Technical guides
Vector Search Guides
Focused engineering notes for non-brand search intent: vector search comparisons, ANN index tradeoffs, cosine similarity, RAG chunking, retrieval failures, and vector database interviews.
01 / vector search vs keyword search
Vector Search vs Keyword Search
Compare vector search and keyword search, including embeddings, BM25, exact tokens, semantic recall, hybrid search, and production RAG tradeoffs.02 / semantic search vs vector search
Semantic Search vs Vector Search
Understand the difference between semantic search as a product behavior and vector search as one retrieval technique using embeddings and nearest-neighbor indexes.03 / HNSW vs IVF
HNSW vs IVF
Compare HNSW graph search and IVF clustering for vector databases, including latency, recall, memory, filtering, updates, and production failure modes.04 / FAISS vs Pinecone vs Qdrant
FAISS vs Pinecone vs Qdrant
Compare FAISS, Pinecone, and Qdrant by control plane, operations, filtering, deployment model, ANN indexes, and RAG retrieval fit.05 / cosine similarity in vector search
Cosine Similarity Explained for Vector Search
Learn how cosine similarity compares embedding vectors, when it works, where it fails, and why high similarity does not guarantee correctness.06 / RAG chunking strategies
RAG Chunking Strategies
Learn practical RAG chunking strategies for documents, code, support content, metadata, overlap, evaluation, and retrieval failure prevention.07 / RAG retrieval problems
RAG Retrieval Failure Modes
Learn common RAG retrieval problems, including wrong chunks, missing context, stale metadata, weak filters, reranking failures, and evaluation gaps.08 / vector database interview questions
Vector Database Interview Questions
Practice vector database interview questions covering embeddings, ANN indexes, HNSW, IVF, PQ, hybrid search, RAG retrieval, filtering, and memory.09 / how much RAM for 1 million embeddings
How Much RAM for 1 Million Embeddings?
Calculate RAM for 1 million embeddings across common dimensions and precisions, including float32, float16, int8, index overhead, metadata, and production headroom.10 / 1536 dimension embedding memory calculator
1536 Dimension Embedding Memory Calculator
Estimate memory for 1536-dimensional embeddings using vector count, precision, HNSW or IVF overhead, metadata, replicas, and production headroom.11 / HNSW memory usage explained