Skip to content

nornic-search-grpc.md

Latest commit

 

History

History
91 lines (68 loc) · 2.58 KB

nornic-search-grpc.md

File metadata and controls

91 lines (68 loc) · 2.58 KB

NornicSearch gRPC With Existing Qdrant Drivers

Use existing Qdrant gRPC drivers unchanged, and add NornicDB's NornicSearch proto as an extra client when you want SearchText.

You do not need to fork Qdrant drivers.

When You Need This

  • You already use Qdrant gRPC APIs (Points.Search, Points.Query, etc.).
  • You also want Nornic-native NornicSearch/SearchText.
  • You want both over the same endpoint (:6334 by default).

Prerequisites

  • NORNICDB_QDRANT_GRPC_ENABLED=true
  • Port 6334 exposed/reachable
  • Existing Qdrant gRPC client in your app (unchanged)
  • Add generated client code from pkg/nornicgrpc/proto/nornicdb_search.proto

Architecture (Side-by-Side Clients)

  • Qdrant client: talks to Qdrant-compatible services
  • Nornic client: talks to NornicSearch service
  • Same gRPC channel: both clients can share one connection to 127.0.0.1:6334

Go Example (Recommended)

  1. Keep your current Qdrant client/protos.

  2. Generate Go stubs for Nornic proto:

protoc \
  --go_out=. \
  --go-grpc_out=. \
  pkg/nornicgrpc/proto/nornicdb_search.proto
  1. Create both clients on the same connection:
conn, err := grpc.Dial("127.0.0.1:6334", grpc.WithTransportCredentials(insecure.NewCredentials()))
if err != nil {
    log.Fatal(err)
}
defer conn.Close()

qdrantClient := qdrant.NewPointsClient(conn)             // Existing Qdrant driver usage
nornicClient := nornicpb.NewNornicSearchClient(conn)     // Additive Nornic client

// Existing Qdrant-compatible vector search
_, err = qdrantClient.Search(ctx, &qdrant.SearchPoints{
    CollectionName: "my_vectors",
    Vector:         make([]float32, 128),
    Limit:          10,
})
if err != nil {
    log.Fatal(err)
}

// Nornic-native text search
resp, err := nornicClient.SearchText(ctx, &nornicpb.SearchTextRequest{
    Query: "machine learning",
    Limit: 10,
})
if err != nil {
    log.Fatal(err)
}
_ = resp

Existing Qdrant Text Query (No Nornic Proto Needed)

If you only need text query with existing Qdrant drivers, use Points.Query with Document.text:

from qdrant_client import QdrantClient, models

client = QdrantClient(host="127.0.0.1", grpc_port=6334, prefer_grpc=True)
client.query_points(
    collection_name="my_vectors",
    query=models.Document(text="machine learning"),
    limit=10,
)

Which Path Should I Use?

  • Use Qdrant Points.Query + Document.text when you want pure Qdrant-driver semantics.
  • Use NornicSearch/SearchText when you want Nornic-native typed search RPC behavior.
  • Use both together if your app needs both compatibility and native features.