at analyzing vast data set.

Traditional relational databases use SQL to query tabular datasets and often rely on joins to expand information by linking related records across multiple tables.