# qdrant-scale

Agent skills for scaling Qdrant vector search deployments

pip install qdrant-scale

```python
from qdrant_scale import QdrantScaler

# attach to existing cluster
scaler = QdrantScaler(
    host="localhost:6333",
    api_key="your-key"
)

# check if scaling needed
metrics = scaler.get_metrics()
if metrics.cpu_usage > 0.8 or metrics.memory_usage > 0.9:
    scaler.scale_up(target_nodes=5)

# auto-scale based on query latency
scaler.watch(
    max_latency_ms=100,
    check_interval=30
)

notes

designed for programmatic control of Qdrant clusters. works with cloud deployments and self-hosted k8s setups.

skills included:

• horizontal scaling (add/remove nodes)
• vertical scaling (change node specs)
• collection sharding operations
• replication factor adjustment
• health monitoring and auto-recovery

requires cluster mode, won't work on single-node standalone instances.

from qdrant_scale import QdrantScaler, ScalingPolicy

# define custom scaling logic
policy = ScalingPolicy(
    scale_up_threshold=0.75,
    scale_down_threshold=0.25,
    cooldown_minutes=10,
    min_nodes=2,
    max_nodes=20
)

scaler = QdrantScaler("localhost:6333")
scaler.apply_policy(policy)

# reshard collection across new nodes
scaler.reshard_collection(
    collection_name="embeddings",
    shard_count=8,
    replication_factor=2
)

# drain node before removal
scaler.drain_node("node-3")
scaler.remove_node("node-3")

# backup before scaling operations
scaler.create_snapshot("pre-scale-snapshot")
scaler.scale_up(target_nodes=10)

MIT

<!-- nit -->