optimize catalog semantic match sby tags

src/Tag/TagNdjsonExporter.php

@@ -19,7 +19,12 @@ final readonly class TagNdjsonExporter
      * Export all tags into NDJSON (streaming) with atomic switch (.tmp + rename()).
      *
      * Line format:
-     * {"tag_id":"...","text":"label\nslug\noptional description","document_ids":["...","..."]}
+     * {
+     *   "tag_id":"...",
+     *   "text":"label\nslug\noptional description",
+     *   "type":"catalog_entity|generic|...",
+     *   "document_ids":["...","..."]
+     * }
      *
      * @return array{tags:int, lines:int, bytes:int, path:string}
      */
@@ -37,12 +42,7 @@ final readonly class TagNdjsonExporter
             throw new \RuntimeException('Cannot write tags NDJSON: ' . $tmpPath);
         }
 
-        // ------------------------------------------------------------------
-        // Fetch tags (small) + join document ids (can be bigger) efficiently.
-        // We avoid repositories and keep it DB-agnostic via DQL/QB.
-        // ------------------------------------------------------------------
-
-        // 1) Load all tags (id, slug, label, description)
+        // 1) Load all tags
         $tags = $this->em->createQueryBuilder()
             ->select('t')
             ->from(Tag::class, 't')
@@ -52,8 +52,6 @@ final readonly class TagNdjsonExporter
 
         if (!\is_array($tags) || $tags === []) {
             \fclose($fh);
-
-            // Write empty file atomically
             $this->atomicReplace($tmpPath, $this->tagsNdjsonPath);
 
             return [
@@ -64,8 +62,7 @@ final readonly class TagNdjsonExporter
             ];
         }
 
-        // 2) Build tagId => docIds map from document_tag
-        //    We query pairs (tag_id, document_id) in one go.
+        // 2) Build tagId => docIds map
         $rows = $this->em->createQueryBuilder()
             ->select('IDENTITY(dt.tag) AS tagId', 'IDENTITY(dt.document) AS docId')
             ->from(DocumentTag::class, 'dt')
@@ -82,7 +79,7 @@ final readonly class TagNdjsonExporter
             $tagToDocs[$tagId][] = $docId;
         }
 
-        // 3) Stream NDJSON lines
+        // 3) Stream NDJSON
         $lines = 0;
 
         foreach ($tags as $tag) {
@@ -93,13 +90,11 @@ final readonly class TagNdjsonExporter
             $tagId = (string) $tag->getId();
             $docIds = $tagToDocs[$tagId] ?? [];
 
-            // de-dupe docIds for safety
             if ($docIds !== []) {
                 $docIds = \array_values(\array_unique($docIds));
             }
 
-            // "text" is the embedding source for tag vectors later:
-            // Keep it short but semantically useful.
+            // Embedding source
             $textParts = [
                 $tag->getLabel(),
                 $tag->getSlug(),
@@ -110,15 +105,23 @@ final readonly class TagNdjsonExporter
                 $textParts[] = \trim($desc);
             }
 
+            $type = method_exists($tag, 'getType')
+                ? (string) $tag->getType()
+                : 'generic';
+
+            if ($type === '') {
+                $type = 'generic';
+            }
+
             $line = [
                 'tag_id' => $tagId,
                 'text' => \implode("\n", $textParts),
+                'type' => $type,                  // 🔥 NEW
                 'document_ids' => $docIds,
             ];
 
             $json = \json_encode($line, JSON_UNESCAPED_UNICODE | JSON_UNESCAPED_SLASHES);
             if (!\is_string($json)) {
-                // skip invalid line but keep export running
                 continue;
             }
 
@@ -127,7 +130,6 @@ final readonly class TagNdjsonExporter
         }
 
         \fclose($fh);
-
         $this->atomicReplace($tmpPath, $this->tagsNdjsonPath);
 
         return [
@@ -140,13 +142,11 @@ final readonly class TagNdjsonExporter
 
     private function atomicReplace(string $tmpPath, string $finalPath): void
     {
-        // Ensure old file can be replaced on Windows-like FS too (best effort)
         if (\is_file($finalPath)) {
             @\chmod($finalPath, 0664);
         }
 
         if (!@\rename($tmpPath, $finalPath)) {
-            // if rename fails, try copy+unlink fallback
             if (!@\copy($tmpPath, $finalPath)) {
                 @\unlink($tmpPath);
                 throw new \RuntimeException('Atomic replace failed for: ' . $finalPath);

src/Tag/TagService.php

@@ -21,8 +21,12 @@ final readonly class TagService
     // TAG CREATE
     // =========================================================
 
-    public function create(string $slug, string $label, ?string $description = null): Tag
-    {
+    public function create(
+        string $slug,
+        string $label,
+        ?string $description = null,
+        string $type = 'generic' // NEU
+    ): Tag {
         $slug  = trim($slug);
         $label = trim($label);
 
@@ -35,6 +39,7 @@ final readonly class TagService
         }
 
         $tag = new Tag($slug, $label, $description);
+        $tag->setType($type); // NEU
 
         $this->em->persist($tag);
         $this->em->flush();
@@ -71,10 +76,6 @@ final readonly class TagService
     // DOCUMENT TAG SYNC
     // =========================================================
 
-    /**
-     * Synchronisiert alle Tags eines Dokuments.
-     * Löst einen Rebuild aus, da document_ids Teil des NDJSON sind.
-     */
     public function syncDocumentTags(Document $document, array $newTagIds): void
     {
         $newTagIds = array_unique($newTagIds);
@@ -114,10 +115,6 @@ final readonly class TagService
     // TAG → DOCUMENT SYNC (Bulk Assign)
     // =========================================================
 
-    /**
-     * Synchronisiert alle Dokumente eines Tags.
-     * Löst einen Rebuild aus, da document_ids Teil des NDJSON sind.
-     */
     public function syncTagDocuments(Tag $tag, array $newDocumentIds): void
     {
         $newDocumentIds = array_unique($newDocumentIds);

src/Tag/TagTypes.php

@@ -0,0 +1,27 @@
+<?php
+
+declare(strict_types=1);
+
+namespace App\Tag;
+
+/**
+ * Zentrale Definition aller erlaubten Tag-Typen.
+ * Verhindert Magic Strings im Code.
+ */
+final class TagTypes
+{
+    public const GENERIC = 'generic';
+    public const CATALOG_ENTITY = 'catalog_entity';
+    public const SALES_SIGNAL = 'sales_signal';
+
+    public static function choices(): array
+    {
+        return [
+            'Generic' => self::GENERIC,
+            'Catalog Entity' => self::CATALOG_ENTITY,
+            'Sales Signal' => self::SALES_SIGNAL,
+        ];
+    }
+
+    private function __construct() {}
+}

src/Tag/TagVectorSearchClient.php

@@ -11,7 +11,6 @@ final readonly class TagVectorSearchClient
 {
     /**
      * Minimum similarity score required for a tag to be considered.
-     * Acts as a confidence gate to avoid noisy routing.
      */
     private const MIN_SCORE = 0.4;
 
@@ -29,7 +28,22 @@ final readonly class TagVectorSearchClient
     /**
      * Executes a vector search against the Python tag index.
      *
-     * @return array<int, array{tag_id:string, score:float}>
+     * Expected response rows:
+     * [
+     *   {
+     *     "tag_id": "...",
+     *     "score": 0.73,
+     *     "label": "Geräte",          // optional (new)
+     *     "tag_type": "catalog_entity" // optional (new)
+     *   }
+     * ]
+     *
+     * @return array<int, array{
+     *   tag_id:string,
+     *   score:float,
+     *   label?:string,
+     *   tag_type?:string
+     * }>
      */
     public function search(string $query, int $limit = 8): array
     {
@@ -94,15 +108,26 @@ final readonly class TagVectorSearchClient
 
             $score = (float) $score;
 
-            // 🔥 Confidence Gate
             if ($score < self::MIN_SCORE) {
                 continue;
             }
 
-            $hits[] = [
+            $hit = [
                 'tag_id' => $tagId,
                 'score'  => $score,
             ];
+
+            // Optional: label
+            if (isset($row['label']) && is_string($row['label'])) {
+                $hit['label'] = $row['label'];
+            }
+
+            // Optional: tag_type
+            if (isset($row['tag_type']) && is_string($row['tag_type'])) {
+                $hit['tag_type'] = $row['tag_type'];
+            }
+
+            $hits[] = $hit;
         }
 
         return $hits;