RAG with Weaviate

Overview

Weaviate is an open-source vector database that powers semantic search, hybrid (keyword + vector) retrieval, and generative feedback loops. Meridian provisions a dedicated Weaviate cluster inside your VPC, pre-configured with TLS, backup snapshots, and a RESTful GraphQL API.

Quickstart

Once your cluster is provisioned, create a collection and start indexing.

import weaviate

client = weaviate.Client(
    url="https://your-cluster.meridian.dev",
    auth_client_secret=weaviate.AuthApiKey("sk-..."),
)

client.schema.create_class({
    "class": "Document",
    "vectorizer": "none",
    "properties": [
        {"name": "text", "dataType": ["text"]},
        {"name": "source", "dataType": ["string"]},
    ],
})

Hybrid Search

Combine BM25 keyword scoring with cosine-similarity vector search. Weaviate fuses results using a configurable alpha parameter — set alpha=0.5 for equal weighting.

{
  Get {
    Document(
      hybrid: {
        query: "transformer attention mechanism"
        alpha: 0.5
      }
      limit: 5
    ) {
      text
      source
      _additional { score }
    }
  }
}

End-to-End RAG Pipeline

Limits & Pricing