From acbc886ecd578b56a745d9187d1064d3ca3cfd87 Mon Sep 17 00:00:00 2001
From: Conner Mo <conner.mo@gmail.com>
Date: Sat, 29 Nov 2025 18:50:21 +0800
Subject: [PATCH] fix: implement score_threshold filtering for OceanBase vector
 search (#28536)

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
---
 .../vdb/oceanbase/oceanbase_vector.py         | 24 +++++++++++++------
 1 file changed, 17 insertions(+), 7 deletions(-)

diff --git a/api/core/rag/datasource/vdb/oceanbase/oceanbase_vector.py b/api/core/rag/datasource/vdb/oceanbase/oceanbase_vector.py
index b3db7332e8..7b53f47419 100644
--- a/api/core/rag/datasource/vdb/oceanbase/oceanbase_vector.py
+++ b/api/core/rag/datasource/vdb/oceanbase/oceanbase_vector.py
@@ -270,6 +270,10 @@ class OceanBaseVector(BaseVector):
             self._client.set_ob_hnsw_ef_search(ef_search)
             self._hnsw_ef_search = ef_search
         topk = kwargs.get("top_k", 10)
+        try:
+            score_threshold = float(val) if (val := kwargs.get("score_threshold")) is not None else 0.0
+        except (ValueError, TypeError) as e:
+            raise ValueError(f"Invalid score_threshold parameter: {e}") from e
         try:
             cur = self._client.ann_search(
                 table_name=self._collection_name,
@@ -285,14 +289,20 @@ class OceanBaseVector(BaseVector):
             raise Exception("Failed to search by vector. ", e)
         docs = []
         for _text, metadata, distance in cur:
-            metadata = json.loads(metadata)
-            metadata["score"] = 1 - distance / math.sqrt(2)
-            docs.append(
-                Document(
-                    page_content=_text,
-                    metadata=metadata,
+            score = 1 - distance / math.sqrt(2)
+            if score >= score_threshold:
+                try:
+                    metadata = json.loads(metadata)
+                except json.JSONDecodeError:
+                    logger.warning("Invalid JSON metadata: %s", metadata)
+                    metadata = {}
+                metadata["score"] = score
+                docs.append(
+                    Document(
+                        page_content=_text,
+                        metadata=metadata,
+                    )
                 )
-            )
         return docs
 
     def delete(self):