132 lines
4.0 KiB
PHP
132 lines
4.0 KiB
PHP
<?php
|
||
/**
|
||
* ACTIONS: Wikipedia Integration
|
||
* ------------------------------
|
||
* @wikiSummary <Title> → returns the lead/summary paragraph
|
||
* @wikiFullArticle <Title> → returns the full article as plain text
|
||
*
|
||
* Examples:
|
||
* @wikiSummary Finland
|
||
* @wikiFullArticle Bitcoin
|
||
*/
|
||
|
||
// 1) Register actions
|
||
registerAction(
|
||
(new Action())
|
||
->setFunctionName("wikiSummary")
|
||
->setArgumentsTypes(["string"]) // title
|
||
);
|
||
|
||
registerAction(
|
||
(new Action())
|
||
->setFunctionName("wikiFullArticle")
|
||
->setArgumentsTypes(["string"]) // title
|
||
);
|
||
|
||
// --- Helpers ---------------------------------------------------------------
|
||
|
||
/**
|
||
* Basic HTTP GET via cURL.
|
||
* @return array [int $httpCode, ?string $body, ?string $err]
|
||
*/
|
||
function http_get(string $url, int $timeout = 6): array {
|
||
$ch = curl_init($url);
|
||
curl_setopt_array($ch, [
|
||
CURLOPT_RETURNTRANSFER => true,
|
||
CURLOPT_FOLLOWLOCATION => true,
|
||
CURLOPT_TIMEOUT => $timeout,
|
||
CURLOPT_CONNECTTIMEOUT => 4,
|
||
CURLOPT_USERAGENT => "LLM-Action-Demo/1.0 (+https://example.com)"
|
||
]);
|
||
$body = curl_exec($ch);
|
||
$err = curl_error($ch) ?: null;
|
||
$code = curl_getinfo($ch, CURLINFO_HTTP_CODE);
|
||
curl_close($ch);
|
||
return [$code, $body, $err];
|
||
}
|
||
|
||
/** Safely pick the first (and only) page object from MediaWiki Action API */
|
||
function mw_first_page(array $json): ?array {
|
||
if (!isset($json['query']['pages']) || !is_array($json['query']['pages'])) return null;
|
||
foreach ($json['query']['pages'] as $page) {
|
||
return $page; // first element
|
||
}
|
||
return null;
|
||
}
|
||
|
||
// --- Actions ---------------------------------------------------------------
|
||
|
||
/**
|
||
* @param string $title Wikipedia article title
|
||
* @return string
|
||
*/
|
||
function wikiSummary(string $title): string {
|
||
$title = str_replace("-"," ",$title);
|
||
$encoded = rawurlencode($title);
|
||
$url = "https://en.wikipedia.org/api/rest_v1/page/summary/{$encoded}";
|
||
|
||
[$code, $body, $err] = http_get($url);
|
||
if ($err) return "Error fetching summary for '{$title}': {$err}";
|
||
if ($code < 200 || $code >= 300 || !$body) return "HTTP {$code}: Failed to fetch summary for '{$title}'.";
|
||
|
||
$data = json_decode($body, true);
|
||
if (isset($data['extract']) && is_string($data['extract']) && $data['extract'] !== '') {
|
||
return $data['extract'];
|
||
}
|
||
|
||
// Common “not found” or disambiguation handling
|
||
if (!empty($data['type']) && $data['type'] === 'disambiguation') {
|
||
return "‘{$title}’ is a disambiguation page. Try a more specific title.";
|
||
}
|
||
|
||
return "No summary found for '{$title}'.";
|
||
}
|
||
|
||
/**
|
||
* Returns full article as plain text (sections + paragraphs).
|
||
* Uses MediaWiki Action API with extracts (plaintext).
|
||
*
|
||
* @param string $title
|
||
* @return string
|
||
*/
|
||
function wikiFullArticle(string $title): string {
|
||
$title = str_replace("-"," ",$title);
|
||
$encoded = rawurlencode($title);
|
||
$url = "https://en.wikipedia.org/w/api.php"
|
||
. "?action=query"
|
||
. "&prop=extracts"
|
||
. "&explaintext=1"
|
||
. "&exsectionformat=plain"
|
||
. "&format=json"
|
||
. "&redirects=1"
|
||
. "&titles={$encoded}";
|
||
|
||
[$code, $body, $err] = http_get($url, 12);
|
||
if ($err) return "Error fetching article for '{$title}': {$err}";
|
||
if ($code < 200 || $code >= 300 || !$body) return "HTTP {$code}: Failed to fetch article for '{$title}'.";
|
||
|
||
$json = json_decode($body, true);
|
||
$page = mw_first_page($json);
|
||
|
||
if (!$page) {
|
||
return "No article found for '{$title}'.";
|
||
}
|
||
|
||
if (isset($page['missing'])) {
|
||
return "No article found for '{$title}'.";
|
||
}
|
||
|
||
if (!isset($page['extract']) || trim($page['extract']) === '') {
|
||
return "Article exists but has no plain-text extract for '{$title}'.";
|
||
}
|
||
|
||
// Optionally trim extremely long responses (LLM-friendly)
|
||
$maxChars = 40000; // adjust for your pipeline
|
||
$text = $page['extract'];
|
||
if (mb_strlen($text, 'UTF-8') > $maxChars) {
|
||
$text = mb_substr($text, 0, $maxChars, 'UTF-8') . "\n\n[Truncated]";
|
||
}
|
||
|
||
return $text;
|
||
}
|