vector-databases-and-embeddings-module

vector databaseA specialized database system optimized for storing, indexing, and searching high-dimensional vector embedding

approximate nearest neighborAn ANN algorithm that trades exact recall for sub-linear query time by probabilistically organizing the embedd

chroma collectionA named namespace within Chroma that stores embeddings, associated documents, metadata, and IDs as an isolated

faissMeta's open-source C++/Python library for dense vector similarity search; implements IVF, PQ, HNSW, ONNG, and

hnsw indexHierarchical Navigable Small World graph; layered skip-list + NN graph enabling logarithmic-layer traversal th

ivf indexInverted File Index; clusters vectors into Voronoi cells via k-means; search prunes to nearest centroids then

nearest neighbor searchQuery operation returning the k closest vectors by a distance or similarity metric in embedding space.

product quantizationA vector quantization technique that splits a high-dimensional vector into subvectors, independently quantizes

sentence embeddingA dense vector that represents the semantic content of an individual sentence, produced by a sentence-transfor

chroma queryA search request containing a query vector (or text to embed) plus optional top-k and metadata filter paramete

embedding modelA trained neural network that converts raw tokens or image patches into fixed-length dense vectors in the embe

idx algorithm selectionChoice among ANN index types HNSW IVF ANNOY LSH determined by data scale query patterns and accuracy requireme

memory footprintRAM or VRAM required to hold index and vectors; quantization (PQ, int8) is the primary compression lever.

pinecone queryA read operation that returns approximate nearest neighbors to a query vector with optional metadata filtering

pinecone recordA data unit in Pinecone containing id, dense vector values, optional sparse values, and key-value metadata.

semantic searchAn information retrieval paradigm that finds results based on semantic meaning rather than exact keyword overl

vector embeddingDense numerical representation of discrete data (text, image, audio) as a point in high-dimensional continuous

bm25 sparse retrievalLexical baseline using term frequency and document length normalization; complements dense retrieval for exact

chroma embeddingA dense vector of floating-point numbers produced by an embedding function, representing the semantic position

embedding aggregationA method for combining multiple embedding vectors (from document chunks, sentences, or modalities) into a sing