Fix chinese character bug in streaming json & fix pickle load bug for…

… bm25 index (#239)

* Fix bug

* Fix linter

src/pai_rag/core/rag_application.py

@@ -43,7 +43,7 @@ async def event_generator_async(
         if token and token != DEFAULT_EMPTY_RESPONSE_GEN:
             chunk = {"delta": token, "is_finished": False}
             content += token
-            yield json.dumps(chunk) + "\n"
+            yield json.dumps(chunk, ensure_ascii=False) + "\n"
 
     if chat_store:
         chat_store.add_message(
@@ -56,7 +56,7 @@ async def event_generator_async(
     else:
         last_chunk = {"delta": "", "is_finished": True}
 
-    yield json.dumps(last_chunk, default=lambda x: x.dict())
+    yield json.dumps(last_chunk, default=lambda x: x.dict(), ensure_ascii=False)
 
 
 class RagApplication:

src/pai_rag/integrations/index/pai/local/local_bm25_index.py

@@ -37,6 +37,20 @@
 DEFAULT_INDEX_MATRIX_FILE = "bm25.index.matrix.pkl"
 
 
+# Handle module mismatch issue.
+class RenameUnpickler(pickle.Unpickler):
+    def find_class(self, module, name):
+        renamed_module = module
+        if module == "pai_rag.modules.index.pai_bm25_index":
+            renamed_module = "pai_rag.integrations.index.pai.local.local_bm25_index"
+
+        return super(RenameUnpickler, self).find_class(renamed_module, name)
+
+
+def renamed_load(file_obj):
+    return RenameUnpickler(file_obj).load()
+
+
 class LocalBm25Index:
     def __init__(self):
         self.doc_count: int = 0
@@ -76,7 +90,7 @@ def __init__(
     def reload(self):
         if os.path.exists(self.parts_path):
             with open(self.index_file, "rb") as f:
-                self.index: LocalBm25Index = pickle.load(f)
+                self.index: LocalBm25Index = renamed_load(f)
             with open(self.index_matrix_file, "rb") as f:
                 self.index_matrix = pickle.load(f)
         else: