Laravel 13 向量搜索实战

摘要

本文将通过实战案例演示如何使用 Laravel 13 构建智能语义搜索系统。内容包括：

  • 向量搜索原理与应用场景
  • 数据库设计与迁移
  • 嵌入生成与存储
  • 相似性搜索实现
  • 搜索结果优化
  • 完整实战案例

本文适合希望构建智能搜索功能的 Laravel 开发者。

1. 向量搜索原理

1.1 核心概念

向量搜索通过将文本转换为高维向量，然后计算向量之间的相似度来实现语义搜索。

1.2 应用场景

  • 文档搜索
  • 产品推荐
  • 相似内容发现
  • 问答系统

2. 数据库设计

2.1 迁移文件

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
<?php

use Illuminate\Database\Migrations\Migration;
use Illuminate\Support\Facades\DB;
use Illuminate\Support\Facades\Schema;

return new class extends Migration
{
    public function up(): void
    {
        DB::statement('CREATE EXTENSION IF NOT EXISTS vector');
        
        Schema::create('documents', function ($table) {
            $table->id();
            $table->string('title');
            $table->text('content');
            $table->vector('embedding', 1536);
            $table->timestamps();
        });
        
        DB::statement('CREATE INDEX documents_embedding_idx ON documents USING ivfflat (embedding vector_cosine_ops) WITH (lists = 100)');
    }
};

3. 模型实现

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
<?php

namespace App\Models;

use Illuminate\Database\Eloquent\Model;
use Laravel\Ai\Embeddings;

class Document extends Model
{
    protected $fillable = ['title', 'content', 'embedding'];
    
    protected $casts = [
        'embedding' => 'vector',
    ];
    
    protected static function booted(): void
    {
        static::creating(function (self $doc) {
            if (empty($doc->embedding)) {
                $doc->embedding = Embeddings::from($doc->content)->generate()->vector;
            }
        });
    }
}

4. 搜索服务

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
<?php

namespace App\Services;

use App\Models\Document;
use Laravel\Ai\Embeddings;

class SemanticSearchService
{
    public function search(string $query, int $limit = 10)
    {
        $embedding = Embeddings::from($query)->generate()->vector;
        
        return Document::query()
            ->whereVectorSimilarTo('embedding', $embedding)
            ->withSimilarityScore()
            ->orderByDesc('similarity')
            ->limit($limit)
            ->get();
    }
}

5. 控制器

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
<?php

namespace App\Http\Controllers;

use App\Services\SemanticSearchService;

class SearchController extends Controller
{
    public function __construct(
        protected SemanticSearchService $search
    ) {}
    
    public function search(Request $request)
    {
        $validated = $request->validate([
            'query' => 'required|string|min:3',
            'limit' => 'nullable|integer|min:1|max:50',
        ]);
        
        $results = $this->search->search(
            $validated['query'],
            $validated['limit'] ?? 10
        );
        
        return response()->json([
            'query' => $validated['query'],
            'results' => $results,
        ]);
    }
}

6. 完整示例

6.1 文档导入

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
<?php

namespace App\Services;

use App\Models\Document;
use Laravel\Ai\Embeddings;

class DocumentImportService
{
    public function import(array $documents): void
    {
        $texts = array_column($documents, 'content');
        $embeddings = Embeddings::batch($texts)->generate();
        
        foreach ($documents as $index => $doc) {
            Document::create([
                'title' => $doc['title'],
                'content' => $doc['content'],
                'embedding' => $embeddings[$index]->vector,
            ]);
        }
    }
}

6.2 高级搜索

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
public function advancedSearch(string $query, array $filters = [])
{
    $embedding = Embeddings::from($query)->generate()->vector;
    
    $query = Document::query()
        ->whereVectorSimilarTo('embedding', $embedding)
        ->withSimilarityScore();
    
    if (!empty($filters['category'])) {
        $query->where('category', $filters['category']);
    }
    
    if (!empty($filters['date_from'])) {
        $query->where('created_at', '>=', $filters['date_from']);
    }
    
    return $query->orderByDesc('similarity')->get();
}

7. 性能优化

7.1 索引优化

1
2
3
4
-- HNSW 索引（更快查询）
CREATE INDEX documents_embedding_hnsw_idx ON documents 
USING hnsw (embedding vector_cosine_ops)
WITH (m = 16, ef_construction = 64);

7.2 缓存策略

1
2
3
4
5
6
7
8
public function search(string $query)
{
    $key = 'search:' . md5($query);
    
    return Cache::remember($key, 3600, function () use ($query) {
        // 执行搜索
    });
}

8. 总结

通过本实战案例，您已经掌握了：

  1. 向量搜索原理：语义相似性计算
  2. 数据库设计：pgvector 集成
  3. 嵌入生成：Laravel AI SDK
  4. 搜索实现：相似性查询
  5. 性能优化：索引与缓存

参考资料