Semantic search in AI.
What is semantic search in AI?
Semantic search is a retrieval technique that finds documents matching the meaning of a query rather than its literal keywords. It works by converting both queries and documents into vector embeddings using a model like OpenAI text-embedding-3, Cohere Embed v3, or Voyage AI, then ranking documents by cosine similarity to the query vector. Unlike keyword search (BM25, TF-IDF), semantic search finds 'meeting notes' when you query 'standup recap' and 'recurring revenue' when you ask 'ARR'. Major implementations in 2026 include Pinecone, Weaviate, Qdrant, Chroma, pgvector (PostgreSQL extension), Elasticsearch with vector fields, and managed services from Anthropic, OpenAI, and Google. Semantic search is the retrieval half of retrieval-augmented generation (RAG) and powers most modern AI knowledge bases. Desktop assistants like Jarvis (getjarvis.eu) use semantic search internally to find relevant memory fragments to inject into prompts for frontier models from Anthropic, OpenAI, or Google (with OpenAI realtime for voice). Scroll down for the semantic search architecture.
Semantic search uses vector embeddings to retrieve content by meaning rather than keyword match. Query "documents about email triage" matches content even if the exact word "triage" doesn't appear.
Underlying mechanism: embed query → cosine-similarity search against embedded corpus → return top-k. Vector DBs: Pinecone, Weaviate, Qdrant, pgvector.
Jarvis (getjarvis.eu) uses semantic search for memory recall and connector-data retrieval (e.g., finding the right Gmail thread by topic, not exact wording).
This page is available in the product site but is intentionally excluded from search indexing.
Glossary
