index.php

<?php
header("Content-Type: application/json");
header("Access-Control-Allow-Origin: *"); // This allows all origins, to only allow requests from a specific
// domain remove this line and change the commented line below.
/// header("Access-Control-Allow-Origin: https://www.example.com");

header("Access-Control-Allow-Methods: GET, POST, OPTIONS");
if ($_SERVER['REQUEST_METHOD'] == 'OPTIONS') {
    http_response_code(200);
    exit();
}

require_once __DIR__ . "/bootstrap.php";
$ParallelRequest = new \App\ParallelRequest(10, 3);
$time_out = $_GET['time_out'] ?? $_POST['time_out'] ?? TIME_OUT;
$time_out = (int)$time_out; // Maximum time that a curl request to links extracted from Google search will wait
$ParallelRequest->setTimeout($time_out); // in seconds

$GoogleCSE = new \App\GoogleCSE();
$MainContentExtractor = new \App\MainContentExtractor();
$TextToChunk = new \App\TextToChunk();
$Cohere = new \App\Cohere();

$query = $_POST['query'] ?? $_GET['query'] ?? '';
$url = $_POST['url'] ?? $_GET['url'] ?? '';
$language = $_POST['language'] ?? $_GET['language'] ?? '';
$errors = [];

$max_results = $_GET['max_results'] ?? $_POST['max_results'] ?? MAX_RESULTS; // Google Search maximum results
$max_results = (int)$max_results;

$max_chunks = $_GET['max_chunks'] ?? $_POST['max_chunks'] ?? MAX_CHUNKS;
$max_chunks = (int)$max_chunks;
if ($max_results > $max_chunks) {
    $errors[] = ["max_results exceeds maximum number of chunks"];
}

$do_rerank = $_GET['do_rerank'] ?? $_POST['do_rerank'] ?? DO_RERANK;
$do_rerank = (bool)$do_rerank;

// $min_char The minimum number of characters a chunk must have
// $max_char The maximum number of characters a chunk must have - must be greater than the sum of ($min_char + $max_seq)
// $max_seq The maximum allowed length for words inside the chunk. Longer sequences then $max_seq  will be removed.
$min_char = $_GET['min_char'] ?? $_POST['min_char'] ?? MIN_CHAR;
$max_char = $_GET['max_char'] ?? $_POST['max_char'] ?? MAX_CHAR;
$max_seq = $_GET['max_seq'] ?? $_POST['max_seq'] ?? MAX_SEQ;

$min_char = (int)$min_char;
$max_char = (int)$max_char;
$max_seq = (int)$max_seq;


// maximum number of characters returned as output.
$max_characters_output = $_GET['max_chars_output'] ?? $_POST['max_chars_output'] ?? MAX_CHARACTERS_OUTPUT;
$max_characters_output = (int)$max_characters_output;

$all_data = [];
$all_links = [];
$is_url = false;
if ($query || ($url && filter_var($url, FILTER_VALIDATE_URL))) {
    if ($url) {
        $is_url = true;
        $do_rerank = false;
        // If query was actually a URL, no search will be performed, instead the contents of the URL will be extracted.
        $all_links[] = $url;
    } else {
        try {
            // make a search on Google and return just the links
            $all_links = $GoogleCSE->search($query, $max_results, 0, $language)->getItems(true);
            $arr_snippets = $GoogleCSE->getSnippets();
        } catch (Exception $e) {
            $errors['google_cse'] = $e->getMessage();
        }
    }

    foreach ($all_links as $link) {
        $ParallelRequest->addUrl($link);
    }
    $results = $ParallelRequest->request();

    $request_errors = $ParallelRequest->getError();
    if ($request_errors) {
        $errors[] = $request_errors;
    }

    $total_results = count($results);
    if ($total_results > 0) {
        $max_chunks_per_url = (int)($max_chunks / count($results));
        foreach ($results as $item) {
            $raw_html = $item->html;
            $url = $item->url;
            try {
                $readability = $MainContentExtractor->getMainContent($raw_html);
                $main_html = $readability->getContent();
                if (empty($main_html)) {
                    $main_html = ''; // null will cause errors
                }
                $chunks = $TextToChunk->makeChunks($main_html, $max_chunks_per_url, $min_char, $max_char, $max_seq);
                $all_data[] = (object)[
                    "url" => $url,
                    'chunks' => $chunks
                ];

            } catch (Exception $e) {
                $errors['readability'] = $e->getMessage() . " - url: $url";
            }
        } // end foreach
    } else {
        $do_rerank = false;
    }


    $reranks = [];
    $text = '';
    $urls_references = [];
    $idx = 1;
    $was_reranked = false;
    if ($all_data) {
        $all_chunks = [];
        foreach ($all_data as $item) {
            $url = $item->url;
            $urls_references[$idx] = $url;
            foreach ($item->chunks as $chunk) {
                $all_chunks[] = (object)["text" => $chunk, "url" => $url, "url_id" => $idx];
            }
            $idx++;
        }
        if ($do_rerank) {
            $reranks = $Cohere->rerank($query, $all_chunks)['result']->results ?? [];
        }
        $last_url = '';
        $text = '';
        $next_text = '';
        $paragraphs_ids = [];
        if ($reranks) {
            $was_reranked = true;
            foreach ($reranks as $item) {
                $url_id = $item->document->url_id;
                if (empty($paragraphs_ids[$url_id])) {
                    $paragraphs_ids[$url_id] = 1;
                }
                $p_id = $paragraphs_ids[$url_id];
                $next_text .= "<p>{$item->document->text} [{$item->document->url_id}][$p_id]</p>";
                if (mb_strlen($next_text) > $max_characters_output) {
                    break;
                } else {
                    $text = $next_text;
                }
                $paragraphs_ids[$url_id]++;
            }
        } else {
            // There was no success in reranking, so it will return chunks without reranking.
            $was_reranked = false;
            foreach ($all_chunks as $item) {
                $url_id = $item->url_id;
                if (empty($paragraphs_ids[$url_id])) {
                    $paragraphs_ids[$url_id] = 1;
                }
                $p_id = $paragraphs_ids[$url_id];
                $next_text .= "<p>{$item->text} [{$item->url_id}][$p_id]</p>";
                if (mb_strlen($next_text) > $max_characters_output) {
                    break;
                } else {
                    $text = $next_text;
                }
                $paragraphs_ids[$url_id]++;
            }

        }
    }

    $data = [
        'text' => $text,
        'urls_references' => $urls_references,
        'reranked' => $was_reranked
    ];

    if (count($errors) >= 1) {
        $data['errors'] = $errors;
    }
} else {
    $data = ["text" => "", "msg" => 'No search query were passed'];
    if (!empty($url)) {
        $data = ["text" => "", "msg" => "Bad URL: $url"];
    }
}

if (!empty($arr_snippets)) {
    $data['snippets'] = $arr_snippets;
}
$json = json_encode($data);
if(!$json){
    if(!empty($data['text'])){
        $data['text'] = mb_convert_encoding($data['text'], 'UTF-8');
    }
}
echo json_encode($data);