Scaled Deployment

ModeScaled distributes storage across dedicated backends: Qdrant for vectors, Redis for key-value and event log, and Postgres for the graph store. This enables higher throughput and independent scaling of each component.

Architecture

Configuration

Go SDK

db, err := client.Open(client.Options{
    Mode:             client.ModeScaled,
    DSN:              "postgres://user:pass@localhost:5432/contextdb?sslmode=disable",
    QdrantAddr:       "localhost:6334",
    RedisAddr:        "localhost:6379",
    VectorDimensions: 1536,
})

Option	Type	Description
Mode	Mode	Must be ModeScaled
DSN	string	Postgres connection string for graph store
QdrantAddr	string	Qdrant gRPC address
RedisAddr	string	Redis address
VectorDimensions	int	Embedding dimensionality (default: 1536)

Docker Compose

The docker-compose.yml includes a scaled profile with Qdrant and Redis:

docker compose --profile scaled up --build

This starts:

Service	Port	Purpose
contextdb	7700, 7701	gRPC + REST server
postgres	5432	Graph store with pgvector
qdrant	6334	Vector index
redis	6379	KV store + event log

All services have persistent volumes and health checks.

Backend details

Qdrant

The Qdrant backend implements store.VectorIndex:

• Collections are named contextdb_{namespace} with cosine distance
• Label filtering is pushed down to Qdrant as metadata filters
• Vectors are stored with node metadata (ID, namespace, labels)
• Requires the integration build tag for compilation

Redis

The Redis backend implements store.KVStore and store.EventLog:

• Keys are prefixed with contextdb:kv: by default
• TTL is supported via SET ... EX for session data and caches
• Event log entries are stored in Redis sorted sets

Postgres

Postgres continues to serve as the graph store in scaled mode, providing:

• Node and edge CRUD with versioning
• Recursive CTE graph traversal for Walk operations
• Source credibility tracking

When to use scaled mode

Scenario	Recommended mode
Development and testing	Embedded (in-memory)
Single-node production	Standard (Postgres only)
Multi-service architecture	Remote (gRPC client)
High vector throughput, large datasets	Scaled
Independent scaling of vector/graph/cache	Scaled