Files
myep/src/Chatbot/Mcp/Tool/GetDocumentTool.php
T

57 lines
1.9 KiB
PHP

<?php
declare(strict_types=1);
namespace App\Chatbot\Mcp\Tool;
use App\Chatbot\Mcp\ToolCallRunner;
use App\Chatbot\Search\ContentSearch;
use App\Repository\ContentDocumentRepository;
use Mcp\Capability\Attribute\McpTool;
use Mcp\Capability\Attribute\Schema;
use Mcp\Exception\ToolCallException;
use Mcp\Schema\Result\CallToolResult;
use Mcp\Schema\ToolAnnotations;
#[McpTool(
name: 'get_document',
description: 'Read the full text of one search_content hit, by its url. Editorial text can be out of date: take dates, prices, pickups and inclusions from search_trips and get_trip instead.',
annotations: new ToolAnnotations(readOnlyHint: true, openWorldHint: false),
)]
final class GetDocumentTool
{
private const BODY_LIMIT = 20000;
public function __construct(
private readonly ContentDocumentRepository $repository,
private readonly ToolCallRunner $runner,
) {
}
public function __invoke(
#[Schema(description: 'A url returned by search_content.', minLength: 1)]
string $url,
): CallToolResult {
return $this->runner->run('get_document', get_defined_vars(), function () use ($url): array {
$document = $this->repository->findOneIndexedByUrl(trim($url));
if (null === $document) {
throw new ToolCallException('No document at this url. Use a url returned by search_content.');
}
$body = $document->getBodyMd();
$truncated = mb_strlen($body) > self::BODY_LIMIT;
return [
'type' => ContentSearch::publicType($document->getType()),
'title' => $document->getTitle(),
'url' => $document->getUrl(),
'site' => $document->getSite(),
'breadcrumb' => $document->getBreadcrumb(),
'body' => $truncated ? mb_substr($body, 0, self::BODY_LIMIT) : $body,
'truncated' => $truncated,
];
});
}
}