Vector store
Where embeddings live, so search finds meaning — Postgres often already does it.
A vector store is what makes semantic search fast: it returns the stored texts closest in meaning to your question in milliseconds, even across millions of entries. Brute-force comparison stops working around ten thousand entries, which is when index structures like HNSW start earning their keep. The catch is that vendors sell you a separate database far more often than you actually need one.
You need this when
- You are searching thousands of documents by meaning — an SOP assistant, a ticket archive, a big product catalogue.
- You have passed roughly ten thousand vectors, where checking them one by one stops being fast enough.
- You need tens of millions of vectors, or specific features like multi-tenancy — that is where dedicated stores earn their fee.
You can skip it when
- Under ~10 million vectors and you already run Postgres — turn on the free pgvector extension instead: no new infrastructure, one backup story.
- A small file-based assistant over a few hundred documents — use the built-in file search in ChatGPT or Gemini instead of standing up a database.
- Your users search error codes and SKUs and get nonsense — add classic keyword search alongside (hybrid search), not a bigger vector store.
The long version — open when you want the depth
What it is — in one coffee-break
A vector store is the database where embeddings live (lesson 2 covered what those are: meaning, expressed as numbers). Its one core trick is nearest-neighbour search: given the embedding of a question, return the stored texts whose embeddings sit closest — in milliseconds, even across millions of entries. Checking every stored vector one by one stops working at around ten thousand entries, so vector databases use clever index structures (HNSW is the default everyone uses) that trade a sliver of accuracy for thousand-fold speed.
When you actually need it (and when you don't)
You need one exactly when you need embeddings at scale: a knowledge assistant over thousands of documents, semantic search across your ticket history, deduplication across a big catalogue. You don't need a separate one nearly as often as vendors suggest. The 2026 practitioner consensus is refreshingly boring: if you already run Postgres — and most businesses do, often without knowing it — the free pgvector extension handles up to ~10 million vectors alongside your normal data, with one backup story and zero new infrastructure. Our SOP answer machine and receipt archive blueprints follow the same keep-it-simple logic.
Dedicated stores (Qdrant, Weaviate — both open-source and self-hostable — or managed Pinecone) earn their place at tens of millions of vectors, or when you need specific features like multi-tenancy. And whatever the store, hybrid search is the production standard: pure meaning-based search misses exact terms like error codes and product SKUs, so real systems combine it with classic keyword search.
How to recognize good vs bad implementations
In the audits, the Vector store element (Vx) again exposes platform strategy. Gemini scores 9 — Google sells the whole retrieval stack, storage included. ChatGPT scores 7 with built-in file search that just works at small scale. Claude scores 2 — bring your own, same deliberate gap as embeddings. Workflow tools cluster mid-table (n8n 6.5, Make 5.5) with connectors to the usual stores. The trap to recognize: tools that store your embeddings but won't let you export them. That's your data, indexed — if it can't leave, the meaning-layer of your knowledge base is locked in.
What this costs
Storage is the meter, and dimensions drive it: a million vectors at 1,536 dimensions is roughly 6 GB before index overhead. Self-hosted pgvector: effectively €0 plus the server you already have. Managed serverless starts around $0.33/GB-month — trivial at small business scale, meaningful at millions of documents. The cost levers are dimension truncation and quantization: shrinking vectors 6–12× costs only a few percent of quality. For almost everyone reading this: your vector bill will be under €10/month, and pgvector makes it zero.
Where to see it scored
Vx with reasons: Gemini (9), ChatGPT (7), Windsurf (6.5), Claude (2 — bring your own). Next lesson: RAG — the pattern that puts embeddings, the vector store and the model together.
Flashcards
Check yourself
1. You already run Postgres and want semantic search over 200,000 documents. First move?
2. Which tool scores highest on Vector store (Vx) in the audits?
3. Support staff search by error code and semantic search returns nonsense. Fix?
4. Why do vector databases use index structures like HNSW?
5. A vendor stores your embeddings but offers no export. What is the real risk?
Cheat sheet
- Core trick: nearest-neighbour search over embeddings, milliseconds at millions of rows.
- HNSW is the default index — trades a sliver of accuracy for thousand-fold speed.
- pgvector handles ~10M vectors inside the Postgres you already run, for effectively EUR 0.
- Qdrant, Weaviate and Pinecone start earning their fee at tens of millions of vectors.
- Hybrid = vector + keyword. Pure semantic search misses exact error codes and SKUs.
- Trap: a store that won't export your embeddings locks in your whole meaning layer.
Who actually does this well
| Best on this element | Score | Why it scored that |
|---|---|---|
| Gemini / Google | 9 | Vector Search 2.0 is GA with hybrid search, built-in re-ranking, auto-embeddings and a storage-optimized tier — the strongest first-party vector store of any frontier lab. |
| Glean | 7.5 | Vector search is core, load-bearing architecture at production scale, but it isn't sold as a standalone vector database independent of Glean's own retrieval. |
| Relevance AI | 7.5 | Knowledge tables are a hosted vector store with hybrid retrieval and reranking, synced automatically from files, websites, Google Drive, SharePoint and Notion, with custom metadata |
| ChatGPT / OpenAI Platform | 7 | Hosted vector stores work out of the box but are opaque: limited tuning, index control and export. |
| Grok (xAI) | 7 | Collections is a real hosted vector store — embedding index, configurable chunking, metadata filters, ChromaDB/Elasticsearch pipelines — at $2.50/1k searches; younger and less prov |
| OpenAI AgentKit | 7 | Hosted vector stores with full CRUD, batches to 500 files, attribute filtering, configurable chunking (800-token default, 400 overlap), hybrid search with tunable embedding and tex |
| Salesforce Agentforce | 7 | Data 360 ships vector, hybrid and enriched search indexes with retrievers — capable and integrated, but a Data Cloud dependency with its own consumption pricing. |
| Cursor | 6.5 | The hosted codebase index is a dependable managed semantic store for code, but it is single-purpose — you cannot build on it as a general vector store. |
And the other end of the same column:
| Weakest | Score | Why it scored that |
|---|---|---|
| Claude / Anthropic | 2 | No hosted vector store product of any kind — competitors at this tier ship one; you must pair Claude with external vector infrastructure. |
| Cornerstone OnDemand | 2 | No vector store, no semantic index a customer can write to or query. Content and skills live in Cornerstone's own structures and are reachable only through the product surfaces and |
| Guild | 2 | No vector or semantic store is shipped, hosted or exposed. Program discovery runs on a curated catalogue plus human coaches. |
| Tines | 2 | No hosted vector store. Records is a structured table, not a semantic index. Semantic storage is a system you point Tines at, not one it gives you. |
Scored on 44 of 144 audited tools. Every score links to the full audit and its reasoning.