PHP code example of boehmmatthias / smartsearch

    
        
<?php
require_once('vendor/autoload.php');

/* Start to develop here. Best regards https://php-download.com/ */

    

boehmmatthias / smartsearch example snippets

use BoehmMatthias\SmartSearch\Service\VectorService;

class MyEventListener
{
    public function __construct(
        private readonly VectorService $vectorService,
    ) {}

    public function afterSave(MyRecord $record): void
    {
        $this->vectorService->embedAndStore(
            collection: 'my-extension-articles',
            identifier: $record->getUid(),
            text: $record->getTitle() . "\n\n" . strip_tags($record->getBodyText()),
        );
    }
}

$hits = $this->vectorService->findSimilar(
    collection: 'my-extension-articles',
    query: 'how do I configure caching?',
    topK: 5,
);

// Returns: [['identifier' => '42', 'score' => 0.87], ['identifier' => '7', 'score' => 0.74], ...]
// Sorted by cosine similarity descending. 'identifier' is always a string.
foreach ($hits as $hit) {
    $record = $this->recordRepository->findByUid((int)$hit['identifier']);
    // filter by threshold if needed: if ($hit['score'] < 0.30) continue;
}

use BoehmMatthias\SmartSearch\Service\GenerationService;
use BoehmMatthias\SmartSearch\Service\VectorService;
use BoehmMatthias\SmartSearch\Configuration\SmartSearchConfiguration;

class SearchController
{
    public function __construct(
        private readonly VectorService $vectorService,
        private readonly GenerationService $generationService,
        private readonly SmartSearchConfiguration $configuration,
    ) {}

    public function answerAction(string $question): string
    {
        // 1. Find the most relevant documents
        $hits = $this->vectorService->findSimilar(
            collection: 'my-extension-articles',
            query: $question,
            topK: $this->configuration->getRagTopK(),
        );

        // 2. Filter by semantic threshold
        $threshold = $this->configuration->getSemanticThreshold();
        $hits = array_filter($hits, fn($h) => $h['score'] >= $threshold);

        if (empty($hits)) {
            return 'No relevant documents found.';
        }

        // 3. Build context blocks — one string per source document
        $maxChars = $this->configuration->getDocumentContextLength();
        $contextBlocks = [];
        foreach ($hits as $hit) {
            $record = $this->recordRepository->findByUid((int)$hit['identifier']);
            $excerpt = mb_substr(strip_tags($record->getBodyText()), 0, $maxChars);
            $contextBlocks[] = sprintf('[%d] %s\n%s', $record->getUid(), $record->getTitle(), $excerpt);
        }

        // 4. Generate a grounded answer
        return $this->generationService->generate(
            query: $question,
            contextBlocks: $contextBlocks,
        );
        // The system prompt instructs the model to answer only from the provided
        // documents and to cite sources by their identifier (e.g. [42]).
    }
}

use BoehmMatthias\SmartSearch\Repository\VectorRepository;

// Remove a single entry
$this->vectorRepository->deleteByIdentifier('my-extension-articles', (string)$uid);

// Remove all entries in a collection (e.g. before a full reindex)
$this->vectorRepository->deleteByCollection('my-extension-articles');

use BoehmMatthias\SmartSearch\Service\ModelAvailabilityService;

if ($this->modelAvailabilityService->isEmbeddingServerAvailable()) {
    // offer semantic search
}

if ($this->modelAvailabilityService->isGenerationServerAvailable()) {
    // offer RAG answers
}

namespace MyVendor\MyExtension\Embedding;

use BoehmMatthias\SmartSearch\Embedding\EmbeddingClientInterface;

final class OpenAiEmbeddingClient implements EmbeddingClientInterface
{
    public function __construct(
        private readonly string $apiKey,
        private readonly string $model = 'text-embedding-3-small',
    ) {}

    /** @return float[] */
    public function embed(string $text): array
    {
        // Call the OpenAI embeddings API and return the float array.
        // ...
    }
}

$vectorRepository->deleteByCollection('your-collection');
// then re-call embedAndStore() for all records