feat: adding youtube transcription tool

README.md

@@ -54,6 +54,7 @@ Provided Tools
 * [x] Clock
 * [x] Wikipedia
 * [x] Weather
+* [x] YouTube Transcriber
 
 Usage Examples
 --------------
@@ -105,6 +106,18 @@ Depending on the example you need to export needed environment variables for API
    php examples/toolchain-serpapi.php
    ```
 
+1. Weather Tool
+   ```bash
+   export OPENAI_API_KEY=sk-...
+   php examples/toolchain-weather.php
+   ```
+
+1. YouTube Transcriber Tool
+   ```bash
+   export OPENAI_API_KEY=sk-...
+   php examples/toolchain-youtube.php
+   ```
+
 ### Structured Output
 
 1. Structured Output Example: OpenAI's GPT

composer.json

@@ -29,12 +29,16 @@
         "probots-io/pinecone-php": "^1.0",
         "symfony/clock": "^6.4 || ^7.1",
         "symfony/console": "^6.4 || ^7.1",
+        "symfony/css-selector": "^6.4 || ^7.1",
+        "symfony/dom-crawler": "^6.4 || ^7.1",
         "symfony/var-dumper": "^6.4 || ^7.1"
     },
     "suggest": {
         "codewithkyrian/chromadb-php": "For using the ChromaDB as retrieval vector store.",
         "probots-io/pinecone-php": "For using the Pinecone as retrieval vector store.",
-        "symfony/clock": "For using the clock tool."
+        "symfony/clock": "For using the clock tool.",
+        "symfony/css-selector": "For using the YouTube transcription tool.",
+        "symfony/dom-crawler": "For using the YouTube transcription tool."
     },
     "config": {
         "sort-packages": true

examples/toolchain-youtube.php

@@ -0,0 +1,28 @@
+<?php
+
+use PhpLlm\LlmChain\Chain;
+use PhpLlm\LlmChain\Message\Message;
+use PhpLlm\LlmChain\Message\MessageBag;
+use PhpLlm\LlmChain\OpenAI\Model\Gpt;
+use PhpLlm\LlmChain\OpenAI\Model\Gpt\Version;
+use PhpLlm\LlmChain\OpenAI\Runtime\OpenAI;
+use PhpLlm\LlmChain\ToolBox\ParameterAnalyzer;
+use PhpLlm\LlmChain\ToolBox\Registry;
+use PhpLlm\LlmChain\ToolBox\Tool\YouTubeTranscriber;
+use PhpLlm\LlmChain\ToolBox\ToolAnalyzer;
+use Symfony\Component\HttpClient\HttpClient;
+
+require_once dirname(__DIR__).'/vendor/autoload.php';
+
+$httpClient = HttpClient::create();
+$runtime = new OpenAI($httpClient, getenv('OPENAI_API_KEY'));
+$llm = new Gpt($runtime, Version::GPT_4o_MINI);
+
+$transcriber = new YouTubeTranscriber($httpClient);
+$registry = new Registry(new ToolAnalyzer(new ParameterAnalyzer()), [$transcriber]);
+$chain = new Chain($llm, $registry);
+
+$messages = new MessageBag(Message::ofUser('Please summarize this video for me: https://www.youtube.com/watch?v=6uXW-ulpj0s'));
+$response = $chain->call($messages);
+
+echo $response.PHP_EOL;

src/ToolBox/Tool/YouTubeTranscriber.php

@@ -0,0 +1,67 @@
+<?php
+
+declare(strict_types=1);
+
+namespace PhpLlm\LlmChain\ToolBox\Tool;
+
+use PhpLlm\LlmChain\ToolBox\AsTool;
+use Symfony\Component\CssSelector\CssSelectorConverter;
+use Symfony\Component\DomCrawler\Crawler;
+use Symfony\Contracts\HttpClient\HttpClientInterface;
+
+#[AsTool('youtube_transcript', 'Fetches the transcript of a YouTube video')]
+final readonly class YouTubeTranscriber
+{
+    public function __construct(
+        private HttpClientInterface $client,
+    ) {
+        if (!class_exists(Crawler::class)) {
+            throw new \LogicException('The Symfony DomCrawler component is required to use this tool.');
+        }
+        if (!class_exists(CssSelectorConverter::class)) {
+            throw new \LogicException('The Symfony CSS Selector component is required to use this tool.');
+        }
+    }
+
+    /**
+     * @param string $videoId The ID of the YouTube video
+     */
+    public function __invoke(string $videoId): string
+    {
+        // Fetch the HTML content of the YouTube video page
+        $htmlResponse = $this->client->request('GET', 'https://youtube.com/watch?v='.$videoId);
+        $html = $htmlResponse->getContent();
+
+        // Use DomCrawler to parse the HTML
+        $crawler = new Crawler($html);
+
+        // Extract the script containing the ytInitialPlayerResponse
+        $scriptContent = $crawler->filter('script')->reduce(function (Crawler $node) {
+            return str_contains($node->text(), 'var ytInitialPlayerResponse = {');
+        })->text();
+
+        // Extract and parse the JSON data from the script
+        $start = strpos($scriptContent, 'var ytInitialPlayerResponse = ') + strlen('var ytInitialPlayerResponse = ');
+        $dataString = substr($scriptContent, $start);
+        $dataString = substr($dataString, 0, strrpos($dataString, ';') ?: null);
+        $data = json_decode(trim($dataString), true);
+
+        // Extract the URL for the captions
+        if (!isset($data['captions']['playerCaptionsTracklistRenderer']['captionTracks'][0]['baseUrl'])) {
+            throw new \Exception('Captions are not available for this video.');
+        }
+        $captionsUrl = $data['captions']['playerCaptionsTracklistRenderer']['captionTracks'][0]['baseUrl'];
+
+        // Fetch and parse the captions XML
+        $xmlResponse = $this->client->request('GET', $captionsUrl);
+        $xmlContent = $xmlResponse->getContent();
+        $xmlCrawler = new Crawler($xmlContent);
+
+        // Collect all text elements from the captions
+        $transcript = $xmlCrawler->filter('text')->each(function (Crawler $node) {
+            return $node->text().' ';
+        });
+
+        return implode(PHP_EOL, $transcript);
+    }
+}