1
2
3

docker run -p 6333:6333 -p 6334:6334 \
  -v $(pwd)/qdrant_storage:/qdrant/storage \
  qdrant/qdrant

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

version: '3.8'
services:
  qdrant:
    image: qdrant/qdrant:latest
    ports:
      - "6333:6333"
      - "6334:6334"
    volumes:
      - ./qdrant_storage:/qdrant/storage
    environment:
      - QDRANT__SERVICE__GRPC_PORT=6334

1

pip install qdrant-client

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35

from qdrant_client import QdrantClient
from qdrant_client.models import Distance, VectorParams, PointStruct

# 連接
client = QdrantClient("localhost", port=6333)

# 建立 Collection
client.create_collection(
    collection_name="security_docs",
    vectors_config=VectorParams(size=384, distance=Distance.COSINE)
)

# 新增向量
client.upsert(
    collection_name="security_docs",
    points=[
        PointStruct(
            id=1,
            vector=[0.1, 0.2, ...],  # 384 維向量
            payload={"title": "SQL Injection", "severity": "high"}
        ),
        PointStruct(
            id=2,
            vector=[0.3, 0.4, ...],
            payload={"title": "XSS Attack", "severity": "medium"}
        )
    ]
)

# 搜尋
results = client.search(
    collection_name="security_docs",
    query_vector=[0.15, 0.25, ...],
    limit=5
)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19

from qdrant_client.models import Distance, VectorParams

# 單一向量欄位
client.create_collection(
    collection_name="docs",
    vectors_config=VectorParams(
        size=768,
        distance=Distance.COSINE
    )
)

# 多向量欄位
client.create_collection(
    collection_name="multimodal",
    vectors_config={
        "text": VectorParams(size=768, distance=Distance.COSINE),
        "image": VectorParams(size=512, distance=Distance.EUCLIDEAN)
    }
)

1
2
3
4
5
6
7
8

# 取得資訊
info = client.get_collection("docs")

# 列出所有
collections = client.get_collections()

# 刪除
client.delete_collection("docs")

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

from qdrant_client.models import PointStruct

# 新增點
client.upsert(
    collection_name="docs",
    points=[
        PointStruct(
            id=1,
            vector=[0.1, 0.2, 0.3, ...],
            payload={
                "title": "OWASP Top 10",
                "category": "security",
                "year": 2025
            }
        )
    ]
)

# 使用 UUID
import uuid
client.upsert(
    collection_name="docs",
    points=[
        PointStruct(
            id=str(uuid.uuid4()),
            vector=[0.1, 0.2, ...],
            payload={"title": "Document"}
        )
    ]
)

1
2
3
4
5
6
7
8

# 批次上傳
client.upload_collection(
    collection_name="docs",
    vectors=vectors_list,
    payload=payloads_list,
    ids=ids_list,
    batch_size=100
)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20

# 依 ID 刪除
client.delete(
    collection_name="docs",
    points_selector=[1, 2, 3]
)

# 依條件刪除
from qdrant_client.models import Filter, FieldCondition, MatchValue

client.delete(
    collection_name="docs",
    points_selector=Filter(
        must=[
            FieldCondition(
                key="category",
                match=MatchValue(value="deprecated")
            )
        ]
    )
)

1
2
3
4
5
6
7
8
9

results = client.search(
    collection_name="docs",
    query_vector=[0.1, 0.2, ...],
    limit=10
)

for result in results:
    print(f"ID: {result.id}, Score: {result.score}")
    print(f"Payload: {result.payload}")

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19

from qdrant_client.models import Filter, FieldCondition, MatchValue, Range

results = client.search(
    collection_name="docs",
    query_vector=query_vector,
    query_filter=Filter(
        must=[
            FieldCondition(
                key="category",
                match=MatchValue(value="security")
            ),
            FieldCondition(
                key="severity",
                match=MatchValue(value="high")
            )
        ]
    ),
    limit=10
)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13

results = client.search(
    collection_name="docs",
    query_vector=query_vector,
    query_filter=Filter(
        must=[
            FieldCondition(
                key="year",
                range=Range(gte=2024, lte=2026)
            )
        ]
    ),
    limit=10
)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

from qdrant_client.models import FieldCondition, MatchValue, Range

# 等於
FieldCondition(key="status", match=MatchValue(value="active"))

# 範圍
FieldCondition(key="score", range=Range(gt=0.5, lt=1.0))

# 在列表中
FieldCondition(key="tag", match=MatchAny(any=["web", "api"]))

1
2
3
4
5

Filter(
    must=[...],      # AND
    should=[...],    # OR
    must_not=[...]   # NOT
)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13

# 加速過濾查詢
client.create_payload_index(
    collection_name="docs",
    field_name="category",
    field_schema="keyword"
)

# 數值索引
client.create_payload_index(
    collection_name="docs",
    field_name="year",
    field_schema="integer"
)

1
2
3
4
5
6
7
8
9

from qdrant_client.models import HnswConfigDiff

client.update_collection(
    collection_name="docs",
    hnsw_config=HnswConfigDiff(
        m=16,
        ef_construct=100
    )
)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15

from langchain_qdrant import Qdrant
from langchain_openai import OpenAIEmbeddings

embeddings = OpenAIEmbeddings()

# 建立向量儲存
vectorstore = Qdrant.from_documents(
    documents=docs,
    embedding=embeddings,
    url="http://localhost:6333",
    collection_name="langchain_docs"
)

# 搜尋
results = vectorstore.similarity_search("security vulnerability", k=3)

1
2
3
4
5
6

client.create_collection(
    collection_name="large_docs",
    vectors_config=VectorParams(size=768, distance=Distance.COSINE),
    shard_number=6,          # 分片數
    replication_factor=2     # 副本數
)

1
2
3
4

client = QdrantClient(
    url="https://your-cluster.qdrant.io",
    api_key="your-api-key"
)