PHP code example of blue-hex / laravel-docling-rag

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

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

    

blue-hex / laravel-docling-rag example snippets

use BlueHex\DoclingRag\Support\HasRagDocuments;
use Illuminate\Database\Eloquent\Model;

class DataSource extends Model
{
    use HasRagDocuments;
}

use BlueHex\DoclingRag\Facades\Rag;

$document = Rag::ingest($request->file('document'), owner: $dataSource);
// or
$document = $dataSource->ingestDocument($request->file('document'));

// config/docling-rag.php
'docling' => [
    'request_options' => [
        'convert_do_ocr' => true,
        'chunking_max_tokens' => 512,
        'chunking_merge_peers' => true,
        'chunking_use_markdown_tables' => true,
    ],
],

// e.g. skip OCR for a document you know is already text, or force it for a scan
$dataSource->ingestDocument($file, options: ['convert_do_ocr' => false]);

Rag::ingest($file, owner: $dataSource, options: [
    'convert_table_mode' => 'fast',
    'convert_page_range' => [1, 10],
]);

  public function search(Request $request, Project $project)
  {
      abort_if($project->user_id !== $request->user()->id, 404);

      return Rag::search($request->string('q'), owner: $project);
  }
  

use BlueHex\DoclingRag\Events\DocumentIngested;
use BlueHex\DoclingRag\Events\IngestionFailed;
use Illuminate\Support\Facades\Event;

Event::listen(function (DocumentIngested $event): void {
    $document = $event->document;                 // status === DocumentStatus::Ready
    $document->owner->notify(new DocumentReady($document));
});

Event::listen(function (IngestionFailed $event): void {
    report(new RuntimeException($event->document->failure_reason));
});

// app/Listeners/BroadcastIngestionStatus.php
public function handle(DocumentIngested $event): void
{
    IngestionStatusChanged::dispatch($event->document->id, 'ready');
}

use BlueHex\DoclingRag\Facades\Rag;

$results = Rag::search('What is the refund window?', owner: $dataSource);

foreach ($results as $chunk) {
    $chunk->text;        // the passage
    $chunk->pageNo;      // cite this
    $chunk->headingPath; // e.g. ['Policies', 'Refunds']
    $chunk->score;
}

use BlueHex\DoclingRag\Enums\ContentType;

$tables = Rag::search('quarterly revenue', $dataSource, filters: [
    'content_type' => ContentType::Table,
], k: 5);

use BlueHex\DoclingRag\Tools\SearchDocuments;

$agent->tools([
    SearchDocuments::for($dataSource),
    // optional: scope and cap it
    // SearchDocuments::for($dataSource, filters: ['content_type' => ContentType::Table], k: 5),
]);

use BlueHex\DoclingRag\Retrieval\ChunkResult;

$fake = Rag::fake([
    new ChunkResult(1, $doc->id, 'Refunds within 30 days.', pageNo: 4, headingPath: ['Refunds'], contentType: null, score: 0.9),
]);

// ... exercise your agent ...

expect($fake->searches)->toHaveCount(1);

bash
composer rag:install
php artisan migrate

bash
php artisan rag:install --with-gotenberg

bash
php artisan rag:index

bash
php artisan rag:health

bash
php artisan rag:reembed
php artisan rag:reembed 42

bash
php artisan rag:index