Database vettoriali self-hosted con Qdrant: un'immersione profonda (Italiano (Italian))

Database vettoriali self-hosted con Qdrant: un'immersione profonda

Sunday, 23 November 2025

//

9 minute read

Introduzione

In relazione alla serie RAG: Questo articolo fornisce un'immersione profonda in Qdrant, il database vettoriale utilizzato in:

QdrantCity name (optional, probably does not need a translation) (pronunciato "quadrant") è un database vettoriale open-source costruito in Rust. Questo articolo copre i concetti fondamentali, il client C#, la messa a punto delle prestazioni e i modelli di produzione.

Che cos'è Qdrant?

A banca dati vettoriale memorizza vettori ad alta dimensione (imbeddings) e consente una rapida ricerca di somiglianza. A differenza dei database tradizionali che trovano corrispondenze esatte, Qdrant trova Simile semanticamente oggetti.

flowchart LR
    A[Text: 'Docker deployment'] --> B[Embedding Model]
    B --> C["Vector: [0.12, -0.34, 0.56, ...]"]
    C --> D[Qdrant]
    E[Query: 'container setup'] --> F[Embedding Model]
    F --> G["Vector: [0.11, -0.32, 0.58, ...]"]
    G --> H[Similarity Search]
    D --> H
    H --> I[Similar Results]

    style B stroke:#6366f1,stroke-width:3px
    style D stroke:#ef4444,stroke-width:3px
    style F stroke:#6366f1,stroke-width:3px
    style H stroke:#10b981,stroke-width:2px

Caratteristiche chiave di Qdrant:

Concetti principali

Collezioni

A raccolta è come una tabella - contiene vettori con una dimensionalità fissa e una metrica di distanza.

flowchart TB
    subgraph Collection["Collection: blog_posts"]
        A[Vector Size: 384]
        B[Distance: Cosine]
        C[HNSW Index]
    end

    subgraph Points
        D[Point 1: slug=docker-intro]
        E[Point 2: slug=kubernetes-basics]
        F[Point N...]
    end

    Collection --> Points

    style A stroke:#6366f1,stroke-width:2px
    style B stroke:#6366f1,stroke-width:2px
    style C stroke:#f59e0b,stroke-width:2px
    style D stroke:#10b981,stroke-width:2px
    style E stroke:#10b981,stroke-width:2px
// Create collection - see https://qdrant.tech/documentation/concepts/collections/#create-a-collection
await client.CreateCollectionAsync(
    collectionName: "blog_posts",
    vectorsConfig: new VectorParams
    {
        Size = 384,              // Must match your embedding model
        Distance = Distance.Cosine  // Best for text embeddings
    }
);

Metri di distanza (docs):

  • Coseno - Misura l'angolo tra i vettori (meglio per il testo)
  • Punto - Prodotto interno grezzo (per vettori prenormalizzati)
  • EuclidCity name (optional, probably does not need a translation) - Distanza geometrica (per dati territoriali)

Punti

A punto è un singolo record contenente:

flowchart LR
    subgraph Point
        A[ID: uuid/int]
        B["Vector: float[384]"]
        C[Payload: JSON metadata]
    end

    style A stroke:#8b5cf6,stroke-width:2px
    style B stroke:#f59e0b,stroke-width:2px
    style C stroke:#10b981,stroke-width:2px
// Upsert points - see https://qdrant.tech/documentation/concepts/points/#upload-points
var point = new PointStruct
{
    Id = new PointId { Uuid = Guid.NewGuid().ToString() },
    Vectors = embedding,  // float[384]
    Payload =
    {
        ["slug"] = "my-post",
        ["title"] = "Vector Databases",
        ["language"] = "en",
        ["categories"] = new[] { "AI", "Databases" },
        ["published"] = DateTimeOffset.UtcNow.ToUnixTimeSeconds()
    }
};

await client.UpsertAsync("blog_posts", points: new[] { point });

Filtraggio

Filtraggio esegui prima ricerca di somiglianza - estremamente efficiente.

flowchart TB
    A[Search Query] --> B{Apply Filters First}
    B --> C[Language = 'en']
    B --> D[Year >= 2024]
    C --> E[Filtered Subset]
    D --> E
    E --> F[Vector Similarity Search]
    F --> G[Ranked Results]

    style B stroke:#ec4899,stroke-width:3px
    style E stroke:#f59e0b,stroke-width:2px
    style F stroke:#6366f1,stroke-width:2px
    style G stroke:#10b981,stroke-width:2px
// Filter conditions - see https://qdrant.tech/documentation/concepts/filtering/#filtering-conditions
var filter = new Filter
{
    Must =  // AND conditions
    {
        new Condition { Field = new FieldCondition
        {
            Key = "language",
            Match = new Match { Keyword = "en" }
        }},
        new Condition { Field = new FieldCondition
        {
            Key = "published",
            Range = new Range { Gte = 1704067200 }  // 2024-01-01
        }}
    },
    MustNot =  // Exclude conditions
    {
        new Condition { Field = new FieldCondition
        {
            Key = "slug",
            Match = new Match { Keyword = "draft-post" }
        }}
    }
};

Tipo di filtro (docs):

  • Match.Keyword - Corrispondenza stringa esatta
  • Match.Text - Full-text match
  • Match.Any - Corrisponde a qualsiasi array
  • Range - Gamme numeriche (Gte, Lte, Gt, Lt)
  • GeoBoundingBox / GeoRadius - Geofiltraggio

Il client C#

Installare il funzionario Qdrant.Client pacchetto (GitHubCity name (optional, probably does not need a translation)):

dotnet add package Qdrant.Client

Configurazione connessione

using Qdrant.Client;
using Qdrant.Client.Grpc;

// gRPC client (recommended) - see https://qdrant.tech/documentation/interfaces/#grpc-interface
var client = new QdrantClient(
    host: "localhost",
    port: 6334,  // gRPC port (6333 is REST)
    https: false
);

// With API key - see https://qdrant.tech/documentation/guides/security/
var secureClient = new QdrantClient(
    host: "your-qdrant.cloud",
    port: 6334,
    https: true,
    apiKey: "your-api-key"
);

Usare sempre gRPC (porto 6334) per la produzione - 3-5x più veloce del REST.

Windows HTTP/2 Fix

Su Windows, abilitare HTTP/2 non cifrato prima creazione del client:

AppContext.SetSwitch("System.Net.Http.SocketsHttpHandler.Http2UnencryptedSupport", true);

Operazioni chiave

Cerca

// Vector search - see https://qdrant.tech/documentation/concepts/search/
var results = await client.SearchAsync(
    collectionName: "blog_posts",
    vector: queryEmbedding,
    limit: 10,
    filter: filter,
    scoreThreshold: 0.5f,  // Minimum similarity
    searchParams: new SearchParams
    {
        HnswEf = 128,  // Search accuracy (higher = better recall)
        Exact = false  // Use approximate search
    },
    withPayload: true
);

foreach (var result in results)
{
    Console.WriteLine($"{result.Payload["title"].StringValue}: {result.Score}");
}

Batch Upsert

// Batch operations - see https://qdrant.tech/documentation/concepts/points/#batch-update
var points = documents.Select(doc => new PointStruct
{
    Id = new PointId { Uuid = doc.Id },
    Vectors = doc.Embedding,
    Payload = { ["slug"] = doc.Slug, ["title"] = doc.Title }
}).ToList();

await client.UpsertAsync(
    collectionName: "blog_posts",
    points: points,
    wait: true  // Wait for indexing
);

Elimina

// Delete by filter - see https://qdrant.tech/documentation/concepts/points/#delete-points
await client.DeleteAsync(
    collectionName: "blog_posts",
    filter: new Filter
    {
        Must = { new Condition { Field = new FieldCondition
        {
            Key = "slug",
            Match = new Match { Keyword = "old-post" }
        }}}
    }
);

Tuning dell'indice HNSW

HNSWCity name (optional, probably does not need a translation) (Hierarchical Navigable Small World) è l'algoritmo indice di Qdrant.

flowchart TB
    subgraph "HNSW Graph Layers"
        L2[Layer 2 - Sparse]
        L1[Layer 1 - Medium]
        L0[Layer 0 - Dense]
    end

    Q[Query] --> L2
    L2 --> L1
    L1 --> L0
    L0 --> R[Nearest Neighbors]

    style L2 stroke:#8b5cf6,stroke-width:2px
    style L1 stroke:#6366f1,stroke-width:2px
    style L0 stroke:#3b82f6,stroke-width:2px
    style Q stroke:#10b981,stroke-width:2px
    style R stroke:#ef4444,stroke-width:2px

Parametri dell'indice

// HNSW config - see https://qdrant.tech/documentation/concepts/indexing/#hnsw-index
var hnswConfig = new HnswConfigDiff
{
    M = 16,              // Edges per node (16-32 recommended)
    EfConstruct = 100,   // Build-time accuracy (100-200)
    FullScanThreshold = 10000  // Brute force threshold
};

await client.UpdateCollectionAsync(
    collectionName: "blog_posts",
    hnswConfig: hnswConfig
);

Precisione del tempo di ricerca:

var searchParams = new SearchParams
{
    HnswEf = 128  // Higher = better recall, slower (64-256)
};

Linee guida per la sintonizzazione:

Caso d'uso M EfConstruct HnswEf
Richiamo rapido e basso 8 64 32
Bilanciato 16 100 128
Richiamo alto 32 200 256

Indici di carico

Crea indici di carico per i campi filtrati frequentemente:

// Keyword index - see https://qdrant.tech/documentation/concepts/indexing/#payload-index
await client.CreatePayloadIndexAsync(
    collectionName: "blog_posts",
    fieldName: "language",
    schemaType: PayloadSchemaType.Keyword
);

// Integer index for ranges
await client.CreatePayloadIndexAsync(
    collectionName: "blog_posts",
    fieldName: "published",
    schemaType: PayloadSchemaType.Integer
);

Impatto: 10-100x filtraggio più veloce su grandi collezioni.

Quantizzazione

Quantizzazione riduce l'utilizzo della memoria:

// Scalar quantization - see https://qdrant.tech/documentation/guides/quantization/#scalar-quantization
await client.UpdateCollectionAsync(
    collectionName: "blog_posts",
    quantizationConfig: new ScalarQuantization
    {
        Scalar = new ScalarQuantizationConfig
        {
            Type = ScalarType.Int8,  // float32 -> int8
            Quantile = 0.99f,
            AlwaysRam = true
        }
    }
);

Trade-off: 4x meno memoria, ~2% perdita di richiamo, 1.5x ricerca più veloce.

Docker Deployment

# docker-compose.yml - see https://qdrant.tech/documentation/guides/installation/
services:
  qdrant:
    image: qdrant/qdrant:v1.12.1  # Pin version!
    ports:
      - "6333:6333"  # REST
      - "6334:6334"  # gRPC
    volumes:
      - qdrant_data:/qdrant/storage
    environment:
      - QDRANT__SERVICE__GRPC_PORT=6334
      - QDRANT__SERVICE__HTTP_PORT=6333
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:6333/health"]
      interval: 30s
      timeout: 10s
      retries: 3

volumes:
  qdrant_data:

Sicurezza

Abilita Autenticazione chiave API:

environment:
  - QDRANT__SERVICE__API_KEY=your-secret-key

Monitoraggio

MetricsCity name (optional, probably does not need a translation)

Qdrant espone Prometheus metriche a /metrics:

curl http://localhost:6333/metrics

metriche chiave:

  • qdrant_collections_vector_count - Totale vettori
  • qdrant_rest_responses_duration_seconds - Interrogazione di latenza
  • qdrant_memory_usage_bytes - Consumo di memoria

Istantanee

Crea backup:

# Create snapshot
curl -X POST http://localhost:6333/collections/blog_posts/snapshots

# List snapshots
curl http://localhost:6333/collections/blog_posts/snapshots

# Restore (copy snapshot to storage/collections/blog_posts/snapshots/)

Common GotchasCity name (optional, probably does not need a translation)

1. Confusione del porto

  • 6333 = API REST
  • 6334 = API GRPC (usa questo!)

2. Errore di dimensione vettoriale

Error: expected dim: 384, got 768

Il modello e la collezione devono corrispondere:

  • all-MiniLM-L6-v2: 384 dimensioni
  • nomic-embed-text: 768 dimensioni
  • OpenAICity name (optional, probably does not need a translation) text-embedding-3-small: 1536 dimensioni

3. Prima domanda lenta

HNSW carica pigro nella memoria. Riscaldamento dopo l'avvio:

await client.SearchAsync("blog_posts", new float[384], limit: 1);

4. Filtraggio dell'array

Uso Match.Any per i campi array:

new Match { Any = new RepeatedStrings { Strings = { "AI", "ML" } } }

Risorse

Documentazione ufficiale Qdrant

Librerie dei clienti

Articoli correlati

Codice sorgente

Tutti i codici disponibili al seguente indirizzo: github.com/scottgal/mostlylucidweb

  • Mostlylucid.SemanticSearch/Services/QdrantVectorStoreService.cs - Integrazione Qdrant
Finding related posts...
logo

© 2026 Scott Galloway — Unlicense — All content and source code on this site is free to use, copy, modify, and sell.