[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6DFPBNRREW313P6C73VXTSS":3,"skill-same:01M6DFPBNRREW313P6C73VXTSS":81},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":39,"linkOnly":40,"markdownContent":41,"markdownBody":42,"frontMatter":43,"refName":59,"commitSha":60,"skillDirectory":61,"forksCount":62,"repositoryHtmlUrl":63,"repositoryHomepageUrl":6,"ownerHtmlUrl":64,"ownerBlogUrl":6,"ownerDescription":65,"updatedAt":66,"install":67,"resources":80},"01M6DFPBNRREW313P6C73VXTSS","qdrant-vector-search",null,"High-performance vector similarity search engine for RAG and semantic search. Use when building production RAG systems requiring fast nearest neighbor search, hybrid search with filtering, or scalable vector storage with Rust-powered performance.","zapabob\u002FSkills","zapabob","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F130830376?v=4","01M6DBMYZAGRTFQDSQYW3W764P","峯岸 亮","01M6DBMZ3GKDSJ01NZ9XTZD7JD",2,"MIT","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Farchive\u002Fskill-md-reviewed\u002Fsha256\u002F50\u002F504b9d36683ec89ee492f9770c8c49c8d763f73ffbf2c677c74da16caf1c90e9\u002FSKILL.md",[],[19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38],"qdrant vector search","Qdrant","Vector Search","vector","search","systems","macos","use","filtering","engine","embeddings","semantic search","rag","distributed","windows","storage","research","rust","performance","linux","2026-11-10 10:12:30",false,"---\r\nname: qdrant-vector-search\r\ndescription: High-performance vector similarity search engine for RAG and semantic search. Use when building production RAG systems requiring fast nearest neighbor search, hybrid search with filtering, or scalable vector storage with Rust-powered performance.\r\nversion: 1.0.0\r\nauthor: Orchestra Research\r\nlicense: MIT\r\ndependencies: [qdrant-client>=1.12.0]\r\nplatforms: [linux, macos, windows]\r\nmetadata:\r\n  hermes:\r\n    tags: [RAG, Vector Search, Qdrant, Semantic Search, Embeddings, Similarity Search, HNSW, Production, Distributed]\r\n\r\n---\r\n\r\n# Qdrant - Vector Similarity Search Engine\r\n\r\nHigh-performance vector database written in Rust for production RAG and semantic search.\r\n\r\n## When to use Qdrant\r\n\r\n**Use Qdrant when:**\r\n- Building production RAG systems requiring low latency\r\n- Need hybrid search (vectors + metadata filtering)\r\n- Require horizontal scaling with sharding\u002Freplication\r\n- Want on-premise deployment with full data control\r\n- Need multi-vector storage per record (dense + sparse)\r\n- Building real-time recommendation systems\r\n\r\n**Key features:**\r\n- **Rust-powered**: Memory-safe, high performance\r\n- **Rich filtering**: Filter by any payload field during search\r\n- **Multiple vectors**: Dense, sparse, multi-dense per point\r\n- **Quantization**: Scalar, product, binary for memory efficiency\r\n- **Distributed**: Raft consensus, sharding, replication\r\n- **REST + gRPC**: Both APIs with full feature parity\r\n\r\n**Use alternatives instead:**\r\n- **Chroma**: Simpler setup, embedded use cases\r\n- **FAISS**: Maximum raw speed, research\u002Fbatch processing\r\n- **Pinecone**: Fully managed, zero ops preferred\r\n- **Weaviate**: GraphQL preference, built-in vectorizers\r\n\r\n## Quick start\r\n\r\n### Installation\r\n\r\n```bash\r\n# Python client\r\npip install qdrant-client\r\n\r\n# Docker (recommended for development)\r\ndocker run -p 6333:6333 -p 6334:6334 qdrant\u002Fqdrant\r\n\r\n# Docker with persistent storage\r\ndocker run -p 6333:6333 -p 6334:6334 \\\r\n    -v $(pwd)\u002Fqdrant_storage:\u002Fqdrant\u002Fstorage \\\r\n    qdrant\u002Fqdrant\r\n```\r\n\r\n### Basic usage\r\n\r\n```python\r\nfrom qdrant_client import QdrantClient\r\nfrom qdrant_client.models import Distance, VectorParams, PointStruct\r\n\r\n# Connect to Qdrant\r\nclient = QdrantClient(host=\"localhost\", port=6333)\r\n\r\n# Create collection\r\nclient.create_collection(\r\n    collection_name=\"documents\",\r\n    vectors_config=VectorParams(size=384, distance=Distance.COSINE)\r\n)\r\n\r\n# Insert vectors with payload\r\nclient.upsert(\r\n    collection_name=\"documents\",\r\n    points=[\r\n        PointStruct(\r\n            id=1,\r\n            vector=[0.1, 0.2, ...],  # 384-dim vector\r\n            payload={\"title\": \"Doc 1\", \"category\": \"tech\"}\r\n        ),\r\n        PointStruct(\r\n            id=2,\r\n            vector=[0.3, 0.4, ...],\r\n            payload={\"title\": \"Doc 2\", \"category\": \"science\"}\r\n        )\r\n    ]\r\n)\r\n\r\n# Search with filtering\r\nresults = client.search(\r\n    collection_name=\"documents\",\r\n    query_vector=[0.15, 0.25, ...],\r\n    query_filter={\r\n        \"must\": [{\"key\": \"category\", \"match\": {\"value\": \"tech\"}}]\r\n    },\r\n    limit=10\r\n)\r\n\r\nfor point in results:\r\n    print(f\"ID: {point.id}, Score: {point.score}, Payload: {point.payload}\")\r\n```\r\n\r\n## Core concepts\r\n\r\n### Points - Basic data unit\r\n\r\n```python\r\nfrom qdrant_client.models import PointStruct\r\n\r\n# Point = ID + Vector(s) + Payload\r\npoint = PointStruct(\r\n    id=123,                              # Integer or UUID string\r\n    vector=[0.1, 0.2, 0.3, ...],        # Dense vector\r\n    payload={                            # Arbitrary JSON metadata\r\n        \"title\": \"Document title\",\r\n        \"category\": \"tech\",\r\n        \"timestamp\": 1699900000,\r\n        \"tags\": [\"python\", \"ml\"]\r\n    }\r\n)\r\n\r\n# Batch upsert (recommended)\r\nclient.upsert(\r\n    collection_name=\"documents\",\r\n    points=[point1, point2, point3],\r\n    wait=True  # Wait for indexing\r\n)\r\n```\r\n\r\n### Collections - Vector containers\r\n\r\n```python\r\nfrom qdrant_client.models import VectorParams, Distance, HnswConfigDiff\r\n\r\n# Create with HNSW configuration\r\nclient.create_collection(\r\n    collection_name=\"documents\",\r\n    vectors_config=VectorParams(\r\n        size=384,                        # Vector dimensions\r\n        distance=Distance.COSINE         # COSINE, EUCLID, DOT, MANHATTAN\r\n    ),\r\n    hnsw_config=HnswConfigDiff(\r\n        m=16,                            # Connections per node (default 16)\r\n        ef_construct=100,                # Build-time accuracy (default 100)\r\n        full_scan_threshold=10000        # Switch to brute force below this\r\n    ),\r\n    on_disk_payload=True                 # Store payload on disk\r\n)\r\n\r\n# Collection info\r\ninfo = client.get_collection(\"documents\")\r\nprint(f\"Points: {info.points_count}, Vectors: {info.vectors_count}\")\r\n```\r\n\r\n### Distance metrics\r\n\r\n| Metric | Use Case | Range |\r\n|--------|----------|-------|\r\n| `COSINE` | Text embeddings, normalized vectors | 0 to 2 |\r\n| `EUCLID` | Spatial data, image features | 0 to ∞ |\r\n| `DOT` | Recommendations, unnormalized | -∞ to ∞ |\r\n| `MANHATTAN` | Sparse features, discrete data | 0 to ∞ |\r\n\r\n## Search operations\r\n\r\n### Basic search\r\n\r\n```python\r\n# Simple nearest neighbor search\r\nresults = client.search(\r\n    collection_name=\"documents\",\r\n    query_vector=[0.1, 0.2, ...],\r\n    limit=10,\r\n    with_payload=True,\r\n    with_vectors=False  # Don't return vectors (faster)\r\n)\r\n```\r\n\r\n### Filtered search\r\n\r\n```python\r\nfrom qdrant_client.models import Filter, FieldCondition, MatchValue, Range\r\n\r\n# Complex filtering\r\nresults = client.search(\r\n    collection_name=\"documents\",\r\n    query_vector=query_embedding,\r\n    query_filter=Filter(\r\n        must=[\r\n            FieldCondition(key=\"category\", match=MatchValue(value=\"tech\")),\r\n            FieldCondition(key=\"timestamp\", range=Range(gte=1699000000))\r\n        ],\r\n        must_not=[\r\n            FieldCondition(key=\"status\", match=MatchValue(value=\"archived\"))\r\n        ]\r\n    ),\r\n    limit=10\r\n)\r\n\r\n# Shorthand filter syntax\r\nresults = client.search(\r\n    collection_name=\"documents\",\r\n    query_vector=query_embedding,\r\n    query_filter={\r\n        \"must\": [\r\n            {\"key\": \"category\", \"match\": {\"value\": \"tech\"}},\r\n            {\"key\": \"price\", \"range\": {\"gte\": 10, \"lte\": 100}}\r\n        ]\r\n    },\r\n    limit=10\r\n)\r\n```\r\n\r\n### Batch search\r\n\r\n```python\r\nfrom qdrant_client.models import SearchRequest\r\n\r\n# Multiple queries in one request\r\nresults = client.search_batch(\r\n    collection_name=\"documents\",\r\n    requests=[\r\n        SearchRequest(vector=[0.1, ...], limit=5),\r\n        SearchRequest(vector=[0.2, ...], limit=5, filter={\"must\": [...]}),\r\n        SearchRequest(vector=[0.3, ...], limit=10)\r\n    ]\r\n)\r\n```\r\n\r\n## RAG integration\r\n\r\n### With sentence-transformers\r\n\r\n```python\r\nfrom sentence_transformers import SentenceTransformer\r\nfrom qdrant_client import QdrantClient\r\nfrom qdrant_client.models import VectorParams, Distance, PointStruct\r\n\r\n# Initialize\r\nencoder = SentenceTransformer(\"all-MiniLM-L6-v2\")\r\nclient = QdrantClient(host=\"localhost\", port=6333)\r\n\r\n# Create collection\r\nclient.create_collection(\r\n    collection_name=\"knowledge_base\",\r\n    vectors_config=VectorParams(size=384, distance=Distance.COSINE)\r\n)\r\n\r\n# Index documents\r\ndocuments = [\r\n    {\"id\": 1, \"text\": \"Python is a programming language\", \"source\": \"wiki\"},\r\n    {\"id\": 2, \"text\": \"Machine learning uses algorithms\", \"source\": \"textbook\"},\r\n]\r\n\r\npoints = [\r\n    PointStruct(\r\n        id=doc[\"id\"],\r\n        vector=encoder.encode(doc[\"text\"]).tolist(),\r\n        payload={\"text\": doc[\"text\"], \"source\": doc[\"source\"]}\r\n    )\r\n    for doc in documents\r\n]\r\nclient.upsert(collection_name=\"knowledge_base\", points=points)\r\n\r\n# RAG retrieval\r\ndef retrieve(query: str, top_k: int = 5) -> list[dict]:\r\n    query_vector = encoder.encode(query).tolist()\r\n    results = client.search(\r\n        collection_name=\"knowledge_base\",\r\n        query_vector=query_vector,\r\n        limit=top_k\r\n    )\r\n    return [{\"text\": r.payload[\"text\"], \"score\": r.score} for r in results]\r\n\r\n# Use in RAG pipeline\r\ncontext = retrieve(\"What is Python?\")\r\nprompt = f\"Context: {context}\\n\\nQuestion: What is Python?\"\r\n```\r\n\r\n### With LangChain\r\n\r\n```python\r\nfrom langchain_community.vectorstores import Qdrant\r\nfrom langchain_community.embeddings import HuggingFaceEmbeddings\r\n\r\nembeddings = HuggingFaceEmbeddings(model_name=\"all-MiniLM-L6-v2\")\r\nvectorstore = Qdrant.from_documents(documents, embeddings, url=\"http:\u002F\u002Flocalhost:6333\", collection_name=\"docs\")\r\nretriever = vectorstore.as_retriever(search_kwargs={\"k\": 5})\r\n```\r\n\r\n### With LlamaIndex\r\n\r\n```python\r\nfrom llama_index.vector_stores.qdrant import QdrantVectorStore\r\nfrom llama_index.core import VectorStoreIndex, StorageContext\r\n\r\nvector_store = QdrantVectorStore(client=client, collection_name=\"llama_docs\")\r\nstorage_context = StorageContext.from_defaults(vector_store=vector_store)\r\nindex = VectorStoreIndex.from_documents(documents, storage_context=storage_context)\r\nquery_engine = index.as_query_engine()\r\n```\r\n\r\n## Multi-vector support\r\n\r\n### Named vectors (different embedding models)\r\n\r\n```python\r\nfrom qdrant_client.models import VectorParams, Distance\r\n\r\n# Collection with multiple vector types\r\nclient.create_collection(\r\n    collection_name=\"hybrid_search\",\r\n    vectors_config={\r\n        \"dense\": VectorParams(size=384, distance=Distance.COSINE),\r\n        \"sparse\": VectorParams(size=30000, distance=Distance.DOT)\r\n    }\r\n)\r\n\r\n# Insert with named vectors\r\nclient.upsert(\r\n    collection_name=\"hybrid_search\",\r\n    points=[\r\n        PointStruct(\r\n            id=1,\r\n            vector={\r\n                \"dense\": dense_embedding,\r\n                \"sparse\": sparse_embedding\r\n            },\r\n            payload={\"text\": \"document text\"}\r\n        )\r\n    ]\r\n)\r\n\r\n# Search specific vector\r\nresults = client.search(\r\n    collection_name=\"hybrid_search\",\r\n    query_vector=(\"dense\", query_dense),  # Specify which vector\r\n    limit=10\r\n)\r\n```\r\n\r\n### Sparse vectors (BM25, SPLADE)\r\n\r\n```python\r\nfrom qdrant_client.models import SparseVectorParams, SparseIndexParams, SparseVector\r\n\r\n# Collection with sparse vectors\r\nclient.create_collection(\r\n    collection_name=\"sparse_search\",\r\n    vectors_config={},\r\n    sparse_vectors_config={\"text\": SparseVectorParams(index=SparseIndexParams(on_disk=False))}\r\n)\r\n\r\n# Insert sparse vector\r\nclient.upsert(\r\n    collection_name=\"sparse_search\",\r\n    points=[PointStruct(id=1, vector={\"text\": SparseVector(indices=[1, 5, 100], values=[0.5, 0.8, 0.2])}, payload={\"text\": \"document\"})]\r\n)\r\n```\r\n\r\n## Quantization (memory optimization)\r\n\r\n```python\r\nfrom qdrant_client.models import ScalarQuantization, ScalarQuantizationConfig, ScalarType\r\n\r\n# Scalar quantization (4x memory reduction)\r\nclient.create_collection(\r\n    collection_name=\"quantized\",\r\n    vectors_config=VectorParams(size=384, distance=Distance.COSINE),\r\n    quantization_config=ScalarQuantization(\r\n        scalar=ScalarQuantizationConfig(\r\n            type=ScalarType.INT8,\r\n            quantile=0.99,        # Clip outliers\r\n            always_ram=True      # Keep quantized in RAM\r\n        )\r\n    )\r\n)\r\n\r\n# Search with rescoring\r\nresults = client.search(\r\n    collection_name=\"quantized\",\r\n    query_vector=query,\r\n    search_params={\"quantization\": {\"rescore\": True}},  # Rescore top results\r\n    limit=10\r\n)\r\n```\r\n\r\n## Payload indexing\r\n\r\n```python\r\nfrom qdrant_client.models import PayloadSchemaType\r\n\r\n# Create payload index for faster filtering\r\nclient.create_payload_index(\r\n    collection_name=\"documents\",\r\n    field_name=\"category\",\r\n    field_schema=PayloadSchemaType.KEYWORD\r\n)\r\n\r\nclient.create_payload_index(\r\n    collection_name=\"documents\",\r\n    field_name=\"timestamp\",\r\n    field_schema=PayloadSchemaType.INTEGER\r\n)\r\n\r\n# Index types: KEYWORD, INTEGER, FLOAT, GEO, TEXT (full-text), BOOL\r\n```\r\n\r\n## Production deployment\r\n\r\n### Qdrant Cloud\r\n\r\n```python\r\nfrom qdrant_client import QdrantClient\r\n\r\n# Connect to Qdrant Cloud\r\nclient = QdrantClient(\r\n    url=\"https:\u002F\u002Fyour-cluster.cloud.qdrant.io\",\r\n    api_key=\"your-api-key\"\r\n)\r\n```\r\n\r\n### Performance tuning\r\n\r\n```python\r\n# Optimize for search speed (higher recall)\r\nclient.update_collection(\r\n    collection_name=\"documents\",\r\n    hnsw_config=HnswConfigDiff(ef_construct=200, m=32)\r\n)\r\n\r\n# Optimize for indexing speed (bulk loads)\r\nclient.update_collection(\r\n    collection_name=\"documents\",\r\n    optimizer_config={\"indexing_threshold\": 20000}\r\n)\r\n```\r\n\r\n## Best practices\r\n\r\n1. **Batch operations** - Use batch upsert\u002Fsearch for efficiency\r\n2. **Payload indexing** - Index fields used in filters\r\n3. **Quantization** - Enable for large collections (>1M vectors)\r\n4. **Sharding** - Use for collections >10M vectors\r\n5. **On-disk storage** - Enable `on_disk_payload` for large payloads\r\n6. **Connection pooling** - Reuse client instances\r\n\r\n## Common issues\r\n\r\n**Slow search with filters:**\r\n```python\r\n# Create payload index for filtered fields\r\nclient.create_payload_index(\r\n    collection_name=\"docs\",\r\n    field_name=\"category\",\r\n    field_schema=PayloadSchemaType.KEYWORD\r\n)\r\n```\r\n\r\n**Out of memory:**\r\n```python\r\n# Enable quantization and on-disk storage\r\nclient.create_collection(\r\n    collection_name=\"large_collection\",\r\n    vectors_config=VectorParams(size=384, distance=Distance.COSINE),\r\n    quantization_config=ScalarQuantization(...),\r\n    on_disk_payload=True\r\n)\r\n```\r\n\r\n**Connection issues:**\r\n```python\r\n# Use timeout and retry\r\nclient = QdrantClient(\r\n    host=\"localhost\",\r\n    port=6333,\r\n    timeout=30,\r\n    prefer_grpc=True  # gRPC for better performance\r\n)\r\n```\r\n\r\n## References\r\n\r\n- **[Advanced Usage](references\u002Fadvanced-usage.md)** - Distributed mode, hybrid search, recommendations\r\n- **[Troubleshooting](references\u002Ftroubleshooting.md)** - Common issues, debugging, performance tuning\r\n\r\n## Resources\r\n\r\n- **GitHub**: https:\u002F\u002Fgithub.com\u002Fqdrant\u002Fqdrant (22k+ stars)\r\n- **Docs**: https:\u002F\u002Fqdrant.tech\u002Fdocumentation\u002F\r\n- **Python Client**: https:\u002F\u002Fgithub.com\u002Fqdrant\u002Fqdrant-client\r\n- **Cloud**: https:\u002F\u002Fcloud.qdrant.io\r\n- **Version**: 1.12.0+\r\n- **License**: Apache 2.0\r\n","# Qdrant - Vector Similarity Search Engine\r\n\r\nHigh-performance vector database written in Rust for production RAG and semantic search.\r\n\r\n## When to use Qdrant\r\n\r\n**Use Qdrant when:**\r\n- Building production RAG systems requiring low latency\r\n- Need hybrid search (vectors + metadata filtering)\r\n- Require horizontal scaling with sharding\u002Freplication\r\n- Want on-premise deployment with full data control\r\n- Need multi-vector storage per record (dense + sparse)\r\n- Building real-time recommendation systems\r\n\r\n**Key features:**\r\n- **Rust-powered**: Memory-safe, high performance\r\n- **Rich filtering**: Filter by any payload field during search\r\n- **Multiple vectors**: Dense, sparse, multi-dense per point\r\n- **Quantization**: Scalar, product, binary for memory efficiency\r\n- **Distributed**: Raft consensus, sharding, replication\r\n- **REST + gRPC**: Both APIs with full feature parity\r\n\r\n**Use alternatives instead:**\r\n- **Chroma**: Simpler setup, embedded use cases\r\n- **FAISS**: Maximum raw speed, research\u002Fbatch processing\r\n- **Pinecone**: Fully managed, zero ops preferred\r\n- **Weaviate**: GraphQL preference, built-in vectorizers\r\n\r\n## Quick start\r\n\r\n### Installation\r\n\r\n```bash\r\n# Python client\r\npip install qdrant-client\r\n\r\n# Docker (recommended for development)\r\ndocker run -p 6333:6333 -p 6334:6334 qdrant\u002Fqdrant\r\n\r\n# Docker with persistent storage\r\ndocker run -p 6333:6333 -p 6334:6334 \\\r\n    -v $(pwd)\u002Fqdrant_storage:\u002Fqdrant\u002Fstorage \\\r\n    qdrant\u002Fqdrant\r\n```\r\n\r\n### Basic usage\r\n\r\n```python\r\nfrom qdrant_client import QdrantClient\r\nfrom qdrant_client.models import Distance, VectorParams, PointStruct\r\n\r\n# Connect to Qdrant\r\nclient = QdrantClient(host=\"localhost\", port=6333)\r\n\r\n# Create collection\r\nclient.create_collection(\r\n    collection_name=\"documents\",\r\n    vectors_config=VectorParams(size=384, distance=Distance.COSINE)\r\n)\r\n\r\n# Insert vectors with payload\r\nclient.upsert(\r\n    collection_name=\"documents\",\r\n    points=[\r\n        PointStruct(\r\n            id=1,\r\n            vector=[0.1, 0.2, ...],  # 384-dim vector\r\n            payload={\"title\": \"Doc 1\", \"category\": \"tech\"}\r\n        ),\r\n        PointStruct(\r\n            id=2,\r\n            vector=[0.3, 0.4, ...],\r\n            payload={\"title\": \"Doc 2\", \"category\": \"science\"}\r\n        )\r\n    ]\r\n)\r\n\r\n# Search with filtering\r\nresults = client.search(\r\n    collection_name=\"documents\",\r\n    query_vector=[0.15, 0.25, ...],\r\n    query_filter={\r\n        \"must\": [{\"key\": \"category\", \"match\": {\"value\": \"tech\"}}]\r\n    },\r\n    limit=10\r\n)\r\n\r\nfor point in results:\r\n    print(f\"ID: {point.id}, Score: {point.score}, Payload: {point.payload}\")\r\n```\r\n\r\n## Core concepts\r\n\r\n### Points - Basic data unit\r\n\r\n```python\r\nfrom qdrant_client.models import PointStruct\r\n\r\n# Point = ID + Vector(s) + Payload\r\npoint = PointStruct(\r\n    id=123,                              # Integer or UUID string\r\n    vector=[0.1, 0.2, 0.3, ...],        # Dense vector\r\n    payload={                            # Arbitrary JSON metadata\r\n        \"title\": \"Document title\",\r\n        \"category\": \"tech\",\r\n        \"timestamp\": 1699900000,\r\n        \"tags\": [\"python\", \"ml\"]\r\n    }\r\n)\r\n\r\n# Batch upsert (recommended)\r\nclient.upsert(\r\n    collection_name=\"documents\",\r\n    points=[point1, point2, point3],\r\n    wait=True  # Wait for indexing\r\n)\r\n```\r\n\r\n### Collections - Vector containers\r\n\r\n```python\r\nfrom qdrant_client.models import VectorParams, Distance, HnswConfigDiff\r\n\r\n# Create with HNSW configuration\r\nclient.create_collection(\r\n    collection_name=\"documents\",\r\n    vectors_config=VectorParams(\r\n        size=384,                        # Vector dimensions\r\n        distance=Distance.COSINE         # COSINE, EUCLID, DOT, MANHATTAN\r\n    ),\r\n    hnsw_config=HnswConfigDiff(\r\n        m=16,                            # Connections per node (default 16)\r\n        ef_construct=100,                # Build-time accuracy (default 100)\r\n        full_scan_threshold=10000        # Switch to brute force below this\r\n    ),\r\n    on_disk_payload=True                 # Store payload on disk\r\n)\r\n\r\n# Collection info\r\ninfo = client.get_collection(\"documents\")\r\nprint(f\"Points: {info.points_count}, Vectors: {info.vectors_count}\")\r\n```\r\n\r\n### Distance metrics\r\n\r\n| Metric | Use Case | Range |\r\n|--------|----------|-------|\r\n| `COSINE` | Text embeddings, normalized vectors | 0 to 2 |\r\n| `EUCLID` | Spatial data, image features | 0 to ∞ |\r\n| `DOT` | Recommendations, unnormalized | -∞ to ∞ |\r\n| `MANHATTAN` | Sparse features, discrete data | 0 to ∞ |\r\n\r\n## Search operations\r\n\r\n### Basic search\r\n\r\n```python\r\n# Simple nearest neighbor search\r\nresults = client.search(\r\n    collection_name=\"documents\",\r\n    query_vector=[0.1, 0.2, ...],\r\n    limit=10,\r\n    with_payload=True,\r\n    with_vectors=False  # Don't return vectors (faster)\r\n)\r\n```\r\n\r\n### Filtered search\r\n\r\n```python\r\nfrom qdrant_client.models import Filter, FieldCondition, MatchValue, Range\r\n\r\n# Complex filtering\r\nresults = client.search(\r\n    collection_name=\"documents\",\r\n    query_vector=query_embedding,\r\n    query_filter=Filter(\r\n        must=[\r\n            FieldCondition(key=\"category\", match=MatchValue(value=\"tech\")),\r\n            FieldCondition(key=\"timestamp\", range=Range(gte=1699000000))\r\n        ],\r\n        must_not=[\r\n            FieldCondition(key=\"status\", match=MatchValue(value=\"archived\"))\r\n        ]\r\n    ),\r\n    limit=10\r\n)\r\n\r\n# Shorthand filter syntax\r\nresults = client.search(\r\n    collection_name=\"documents\",\r\n    query_vector=query_embedding,\r\n    query_filter={\r\n        \"must\": [\r\n            {\"key\": \"category\", \"match\": {\"value\": \"tech\"}},\r\n            {\"key\": \"price\", \"range\": {\"gte\": 10, \"lte\": 100}}\r\n        ]\r\n    },\r\n    limit=10\r\n)\r\n```\r\n\r\n### Batch search\r\n\r\n```python\r\nfrom qdrant_client.models import SearchRequest\r\n\r\n# Multiple queries in one request\r\nresults = client.search_batch(\r\n    collection_name=\"documents\",\r\n    requests=[\r\n        SearchRequest(vector=[0.1, ...], limit=5),\r\n        SearchRequest(vector=[0.2, ...], limit=5, filter={\"must\": [...]}),\r\n        SearchRequest(vector=[0.3, ...], limit=10)\r\n    ]\r\n)\r\n```\r\n\r\n## RAG integration\r\n\r\n### With sentence-transformers\r\n\r\n```python\r\nfrom sentence_transformers import SentenceTransformer\r\nfrom qdrant_client import QdrantClient\r\nfrom qdrant_client.models import VectorParams, Distance, PointStruct\r\n\r\n# Initialize\r\nencoder = SentenceTransformer(\"all-MiniLM-L6-v2\")\r\nclient = QdrantClient(host=\"localhost\", port=6333)\r\n\r\n# Create collection\r\nclient.create_collection(\r\n    collection_name=\"knowledge_base\",\r\n    vectors_config=VectorParams(size=384, distance=Distance.COSINE)\r\n)\r\n\r\n# Index documents\r\ndocuments = [\r\n    {\"id\": 1, \"text\": \"Python is a programming language\", \"source\": \"wiki\"},\r\n    {\"id\": 2, \"text\": \"Machine learning uses algorithms\", \"source\": \"textbook\"},\r\n]\r\n\r\npoints = [\r\n    PointStruct(\r\n        id=doc[\"id\"],\r\n        vector=encoder.encode(doc[\"text\"]).tolist(),\r\n        payload={\"text\": doc[\"text\"], \"source\": doc[\"source\"]}\r\n    )\r\n    for doc in documents\r\n]\r\nclient.upsert(collection_name=\"knowledge_base\", points=points)\r\n\r\n# RAG retrieval\r\ndef retrieve(query: str, top_k: int = 5) -> list[dict]:\r\n    query_vector = encoder.encode(query).tolist()\r\n    results = client.search(\r\n        collection_name=\"knowledge_base\",\r\n        query_vector=query_vector,\r\n        limit=top_k\r\n    )\r\n    return [{\"text\": r.payload[\"text\"], \"score\": r.score} for r in results]\r\n\r\n# Use in RAG pipeline\r\ncontext = retrieve(\"What is Python?\")\r\nprompt = f\"Context: {context}\\n\\nQuestion: What is Python?\"\r\n```\r\n\r\n### With LangChain\r\n\r\n```python\r\nfrom langchain_community.vectorstores import Qdrant\r\nfrom langchain_community.embeddings import HuggingFaceEmbeddings\r\n\r\nembeddings = HuggingFaceEmbeddings(model_name=\"all-MiniLM-L6-v2\")\r\nvectorstore = Qdrant.from_documents(documents, embeddings, url=\"http:\u002F\u002Flocalhost:6333\", collection_name=\"docs\")\r\nretriever = vectorstore.as_retriever(search_kwargs={\"k\": 5})\r\n```\r\n\r\n### With LlamaIndex\r\n\r\n```python\r\nfrom llama_index.vector_stores.qdrant import QdrantVectorStore\r\nfrom llama_index.core import VectorStoreIndex, StorageContext\r\n\r\nvector_store = QdrantVectorStore(client=client, collection_name=\"llama_docs\")\r\nstorage_context = StorageContext.from_defaults(vector_store=vector_store)\r\nindex = VectorStoreIndex.from_documents(documents, storage_context=storage_context)\r\nquery_engine = index.as_query_engine()\r\n```\r\n\r\n## Multi-vector support\r\n\r\n### Named vectors (different embedding models)\r\n\r\n```python\r\nfrom qdrant_client.models import VectorParams, Distance\r\n\r\n# Collection with multiple vector types\r\nclient.create_collection(\r\n    collection_name=\"hybrid_search\",\r\n    vectors_config={\r\n        \"dense\": VectorParams(size=384, distance=Distance.COSINE),\r\n        \"sparse\": VectorParams(size=30000, distance=Distance.DOT)\r\n    }\r\n)\r\n\r\n# Insert with named vectors\r\nclient.upsert(\r\n    collection_name=\"hybrid_search\",\r\n    points=[\r\n        PointStruct(\r\n            id=1,\r\n            vector={\r\n                \"dense\": dense_embedding,\r\n                \"sparse\": sparse_embedding\r\n            },\r\n            payload={\"text\": \"document text\"}\r\n        )\r\n    ]\r\n)\r\n\r\n# Search specific vector\r\nresults = client.search(\r\n    collection_name=\"hybrid_search\",\r\n    query_vector=(\"dense\", query_dense),  # Specify which vector\r\n    limit=10\r\n)\r\n```\r\n\r\n### Sparse vectors (BM25, SPLADE)\r\n\r\n```python\r\nfrom qdrant_client.models import SparseVectorParams, SparseIndexParams, SparseVector\r\n\r\n# Collection with sparse vectors\r\nclient.create_collection(\r\n    collection_name=\"sparse_search\",\r\n    vectors_config={},\r\n    sparse_vectors_config={\"text\": SparseVectorParams(index=SparseIndexParams(on_disk=False))}\r\n)\r\n\r\n# Insert sparse vector\r\nclient.upsert(\r\n    collection_name=\"sparse_search\",\r\n    points=[PointStruct(id=1, vector={\"text\": SparseVector(indices=[1, 5, 100], values=[0.5, 0.8, 0.2])}, payload={\"text\": \"document\"})]\r\n)\r\n```\r\n\r\n## Quantization (memory optimization)\r\n\r\n```python\r\nfrom qdrant_client.models import ScalarQuantization, ScalarQuantizationConfig, ScalarType\r\n\r\n# Scalar quantization (4x memory reduction)\r\nclient.create_collection(\r\n    collection_name=\"quantized\",\r\n    vectors_config=VectorParams(size=384, distance=Distance.COSINE),\r\n    quantization_config=ScalarQuantization(\r\n        scalar=ScalarQuantizationConfig(\r\n            type=ScalarType.INT8,\r\n            quantile=0.99,        # Clip outliers\r\n            always_ram=True      # Keep quantized in RAM\r\n        )\r\n    )\r\n)\r\n\r\n# Search with rescoring\r\nresults = client.search(\r\n    collection_name=\"quantized\",\r\n    query_vector=query,\r\n    search_params={\"quantization\": {\"rescore\": True}},  # Rescore top results\r\n    limit=10\r\n)\r\n```\r\n\r\n## Payload indexing\r\n\r\n```python\r\nfrom qdrant_client.models import PayloadSchemaType\r\n\r\n# Create payload index for faster filtering\r\nclient.create_payload_index(\r\n    collection_name=\"documents\",\r\n    field_name=\"category\",\r\n    field_schema=PayloadSchemaType.KEYWORD\r\n)\r\n\r\nclient.create_payload_index(\r\n    collection_name=\"documents\",\r\n    field_name=\"timestamp\",\r\n    field_schema=PayloadSchemaType.INTEGER\r\n)\r\n\r\n# Index types: KEYWORD, INTEGER, FLOAT, GEO, TEXT (full-text), BOOL\r\n```\r\n\r\n## Production deployment\r\n\r\n### Qdrant Cloud\r\n\r\n```python\r\nfrom qdrant_client import QdrantClient\r\n\r\n# Connect to Qdrant Cloud\r\nclient = QdrantClient(\r\n    url=\"https:\u002F\u002Fyour-cluster.cloud.qdrant.io\",\r\n    api_key=\"your-api-key\"\r\n)\r\n```\r\n\r\n### Performance tuning\r\n\r\n```python\r\n# Optimize for search speed (higher recall)\r\nclient.update_collection(\r\n    collection_name=\"documents\",\r\n    hnsw_config=HnswConfigDiff(ef_construct=200, m=32)\r\n)\r\n\r\n# Optimize for indexing speed (bulk loads)\r\nclient.update_collection(\r\n    collection_name=\"documents\",\r\n    optimizer_config={\"indexing_threshold\": 20000}\r\n)\r\n```\r\n\r\n## Best practices\r\n\r\n1. **Batch operations** - Use batch upsert\u002Fsearch for efficiency\r\n2. **Payload indexing** - Index fields used in filters\r\n3. **Quantization** - Enable for large collections (>1M vectors)\r\n4. **Sharding** - Use for collections >10M vectors\r\n5. **On-disk storage** - Enable `on_disk_payload` for large payloads\r\n6. **Connection pooling** - Reuse client instances\r\n\r\n## Common issues\r\n\r\n**Slow search with filters:**\r\n```python\r\n# Create payload index for filtered fields\r\nclient.create_payload_index(\r\n    collection_name=\"docs\",\r\n    field_name=\"category\",\r\n    field_schema=PayloadSchemaType.KEYWORD\r\n)\r\n```\r\n\r\n**Out of memory:**\r\n```python\r\n# Enable quantization and on-disk storage\r\nclient.create_collection(\r\n    collection_name=\"large_collection\",\r\n    vectors_config=VectorParams(size=384, distance=Distance.COSINE),\r\n    quantization_config=ScalarQuantization(...),\r\n    on_disk_payload=True\r\n)\r\n```\r\n\r\n**Connection issues:**\r\n```python\r\n# Use timeout and retry\r\nclient = QdrantClient(\r\n    host=\"localhost\",\r\n    port=6333,\r\n    timeout=30,\r\n    prefer_grpc=True  # gRPC for better performance\r\n)\r\n```\r\n\r\n## References\r\n\r\n- **[Advanced Usage](references\u002Fadvanced-usage.md)** - Distributed mode, hybrid search, recommendations\r\n- **[Troubleshooting](references\u002Ftroubleshooting.md)** - Common issues, debugging, performance tuning\r\n\r\n## Resources\r\n\r\n- **GitHub**: https:\u002F\u002Fgithub.com\u002Fqdrant\u002Fqdrant (22k+ stars)\r\n- **Docs**: https:\u002F\u002Fqdrant.tech\u002Fdocumentation\u002F\r\n- **Python Client**: https:\u002F\u002Fgithub.com\u002Fqdrant\u002Fqdrant-client\r\n- **Cloud**: https:\u002F\u002Fcloud.qdrant.io\r\n- **Version**: 1.12.0+\r\n- **License**: Apache 2.0\r\n",{"name":5,"author":44,"license":15,"version":45,"metadata":46,"platforms":56,"description":7,"dependencies":57},"Orchestra Research","1.0.0",{"hermes":47},{"tags":48},[49,21,20,50,51,52,53,54,55],"RAG","Semantic Search","Embeddings","Similarity Search","HNSW","Production","Distributed",[38,25,33],[58],"qdrant-client>=1.12.0","main","0d1434f89f4bc6b07e9d272accd8c1013768820f","archive\u002Fskill-md-reviewed\u002Fsha256\u002F50\u002F504b9d36683ec89ee492f9770c8c49c8d763f73ffbf2c677c74da16caf1c90e9",0,"https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills","https:\u002F\u002Fgithub.com\u002Fzapabob","The Open University of JAPAN　2019-now\r\n\r\nTokyo University of Science　2007-2010","2026-10-26 16:29:01",{"source":8,"skillName":5,"commands":68,"prompt":78,"zipAvailable":79},[69,72,75],{"tool":70,"command":71},"npx","npx skills add zapabob\u002FSkills --skill qdrant-vector-search",{"tool":73,"command":74},"bunx","bunx skills add zapabob\u002FSkills --skill qdrant-vector-search",{"tool":76,"command":77},"pnpm","pnpm dlx skills add zapabob\u002FSkills --skill qdrant-vector-search","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Farchive\u002Fskill-md-reviewed\u002Fsha256\u002F50\u002F504b9d36683ec89ee492f9770c8c49c8d763f73ffbf2c677c74da16caf1c90e9\u002FSKILL.md",true,[],[82,116,151,177,204,221],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":101,"publishedAt":39},"01M6GJTE1GESGGF9K028Y7753E","xurl","X\u002FTwitter via xurl CLI: post, search, DM, media, v2 API.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fxurl\u002Fvariants\u002Fhermes-agent\u002FSKILL.md",[88,93,96],{"categoryNo":89,"categoryName":90,"taxonomyType":91,"taxonomyTypeDesc":92},"01JBSEEDCAT000000000000010","后端与接口","SCENARIO","使用场景",{"categoryNo":94,"categoryName":95,"taxonomyType":91,"taxonomyTypeDesc":92},"01JBSEEDCAT000000000000023","Skill与智能体开发",{"categoryNo":97,"categoryName":98,"taxonomyType":99,"taxonomyTypeDesc":100},"01JBSEEDCAT000000000000030","开发工程师","PROFESSION","适用职业",[84,102,25,103,104,38,33,105,23,106,107,108,109,110,111,112,59,113,114,115],"openclaw","cli","github","official","https","social-media","api","Media","hermes","hermes agent","twitter","social","via","post",{"skillNo":117,"name":118,"nameZh":6,"description":119,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":120,"categories":121,"keywords":133,"publishedAt":39},"01M6GJT8S1PXQ8WPXT1HSNBWH2","writing-plans","Write implementation plans: bite-sized tasks, paths, code.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fwriting-plans\u002Fvariants\u002Fhermes-agent\u002FSKILL.md",[122,125,128,129,130],{"categoryNo":123,"categoryName":124,"taxonomyType":91,"taxonomyTypeDesc":92},"01JBSEEDCAT000000000000015","视觉设计",{"categoryNo":126,"categoryName":127,"taxonomyType":91,"taxonomyTypeDesc":92},"01JBSEEDCAT000000000000018","写作与文案",{"categoryNo":94,"categoryName":95,"taxonomyType":91,"taxonomyTypeDesc":92},{"categoryNo":97,"categoryName":98,"taxonomyType":99,"taxonomyTypeDesc":100},{"categoryNo":131,"categoryName":132,"taxonomyType":99,"taxonomyTypeDesc":100},"01JBSEEDCAT000000000000036","内容创作者",[134,135,136,137,25,138,38,139,33,140,141,142,143,144,145,146,147,148,149,150],"writing","writing plans","plans","design","review","workflow","documentation","tasks","development","code","planning","code-review","subagent","write","superpowers","paths","test-driven-development",{"skillNo":152,"name":153,"nameZh":6,"description":154,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":155,"categories":156,"keywords":160,"publishedAt":39},"01M6GJST0H5X7WHBJ3BZK0DVKH","whisper","OpenAI's general-purpose speech recognition model. Supports 99 languages, transcription, translation to English, and language identification. Six model sizes from tiny (39M params) to large (1550M params). Use for speech-to-text, podcast transcription, or multilingual audio processing. Best for robust, multilingual ASR.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fwhisper\u002Fvariants\u002Fhermes-agent-optional\u002FSKILL.md",[157],{"categoryNo":158,"categoryName":159,"taxonomyType":91,"taxonomyTypeDesc":92},"01JBSEEDCAT000000000000016","图像与多媒体",[153,25,38,161,162,163,164,165,166,167,168,169,26,170,171,172,173,174,175,176],"openai","speech","multimodal","Transformers","ASR","Audio","text","model","speech recognition","general","multilingual","translation","english","identification","processing","Audio Processing",{"skillNo":178,"name":179,"nameZh":6,"description":180,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":181,"categories":182,"keywords":186,"publishedAt":39},"01M6GJSMRQ8EETZTYMG72P70EA","weights-and-biases","W&B: log ML experiments, sweeps, model registry, dashboards.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fweights-and-biases\u002Fvariants\u002Fhermes-agent\u002FSKILL.md",[183],{"categoryNo":184,"categoryName":185,"taxonomyType":91,"taxonomyTypeDesc":92},"01JBSEEDCAT000000000000013","分析与可视化",[187,25,188,189,33,190,191,192,193,194,195,168,196,197,198,199,200,201,202,203],"weights and biases","tensorflow","experiments","pytorch","visualization","registry","real-time","tuning","dashboards","experiment","log","time","Experiment Tracking","Model Registry","Hyperparameter Tuning","WandB","Real-Time Visualization",{"skillNo":205,"name":206,"nameZh":6,"description":207,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":208,"categories":209,"keywords":211,"publishedAt":39},"01M6GJS19KDPRY112JTR326S06","watchers","Poll RSS, JSON APIs, and GitHub with watermark dedup.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fwatchers\u002Fvariants\u002Fhermes-agent-optional\u002FSKILL.md",[210],{"categoryNo":94,"categoryName":95,"taxonomyType":91,"taxonomyTypeDesc":92},[206,25,212,104,38,213,214,215,216,217,218,110,111,219,220],"terminal","automation","http","monitoring","cron","json","rss","devops","with",{"skillNo":222,"name":223,"nameZh":6,"description":224,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":225,"categories":226,"keywords":229,"publishedAt":39},"01M6GJRHJD2A6Q830SWF9W48BC","serving-llms-vllm","vLLM: high-throughput LLM serving, OpenAI API, quantization.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fvllm\u002Fvariants\u002Fhermes-agent\u002FSKILL.md",[227,228],{"categoryNo":89,"categoryName":90,"taxonomyType":91,"taxonomyTypeDesc":92},{"categoryNo":97,"categoryName":98,"taxonomyType":99,"taxonomyTypeDesc":100},[230,231,232,38,35,161,233,234,235,164,54,108,236,237,238,239,240,241,242,243],"llms","vLLM","serving llms vllm","openai api","tensor","Inference Serving","llm serving","High Throughput","PagedAttention","Continuous Batching","Quantization","Tensor Parallelism","high-throughput","throughput"]