Open-source vector database with an embedded mode — the searchable memory a RAG system stores its chunks in.

Pick this when

You want vector search inside the process (single-tenant, on-prem, laptop-sized) instead of a hosted vector-database service.

What makes it different

Embedded mode runs in-process — Session 22's build used one collection per conversation with no service to operate

Part of a ~107MB install alongside Haystack — vector search without infrastructure

Scales up to a real clustered service later without changing the mental model

What it does

Stores embeddings and runs similarity search over them

Isolates corpora cleanly (collection-per-conversation in the Talent Lens build)

Serves both the demo (OpenAI embeddings) and production (local) paths unchanged

What it does not do

Common wrong expectations

Doesn't create embeddings — an embedding model does that upstream

Doesn't replace retrieval logic — chunking, overlap, and evidence gates live in the orchestrator

Embedded mode isn't multi-tenant infrastructure — that's what the server product is for

Where it sits in a stack

The vector store slot in a lightweight RAG stack; the 'database is just a library' answer for on-prem builds.

Shelf

Dev libraries & frameworks · Backend, hosting & code

Our status · watch

Would be the storage layer if the KB gains semantic search; nothing built on it yet.

For your projects

Embedded mode fits Paul's single-tenant KB exactly — no new service to run.

Freshness

current

Related tools

complementsHaystackcompetes withSupabase

haystack: Chosen together in Session 22 after the measured comparison

supabase: pgvector on Supabase is the hosted-Postgres answer to the same storage question

Every moment the course touched it (1)

SessionCoverageMomentWhat happened
Session 22: Automate Admin Tasks — Advanced RAG (Enterprise 'Internal Perplexity' Built Live)demonstratedEmbedded mode — one collection per conversation, no separate service

← All tools by category