AI 项目在线编辑器

<?php
declare(strict_types=1);

function http_get_text(string $url, int $timeoutSec = 10, array $headers = []): string
{
  $ch = curl_init($url);
  if ($ch === false) throw new RuntimeException('curl_init failed');

$hdr = array_merge(['Accept: application/atom+xml, application/xml, text/xml'], $headers);

curl_setopt_array($ch, [
    CURLOPT_RETURNTRANSFER => true,
    CURLOPT_CONNECTTIMEOUT => $timeoutSec,
    CURLOPT_TIMEOUT => $timeoutSec,
    CURLOPT_HTTPHEADER => $hdr,
  ]);

$resp = curl_exec($ch);
  $errno = curl_errno($ch);
  $err = curl_error($ch);
  $status = (int)curl_getinfo($ch, CURLINFO_HTTP_CODE);
  curl_close($ch);

if ($errno !== 0) throw new RuntimeException('cURL error: ' . $err);
  if ($status < 200 || $status >= 300) throw new RuntimeException('HTTP ' . $status);
  return (string)$resp;
}

return [
  'name' => 'search_arxiv',
  'description' => '在 arXiv 检索论文（Atom API，返回标题/作者/日期/abs链接/pdf链接/摘要）。',
  'parameters' => [
    'type' => 'object',
    'properties' => [
      'query' => ['type' => 'string', 'description' => '检索关键词，例如：diffusion model'],
      'max_results' => ['type' => 'integer', 'description' => '返回条数，默认 5', 'default' => 5],
    ],
    'required' => ['query'],
  ],
  'run' => function(array $args, array $context) {
    $query = trim((string)($args['query'] ?? ''));
    $max = (int)($args['max_results'] ?? 5);
    if ($max <= 0) $max = 5;
    if ($query === '') throw new RuntimeException('query 不能为空');

// arXiv Atom API
    // search_query=all:<query>
    $apiUrl = 'https://export.arxiv.org/api/query?' . http_build_query([
      'search_query' => 'all:' . $query,
      'start' => 0,
      'max_results' => $max,
      'sortBy' => 'relevance',
      'sortOrder' => 'descending',
    ]);

try {
      $xmlText = http_get_text($apiUrl, 15);
    } catch (Throwable $e) {
      return "An error occurred while searching arXiv: " . $e->getMessage();
    }

libxml_use_internal_errors(true);
    $feed = simplexml_load_string($xmlText);
    if ($feed === false) {
      return "An error occurred while searching arXiv: Invalid XML response.";
    }

// Atom namespace
    $feed->registerXPathNamespace('a', 'http://www.w3.org/2005/Atom');

$entries = $feed->xpath('//a:entry');
    if (!is_array($entries) || count($entries) === 0) {
      return "No results found on arXiv for '{$query}'.";
    }

$resultsList = [];
    $citationsList = [];

$emitter = $context['event_emitter'] ?? null;
    $canEmit = is_callable($emitter);

$i = 0;
    foreach ($entries as $entry) {
      $i++;

// title / summary 里经常有换行和多空格
      $title = trim(preg_replace('/\s+/', ' ', (string)$entry->title));
      if ($title === '') $title = 'N/A';

$published = (string)$entry->published;
      $publishedDate = 'N/A';
      if ($published) {
        // 取 YYYY-MM-DD
        $publishedDate = substr($published, 0, 10) ?: 'N/A';
      }

$summary = trim(preg_replace('/\s+/', ' ', (string)$entry->summary));
      if ($summary === '') $summary = 'N/A';

// authors
      $authors = [];
      if (isset($entry->author)) {
        foreach ($entry->author as $a) {
          $nm = trim((string)$a->name);
          if ($nm !== '') $authors[] = $nm;
        }
      }
      $authorsStr = count($authors) ? implode(', ', $authors) : 'N/A';

// abs url (entry->id)
      $absUrl = trim((string)$entry->id);
      if ($absUrl === '') $absUrl = 'N/A';

// pdf url: 在 link 里找 title="pdf" 或 type="application/pdf"
      $pdfUrl = '';
      if (isset($entry->link)) {
        foreach ($entry->link as $lnk) {
          $attrs = $lnk->attributes();
          $href = (string)($attrs['href'] ?? '');
          $type = (string)($attrs['type'] ?? '');
          $titleAttr = (string)($attrs['title'] ?? '');
          if ($href !== '' && (strtolower($titleAttr) === 'pdf' || strtolower($type) === 'application/pdf')) {
            $pdfUrl = $href;
            break;
          }
        }
      }

$resultsList[] =
        "{$i}. {$title}\n" .
        "   Authors: {$authorsStr}\n" .
        "   Published: {$publishedDate}\n" .
        "   URL: {$absUrl}\n" .
        "   PDF URL: " . ($pdfUrl !== '' ? $pdfUrl : 'N/A') . "\n" .
        "   Summary: {$summary}";

if ($pdfUrl !== '') {
        $citationsList[] = "[{$i}] {$pdfUrl}";
        if ($canEmit) {
          $emitter([
            'type' => 'citation',
            'data' => [
              'document' => [$summary],
              'metadata' => [['source' => $pdfUrl]],
              'source' => ['name' => $title],
            ],
          ]);
        }
      }

if ($i >= $max) break;
    }

$output = implode("\n\n", $resultsList);
    if (count($citationsList)) {
      $output .= "\n\n---\n\n**Citation Links:**\n" . implode("\n", $citationsList);
    }
    return $output;
  },
];

Models

Tools