已经完成一个文章校对了，但换个文章id就报错了，排查前备份

application/common/ReferenceCheckService.php

@@ -3,6 +3,7 @@
 namespace app\common;
 
 use think\Db;
+use think\Env;
 use think\Queue;
 
 /**
@@ -131,8 +132,39 @@ class ReferenceCheckService
 
         $this->setAmRefCheckStatus($amId, self::AM_STATUS_RUNNING);
     }
-    public function checkOne(){
-        $this->pushJob(intval(724), 0);
+    /**
+     * 手工触发：对已完成且 confidence<=0.65 的记录入队 DOI 第二轮复核
+     */
+    public function enqueueSecondPassByArticle($articleId)
+    {
+        $articleId = intval($articleId);
+        if ($articleId <= 0) {
+            throw new \InvalidArgumentException('article_id is required');
+        }
+
+        $rows = Db::name('article_reference_check_result')
+            ->where('article_id', $articleId)
+            ->where('status', 1)
+            ->where('confidence', '<=', 0.65)
+            ->orderRaw('rand()')
+            ->limit(2)
+            ->select();
+
+        $checkIds2 = [];
+        $delay2 = 0;
+        foreach ($rows as $checkLog) {
+            $rowId = $this->resolveCheckRowId($checkLog);
+            if ($this->maybeEnqueueSecondPass($rowId, floatval($checkLog['confidence']))) {
+                $checkIds2[] = $rowId;
+                $delay2 += 1;
+            }
+        }
+
+        return [
+            'article_id'  => $articleId,
+            'check_ids2'  => $checkIds2,
+            'queued'      => count($checkIds2),
+        ];
     }
     public function enqueueByArticle($articleId){
         if ($articleId <= 0) {
@@ -140,7 +172,7 @@ class ReferenceCheckService
         }
         $prod = Db::name('production_article')
             ->where('article_id', $articleId)
-            ->where('state', 0)
+            ->where('state', [0, 2])
             ->find();
         if (empty($prod)) {
             throw new \RuntimeException('production_article not found for article_id=' . $articleId);
@@ -296,12 +328,78 @@ class ReferenceCheckService
         return isset($map[$status]) ? $map[$status] : 'unknown';
     }
 
+    /**
+     * 表主键为 id（对外 API 参数名仍叫 check_id）
+     */
+    public function resolveCheckRowId($row)
+    {
+        if (!is_array($row)) {
+            return 0;
+        }
+        if (isset($row['id']) && intval($row['id']) > 0) {
+            return intval($row['id']);
+        }
+        if (isset($row['check_id']) && intval($row['check_id']) > 0) {
+            return intval($row['check_id']);
+        }
+        return 0;
+    }
+
+    /**
+     * 解析 LLM 返回的 is_match（兼容 bool / 0|1 / "true"|"false" 字符串）
+     */
+    public function parseLlmIsMatch($value)
+    {
+        if (is_bool($value)) {
+            return $value;
+        }
+        if (is_int($value) || is_float($value)) {
+            return intval($value) === 1;
+        }
+        $s = strtolower(trim((string)$value));
+        return in_array($s, ['1', 'true', 'yes', 'match', 'matched'], true);
+    }
+
+    /**
+     * 写入单条校对结果（统一截断 reason/error_msg，避免 varchar(512) 导致 UPDATE 失败）
+     *
+     * @throws \RuntimeException
+     */
+    public function updateCheckResult($checkId, array $fields)
+    {
+        $checkId = intval($checkId);
+        if ($checkId <= 0) {
+            throw new \InvalidArgumentException('invalid check id');
+        }
+
+        if (isset($fields['reason'])) {
+            $fields['reason'] = mb_substr(trim((string)$fields['reason']), 0, 512);
+        }
+        if (isset($fields['error_msg'])) {
+            $fields['error_msg'] = mb_substr(trim((string)$fields['error_msg']), 0, 512);
+        }
+        $fields['updated_at'] = date('Y-m-d H:i:s');
+
+        $exists = Db::name('article_reference_check_result')->where('id', $checkId)->find();
+        if (empty($exists)) {
+            throw new \RuntimeException('article_reference_check_result not found, id=' . $checkId);
+        }
+
+        $affected = Db::name('article_reference_check_result')->where('id', $checkId)->update($fields);
+        if ($affected === false) {
+            throw new \RuntimeException('article_reference_check_result update failed, id=' . $checkId);
+        }
+
+        \think\Log::info('updateCheckResult id=' . $checkId . ' affected=' . intval($affected));
+        return intval($affected);
+    }
+
     public function getResult($checkId)
     {
         if ($checkId <= 0) {
             return null;
         }
-        $row = Db::name('article_reference_check_result')->where('check_id', $checkId)->find();
+        $row = Db::name('article_reference_check_result')->where('id', $checkId)->find();
         return $row ?: null;
     }
 
@@ -435,7 +533,7 @@ class ReferenceCheckService
                     'ref_nos'    => [],
                 ];
             }
-            $byAm[$amId]['contexts'][$ctxKey]['check_ids'][] = intval($row['check_id']);
+            $byAm[$amId]['contexts'][$ctxKey]['check_ids'][] = $this->resolveCheckRowId($row);
             $byAm[$amId]['contexts'][$ctxKey]['ref_nos'][] = $refNo;
             $reason = trim((string)$this->arrGet($row, 'reason', ''));
             if ($reason !== '') {
@@ -501,7 +599,7 @@ class ReferenceCheckService
                         $issueCount++;
                         $issues[] = array(
                             'am_id'         => $amId,
-                            'check_id'      => intval($row['check_id']),
+                            'check_id'      => $this->resolveCheckRowId($row),
                             'reference_no'  => $num,
                             'reference_raw' => $inner,
                             'reason'        => $rowReason,
@@ -512,7 +610,7 @@ class ReferenceCheckService
                             ENT_QUOTES,
                             'UTF-8'
                         );
-                        return '<span class="ref-no-error" data-check-id="' . intval($row['check_id'])
+                        return '<span class="ref-no-error" data-check-id="' . $this->resolveCheckRowId($row)
                             . '" data-ref-no="' . $num . '" title="' . $title . '">'
                             . $numMatch[0] . '</span>';
                     },
@@ -627,6 +725,448 @@ class ReferenceCheckService
         return implode("\n", $parts);
     }
 
+    /**
+     * 仅使用 refer_doi 字段（二次 Crossref 摘要用）
+     */
+    public function extractReferDoiOnly($refer)
+    {
+        if (!is_array($refer)) {
+            return '';
+        }
+        $raw = trim((string)$this->arrGet($refer, 'refer_doi', ''));
+        if ($raw === '' || stripos($raw, 'not available') !== false) {
+            return '';
+        }
+        $dois = $this->extractDoisFromString($raw);
+        return empty($dois) ? '' : $dois[0];
+    }
+
+    /**
+     * 根据 refer_doi 调用 Crossref works API 获取摘要（二次校对专用）
+     *
+     * @return array{text:string, has_abstract:bool, doi:string}
+     */
+    public function fetchCrossrefAbstractByReferDoi($refer)
+    {
+        $doi = $this->extractReferDoiOnly($refer);
+        if ($doi === '') {
+            return ['text' => '', 'has_abstract' => false, 'doi' => ''];
+        }
+
+        $crossref = new CrossrefService([
+            'mailto' => trim((string)Env::get('crossref_mailto', '')),
+        ]);
+        $block = $this->extractCrossrefBlock($doi, $crossref);
+        if ($block === null) {
+            return ['text' => '', 'has_abstract' => false, 'doi' => $doi];
+        }
+
+        return [
+            'text'         => $block['text'],
+            'has_abstract' => !empty($block['has_abstract']),
+            'doi'          => $doi,
+        ];
+    }
+
+    /**
+     * 解析 LLM 返回的 can_support
+     */
+    public function parseLlmCanSupport($llmResult)
+    {
+        if (!is_array($llmResult)) {
+            return false;
+        }
+        if (array_key_exists('can_support', $llmResult)) {
+            return $this->parseLlmIsMatch($llmResult['can_support']);
+        }
+        return $this->parseLlmIsMatch(isset($llmResult['is_match']) ? $llmResult['is_match'] : false);
+    }
+
+    /**
+     * 第一次校对：取 article_main.content（整节正文）
+     */
+    public function resolveMainContentForJob(array $row, $maxChars = 8000)
+    {
+        $amId = intval($this->arrGet($row, 'am_id', 0));
+        if ($amId <= 0) {
+            return '';
+        }
+        $main = Db::name('article_main')
+            ->field('content')
+            ->where('am_id', $amId)
+            ->find();
+        if (empty($main)) {
+            return '';
+        }
+
+        $text = trim((string)$this->arrGet($main, 'content', ''));
+        if ($text === '') {
+            return '';
+        }
+
+        $text = preg_replace('/<blue>\[([\d,\-\s]+)\]<\/blue>/', '[$1]', $text);
+        $text = strip_tags($text);
+        $text = html_entity_decode($text, ENT_QUOTES | ENT_HTML5, 'UTF-8');
+        $text = preg_replace('/\s+/u', ' ', $text);
+        $text = trim($text);
+
+        $maxChars = max(500, intval($maxChars));
+        if (mb_strlen($text) > $maxChars) {
+            $text = mb_substr($text, 0, $maxChars) . '...';
+        }
+
+        return $text;
+    }
+
+    /**
+     * 引用处局部上下文（origin_text），供其它场景使用
+     */
+    public function resolveCitationContextForJob(array $row)
+    {
+        $text = trim((string)$this->arrGet($row, 'origin_text', ''));
+        if ($text === '') {
+            $text = trim((string)$this->arrGet($row, 'content_a', ''));
+        }
+        return $text;
+    }
+
+    /**
+     * 从 refer 行提取标准 DOI（10.xxxx/...）
+     *
+     * 优先级：refer_content（原始引用文本里的 DOI 最贴近实际被引用的文献）
+     *        > refer_doi > doi > doilink
+     */
+    public function extractDoiFromRefer($refer)
+    {
+        $list = $this->extractAllDoiCandidatesFromRefer($refer);
+        return empty($list) ? '' : $list[0];
+    }
+
+    /**
+     * 返回 refer 行可能对应的全部 DOI 候选（去重，按优先级排序）
+     *
+     * 用于第二轮 DOI 复核场景：当 metadata 的 refer_doi 与原始引用文本里的 DOI
+     * 不一致时（数据漂移），优先尝试原始引用文本里的 DOI 抓真实摘要。
+     *
+     * @return string[]
+     */
+    public function extractAllDoiCandidatesFromRefer($refer)
+    {
+        if (!is_array($refer)) {
+            return [];
+        }
+        $ordered = [
+            (string)$this->arrGet($refer, 'refer_content', ''),
+            (string)$this->arrGet($refer, 'refer_doi', ''),
+            (string)$this->arrGet($refer, 'doi', ''),
+            (string)$this->arrGet($refer, 'doilink', ''),
+        ];
+
+        $result = [];
+        foreach ($ordered as $raw) {
+            foreach ($this->extractDoisFromString($raw) as $doi) {
+                if (!in_array($doi, $result, true)) {
+                    $result[] = $doi;
+                }
+            }
+        }
+        return $result;
+    }
+
+    /**
+     * 从任意文本里抽取所有形如 10.xxxx/yyy 的 DOI
+     * @return string[]
+     */
+    private function extractDoisFromString($text)
+    {
+        $text = trim((string)$text);
+        if ($text === '' || stripos($text, 'not available') !== false) {
+            return [];
+        }
+
+        $dois = [];
+
+        if (preg_match_all('~doi\.org/([^\s?#"\'<>]+)~i', $text, $m)) {
+            foreach ($m[1] as $cand) {
+                $cand = $this->trimDoiTail(trim($cand));
+                if ($this->isValidDoi($cand)) {
+                    $dois[] = $cand;
+                }
+            }
+        }
+
+        if (preg_match_all('~\b(10\.\d{3,9}/[^\s?#"\'<>]+)~i', $text, $m)) {
+            foreach ($m[1] as $cand) {
+                $cand = $this->trimDoiTail(trim($cand));
+                if ($this->isValidDoi($cand)) {
+                    $dois[] = $cand;
+                }
+            }
+        }
+
+        if ($dois === [] && strpos($text, '10.') === 0) {
+            $cand = $this->trimDoiTail($text);
+            if ($this->isValidDoi($cand)) {
+                $dois[] = $cand;
+            }
+        }
+
+        return array_values(array_unique($dois));
+    }
+
+    private function trimDoiTail($doi)
+    {
+        return rtrim($doi, ".,;:)]}>\"'\\ \t\n\r");
+    }
+
+    private function isValidDoi($doi)
+    {
+        return (bool)preg_match('~^10\.\d{3,9}/[^\s]+$~i', (string)$doi);
+    }
+
+    /**
+     * 通过 PubMed / Crossref 拉取 DOI 对应文献内容（本地 LLM 无法打开网页，须预先抓取）
+     *
+     * 行为：
+     * - 尝试 refer 行内所有 DOI 候选（refer_content > refer_doi > doi > doilink）
+     * - 优先采用第一个能拿到 abstract 的 DOI
+     * - PubMed 无摘要时回落到 Crossref raw 解析摘要（清理 JATS 标签）
+     * - 全部失败则返回空字符串（调用方据此跳过二次复核）
+     */
+    public function fetchDoiLiteratureBlock($refer)
+    {
+        $candidates = $this->extractAllDoiCandidatesFromRefer($refer);
+        if (empty($candidates)) {
+            return '';
+        }
+
+        $pubmed = new PubmedService([
+            'email' => trim((string)Env::get('pubmed_email', '')),
+            'tool'  => trim((string)Env::get('pubmed_tool', 'tmrjournals')),
+        ]);
+        $crossref = new CrossrefService([
+            'mailto' => trim((string)Env::get('crossref_mailto', '')),
+        ]);
+
+        $best = null;
+        $fallback = null;
+
+        foreach ($candidates as $doi) {
+            $block = $this->buildDoiBlockFromSources($doi, $pubmed, $crossref);
+            if ($block === null) {
+                continue;
+            }
+            if (!empty($block['has_abstract'])) {
+                $best = $block;
+                break;
+            }
+            if ($fallback === null) {
+                $fallback = $block;
+            }
+        }
+
+        $chosen = $best ?: $fallback;
+        if ($chosen === null) {
+            return '';
+        }
+        return $chosen['text'];
+    }
+
+    /**
+     * 拉单个 DOI 的真实内容，返回 ['text' => string, 'has_abstract' => bool] 或 null
+     */
+    private function buildDoiBlockFromSources($doi, PubmedService $pubmed, CrossrefService $crossref)
+    {
+        $doi = trim((string)$doi);
+        if ($doi === '') {
+            return null;
+        }
+
+        $pub = $pubmed->fetchByDoi($doi);
+        $pubAbstract = is_array($pub) ? trim((string)$this->arrGet($pub, 'abstract', '')) : '';
+
+        if (is_array($pub) && ($pubAbstract !== '' || trim((string)$this->arrGet($pub, 'title', '')) !== '')) {
+            $lines = ['Source: PubMed (DOI ' . $doi . ')'];
+            if (!empty($pub['title'])) {
+                $lines[] = 'Actual Title: ' . trim((string)$pub['title']);
+            }
+            if (!empty($pub['journal'])) {
+                $lines[] = 'Journal: ' . trim((string)$pub['journal']);
+            }
+            if (!empty($pub['year'])) {
+                $lines[] = 'Year: ' . trim((string)$pub['year']);
+            }
+            if (!empty($pub['publication_types'])) {
+                $lines[] = 'Publication Types: ' . implode('; ', (array)$pub['publication_types']);
+            }
+            if (!empty($pub['mesh_terms'])) {
+                $lines[] = 'MeSH: ' . implode('; ', (array)$pub['mesh_terms']);
+            }
+            if ($pubAbstract !== '') {
+                $lines[] = 'Abstract: ' . $this->truncate($pubAbstract, 3500);
+            }
+
+            if ($pubAbstract === '') {
+                $cr = $this->extractCrossrefBlock($doi, $crossref);
+                if ($cr !== null && $cr['has_abstract']) {
+                    $lines[] = "\n--- Crossref 补充 ---\n" . $cr['text'];
+                    return ['text' => implode("\n", $lines), 'has_abstract' => true];
+                }
+            }
+
+            return ['text' => implode("\n", $lines), 'has_abstract' => $pubAbstract !== ''];
+        }
+
+        return $this->extractCrossrefBlock($doi, $crossref);
+    }
+
+    /**
+     * 从 Crossref 拉取标题/期刊/作者/摘要（abstract 通常包裹 JATS XML，需清洗）
+     * @return array|null ['text' => string, 'has_abstract' => bool]
+     */
+    private function extractCrossrefBlock($doi, CrossrefService $crossref)
+    {
+        $msg = $crossref->fetchWork($doi);
+        if (!is_array($msg)) {
+            return null;
+        }
+
+        $summary = $crossref->fetchWorkSummary($doi);
+        if (!is_array($summary)) {
+            $summary = [];
+        }
+
+        $lines = ['Source: Crossref api.crossref.org/works/' . rawurlencode($doi)];
+        $title = isset($msg['title'][0]) ? trim((string)$msg['title'][0]) : trim((string)$this->arrGet($summary, 'title', ''));
+        if ($title !== '') {
+            $lines[] = 'Actual Title: ' . $title;
+        }
+        if (!empty($summary['joura'])) {
+            $lines[] = 'Journal: ' . trim((string)$summary['joura']);
+        }
+        if (!empty($summary['author_str'])) {
+            $lines[] = 'Authors: ' . trim((string)$summary['author_str']);
+        }
+        if (!empty($summary['dateno'])) {
+            $lines[] = 'Publication: ' . trim((string)$summary['dateno']);
+        }
+        if (!empty($summary['doilink'])) {
+            $lines[] = 'DOI Link: ' . trim((string)$summary['doilink']);
+        }
+        if (!empty($summary['is_retracted'])) {
+            $lines[] = 'Retraction: yes - ' . trim((string)$this->arrGet($summary, 'retract_reason', ''));
+        }
+
+        $abstract = $this->cleanCrossrefAbstract((string)$this->arrGet($msg, 'abstract', ''));
+        $hasAbstract = $abstract !== '';
+        if ($hasAbstract) {
+            $lines[] = 'Abstract: ' . $this->truncate($abstract, 3500);
+        } else {
+            $lines[] = 'Note: Crossref 未返回摘要，请结合标题/期刊/作者与正文谨慎判断。';
+        }
+
+        return ['text' => implode("\n", $lines), 'has_abstract' => $hasAbstract];
+    }
+
+    private function cleanCrossrefAbstract($raw)
+    {
+        $raw = trim((string)$raw);
+        if ($raw === '') {
+            return '';
+        }
+        $raw = preg_replace('~<jats:title[^>]*>.*?</jats:title>~is', '', $raw);
+        $raw = preg_replace('~<jats:p[^>]*>~i', "\n", $raw);
+        $raw = preg_replace('~</jats:p>~i', '', $raw);
+        $raw = preg_replace('~</?jats:[^>]+>~i', '', $raw);
+        $raw = strip_tags($raw);
+        $raw = preg_replace('/[ \t]+/u', ' ', $raw);
+        $raw = preg_replace("/\r\n|\r/u", "\n", $raw);
+        $raw = preg_replace("/\n{2,}/u", "\n", $raw);
+        return trim($raw);
+    }
+
+    private function truncate($text, $max)
+    {
+        $text = (string)$text;
+        if (mb_strlen($text) <= $max) {
+            return $text;
+        }
+        return mb_substr($text, 0, $max) . '...';
+    }
+
+    /**
+     * 第二次 DOI 复核数据准备：返回书目信息 + 真实抓取内容
+     *
+     * @return array{refer_text:string, doi_block:string, has_abstract:bool, doi_used:string}
+     */
+    public function prepareRecheckPayload($refer, $referText = '')
+    {
+        $base = trim($referText) !== '' ? trim($referText) : $this->formatReferForLlm($refer);
+        $cr = $this->fetchCrossrefAbstractByReferDoi($refer);
+        return [
+            'refer_text'   => $base,
+            'doi_block'    => $cr['text'],
+            'has_abstract' => $cr['has_abstract'],
+            'doi_used'     => $cr['doi'],
+        ];
+    }
+
+    /**
+     * 旧接口：拼接成单块文本（向后兼容，建议调用方改用 prepareRecheckPayload）
+     */
+    public function formatReferForDoiRecheck($refer, $referText = '')
+    {
+        $payload = $this->prepareRecheckPayload($refer, $referText);
+        if ($payload['doi_block'] === '') {
+            return $payload['refer_text']
+                . "\n\n【DOI 文献真实内容】\n未能从 PubMed/Crossref 获取该 DOI 的摘要或元数据，请依据书目条目与正文谨慎判断。";
+        }
+        return $payload['refer_text']
+            . "\n\n【Crossref 摘要（依据 Refer_doi 从 api.crossref.org/works 获取）】\n"
+            . $payload['doi_block'];
+    }
+
+    /**
+     * 第一轮 confidence<=0.65 且能抓到 DOI 真实内容时，延迟入队第二轮复核
+     *
+     * 跳过条件（避免无意义重跑得到相同结果）：
+     * - check_id 不合法 / 一次置信度高于阈值
+     * - refer 行不存在
+     * - refer_doi 为空或 Crossref 未返回摘要
+     */
+    public function maybeEnqueueSecondPass($checkId, $confidence)
+    {
+        $checkId = intval($checkId);
+        $confidence = floatval($confidence);
+        if ($checkId <= 0 || $confidence > 0.65) {
+            return false;
+        }
+
+        $row = Db::name('article_reference_check_result')->where('id', $checkId)->find();
+        if (empty($row)) {
+            return false;
+        }
+
+        $refer = null;
+        if (intval($row['p_refer_id']) > 0) {
+            $refer = Db::name('production_article_refer')
+                ->where('p_refer_id', intval($row['p_refer_id']))
+                ->where('state', 0)
+                ->find();
+        }
+        if (empty($refer) || $this->extractReferDoiOnly($refer) === '') {
+            return false;
+        }
+
+        $cr = $this->fetchCrossrefAbstractByReferDoi($refer);
+        if (empty($cr['has_abstract'])) {
+            return false;
+        }
+
+        $this->pushJob2($checkId, 5);
+        return true;
+    }
+
     /**
      * 从 article_main.content 提取 blue 引用
      */
@@ -1021,10 +1561,24 @@ class ReferenceCheckService
             } else {
                 $jobId = Queue::push($jobClass, $data, self::QUEUE_NAME);
             }
-            var_dump("=====jobId:".$jobId);
         } catch (\Exception $e) {
             \think\Log::error('ReferenceCheck pushJob failed check_id=' . $checkId . ' ' . $e->getMessage());
             throw $e;
         }
     }
+    private function pushJob2($checkId, $delaySeconds = 0)
+    {
+        $jobClass = 'app\api\job\ReferenceCheckTwo@fire';
+        $data     = ['check_id' => $checkId];
+        try {
+            if ($delaySeconds > 0) {
+                $jobId = Queue::later($delaySeconds, $jobClass, $data, self::QUEUE_NAME);
+            } else {
+                $jobId = Queue::push($jobClass, $data, self::QUEUE_NAME);
+            }
+        } catch (\Exception $e) {
+            \think\Log::error('ReferenceCheckTwo pushJob failed check_id=' . $checkId . ' ' . $e->getMessage());
+            throw $e;
+        }
+    }
 }