Store Options¶

ragwise supports three store backends that cover the full spectrum from local development to production.

Comparison¶

Store	Setup	Persistence	Scale	Use case
`memory`	zero	none (volatile)	≤500K docs	tests, demos, CI
`lance://`	zero	disk	≤50M docs	dev, prototyping
`postgresql://`	Postgres + pgvector	disk	50M+ docs	production

InMemoryStore¶

RAG(store="memory")

Zero setup — no files, no server
Data lost when the process exits
Uses numpy for dense search and rank-bm25 for sparse search
Recommended for: unit tests, CI, quick demos

LanceDBStore¶

pip install ragwise[lance]

RAG(store="lance://./ragwise-index")

Embedded — no server process
Persistent across restarts
Scales to tens of millions of documents
Full-text search (FTS) index built automatically
Recommended for: local development, prototyping, single-machine deployments

The path after lance:// is relative to where your script runs.

PgVectorStore¶

pip install ragwise[postgres]

RAG(store="postgresql://user:pass@localhost/mydb")

Uses PostgreSQL with the pgvector extension
Native hybrid search: pgvector for dense, tsvector/ts_rank for sparse — combined in one SQL query
pgvector benchmark: 471 QPS at 50M vectors, 99% recall (source)
No new infrastructure if your team already runs PostgreSQL
Table ragwise_docs and indexes are created automatically on first use
Recommended for: production, multi-instance deployments, teams on PostgreSQL

Quick setup with Docker (development only)¶

docker run -d \
  -e POSTGRES_PASSWORD=password \
  -p 5432:5432 \
  pgvector/pgvector:pg17

ragwise_store = "postgresql://postgres:password@localhost/postgres"

Upgrade path¶

# Phase 1 — local dev
async with RAG(store="memory") as rag:
    ...

# Phase 2 — persistent dev, survives restarts
async with RAG(store="lance://./ragwise-index") as rag:
    ...

# Phase 3 — production
async with RAG(store="postgresql://user:pass@db.example.com/mydb") as rag:
    ...

The API stays identical across all three. Only the store= string changes.