Skip to main content

Vector Search

FluidGrids provides advanced vector search capabilities that enable semantic search and similarity matching across your data. Our platform integrates with leading vector databases and provides optimized search algorithms for various use cases.

Core Features​

Vector Database Integration Supported vector stores:

FluidGrids integrates with major vector databases:

  • Pinecone for scalable vector search
  • Weaviate for semantic search
  • Milvus for high-performance queries
  • Qdrant for filtered search
  • ChromaDB for local development

Embedding Generation Multiple embedding options:

Support for various embedding models:

  • OpenAI embeddings
  • Cohere embeddings
  • HuggingFace models
  • Custom embeddings
  • Multi-modal embeddings

Implementation Guide​

Basic Setup Configure vector search:

from fluidgrids.vector import VectorStore

store = VectorStore(
    provider="pinecone",
    dimension=1536,
    metric="cosine"
)

# Index documents
store.index_documents(
    documents=[doc1, doc2],
    batch_size=100
)

Search Configuration Implement search functionality:

# Semantic search
results = store.search(
    query="example query",
    top_k=5,
    threshold=0.8
)

# Hybrid search
results = store.hybrid_search(
    query="example query",
    filters={"category": "tech"},
    weights={"semantic": 0.7, "keyword": 0.3}
)

Advanced Features​

Filtering & Faceting Advanced query capabilities:

# Filtered search
results = store.search(
    query="example",
    filters={
        "date": {"$gt": "2023-01-01"},
        "category": ["tech", "ai"]
    }
)

# Faceted search
results = store.faceted_search(
    query="example",
    facets=["category", "author"],
    facet_size=10
)

Clustering & Analysis Data analysis features:

from fluidgrids.vector import VectorAnalytics

# Cluster vectors
clusters = VectorAnalytics.cluster(
    vectors,
    n_clusters=5,
    algorithm="kmeans"
)

# Analyze similarity
similarity = VectorAnalytics.similarity_matrix(
    vectors,
    metric="cosine"
)

Performance Optimization​

Index Configuration Optimize index performance:

index_config:
  shards: 3
  replicas: 2
  pod_type: p1.x1
  metadata_config:
    indexed: ["category", "date"]

Query Optimization Enhance search performance:

# Optimized batch search
results = store.batch_search(
    queries=["query1", "query2"],
    config={
        "ef": 100,
        "nprobe": 10
    }
)

Monitoring & Analytics​

Performance Metrics Track search performance:

metrics = store.get_metrics(
    timeframe="1h",
    include=[
        "latency",
        "throughput",
        "cache_hits"
    ]
)

Usage Analytics Monitor system usage:

analytics = store.get_analytics(
    start_date="2024-01-01",
    metrics=["queries", "indexing"]
)

Best Practices​

Implementation Guidelines Follow these practices:

  • Choose appropriate embedding models
  • Optimize index configuration
  • Implement proper filtering
  • Monitor performance metrics
  • Regular index maintenance

Search Optimization Enhance search quality:

  • Fine-tune similarity thresholds
  • Implement hybrid search
  • Use appropriate filters
  • Optimize batch sizes
  • Regular performance testing

Getting Started​

Begin implementing vector search:

For vector search support, contact our AI Team.