optimize catalog semantic match sby tags

python/vector/vector_service.py

@@ -42,6 +42,9 @@ INDEX_META_PATH = KNOWLEDGE_DIR / "index_meta.json"
 INDEX_RUNTIME_PATH = KNOWLEDGE_DIR / "index_runtime.json"
 INDEX_NDJSON_PATH = KNOWLEDGE_DIR / "index.ndjson"
 
+# NEW: Tags NDJSON (exported by PHP) used to enrich /search-tags responses
+TAGS_NDJSON_PATH = KNOWLEDGE_DIR / "tags.ndjson"
+
 
 # ============================================================
 # Logging
@@ -111,6 +114,9 @@ chunk_pos_map: Dict[str, int] = {}
 tag_index = None
 tag_ids: Optional[List[Any]] = None
 
+# NEW: tag_id -> {"label": "...", "tag_type": "..."}
+tag_meta_map: Dict[str, Dict[str, str]] = {}
+
 loaded_embedding_model_name: Optional[str] = None
 current_index_version: Optional[int] = None
 current_runtime_stamp: Optional[str] = None
@@ -210,6 +216,61 @@ def load_chunk_maps_from_ndjson() -> None:
         logger.warning("Failed to load chunk maps from ndjson: %s", str(e))
 
 
+def load_tag_meta_from_tags_ndjson() -> None:
+    """
+    Loads minimal tag metadata from tags.ndjson to enrich /search-tags results.
+    Expected line format (from PHP exporter / ingester pipeline):
+      {"tag_id":"...","text":"LABEL\\nSLUG\\noptional description", ...}
+    We extract:
+      label = first line of "text" (fallback: "")
+      tag_type = "type" if present (preferred), else "generic"
+    """
+    global tag_meta_map
+
+    tag_meta_map = {}
+
+    if not TAGS_NDJSON_PATH.exists():
+        logger.info("[Reload] tags.ndjson missing -> tag_meta_map empty (%s)", str(TAGS_NDJSON_PATH))
+        return
+
+    try:
+        with TAGS_NDJSON_PATH.open("r", encoding="utf-8") as f:
+            for line in f:
+                line = line.strip()
+                if not line:
+                    continue
+                try:
+                    row = json.loads(line)
+                except Exception:
+                    continue
+
+                tag_id = _as_key(row.get("tag_id"))
+                if not tag_id:
+                    continue
+
+                # Prefer explicit fields if present
+                ttype = row.get("type")
+                if isinstance(ttype, str) and ttype.strip():
+                    tag_type = ttype.strip()
+                else:
+                    tag_type = "generic"
+
+                label = ""
+                txt = row.get("text")
+                if isinstance(txt, str) and txt.strip():
+                    first = txt.splitlines()[0].strip() if txt.splitlines() else ""
+                    label = first
+
+                if label:
+                    tag_meta_map[tag_id] = {"label": label, "tag_type": tag_type}
+                else:
+                    tag_meta_map[tag_id] = {"label": "", "tag_type": tag_type}
+
+    except Exception as e:
+        logger.warning("Failed to load tag meta from tags.ndjson: %s", str(e))
+        tag_meta_map = {}
+
+
 def _normalize_meta_list(value: Any) -> Optional[List[Any]]:
     """
     Accepts:
@@ -282,6 +343,10 @@ def load_all() -> None:
             tag_index = None
             tag_ids = None
 
+        # NEW: load tag meta for enrichment
+        logger.info("[Reload] Loading tag meta from tags.ndjson")
+        load_tag_meta_from_tags_ndjson()
+
         runtime = _safe_read_json(INDEX_RUNTIME_PATH)
         if isinstance(runtime, dict):
             v = runtime.get("last_rebuild_at")
@@ -292,10 +357,11 @@ def load_all() -> None:
         current_index_version = index_version if isinstance(index_version, int) else None
 
         logger.info(
-            "[Reload] Completed (index_version=%s runtime=%s embedding_model=%s stamp=%s file=%s)",
+            "[Reload] Completed (index_version=%s runtime=%s embedding_model=%s tag_meta=%s stamp=%s file=%s)",
             str(current_index_version),
             str(current_runtime_stamp),
             str(loaded_embedding_model_name),
+            str(len(tag_meta_map)),
             SERVICE_STAMP,
             str(Path(__file__).resolve()),
         )
@@ -390,6 +456,8 @@ def health():
         "tag_meta_len": len(tag_ids) if isinstance(tag_ids, list) else None,
         "chunk_meta_type": type(chunk_ids).__name__ if chunk_ids is not None else None,
         "chunk_meta_len": len(chunk_ids) if isinstance(chunk_ids, list) else None,
+        "tag_meta_map_len": len(tag_meta_map),
+        "tags_ndjson_path": str(TAGS_NDJSON_PATH),
         "log_file": str(LOG_FILE),
     }
 
@@ -502,7 +570,26 @@ def search_tags(req: SearchRequest):
                 continue
             if idx < 0 or idx >= len(tag_ids):
                 continue
-            results.append({"tag_id": tag_ids[idx], "score": float(score)})
+
+            tag_id = tag_ids[idx]
+            tag_id_key = _as_key(tag_id) or ""
+
+            payload: Dict[str, Any] = {
+                "tag_id": tag_id,
+                "score": float(score),
+            }
+
+            meta = tag_meta_map.get(tag_id_key)
+            if isinstance(meta, dict):
+                label = meta.get("label")
+                ttype = meta.get("tag_type")
+
+                if isinstance(label, str) and label.strip():
+                    payload["label"] = label
+                if isinstance(ttype, str) and ttype.strip():
+                    payload["tag_type"] = ttype
+
+            results.append(payload)
 
         return results