Bump qdrant-client from 1.8.2 to 1.9.0 in /intel_extension_for_transformers/neural_chat/tests

intel_extension_for_transformers/neural_chat/pipeline/plugins/retrieval/retrieval_agent.py

@@ -189,9 +189,6 @@ def __init__(self,
                 knowledge_base = self.database.build(documents=langchain_documents, embedding=self.embeddings, **kwargs)
             self.retriever = RetrieverAdapter(retrieval_type=self.retrieval_type, document_store=knowledge_base, \
                                               **kwargs)
-            if self.vector_database == "Qdrant" and knowledge_base.is_local():
-               # one local storage folder cannot be accessed by multiple instances of Qdrant client simultaneously.
-               knowledge_base.client.close()
         elif self.retrieval_type == "child_parent":    # Using child-parent store retriever
             child_documents = self.splitter.split_documents(langchain_documents)
             langchain_documents = document_append_id(langchain_documents)
@@ -206,12 +203,6 @@ def __init__(self,
                                             sign='child', **kwargs)
             self.retriever = RetrieverAdapter(retrieval_type=self.retrieval_type, document_store=knowledge_base, \
                                child_document_store=child_knowledge_base, **kwargs)
-            if self.vector_database == "Qdrant" :
-                # one local storage folder cannot be accessed by multiple instances of Qdrant client simultaneously.
-                if knowledge_base.is_local():
-                    knowledge_base.client.close()
-                if child_knowledge_base.is_local():
-                    child_knowledge_base.client.close()
         elif self.retrieval_type == "bm25":
             self.docs = document_append_id(langchain_documents)
             self.retriever = RetrieverAdapter(retrieval_type=self.retrieval_type, docs=self.docs, **kwargs)
@@ -341,4 +332,23 @@ def pre_llm_inference_actions(self, model_name, query):
                 prompt = generate_qa_prompt(query, context)
         else:
             logging.error("The selected generation mode is invalid!")
+
+        # qdrant local vector db need to be closed
+        # one local storage folder cannot be accessed by multiple instances of Qdrant client simultaneously.
+        if self.vector_database == "Qdrant":
+            to_close = []
+            if self.retrieval_type == "default":
+                knowledge_base = self.retriever.retriever.vectorstore
+                if knowledge_base.is_local():
+                    to_close.append(knowledge_base)
+            if self.retrieval_type == "child_parent":
+                knowledge_base = self.retriever.retriever.parentstore
+                child_knowledge_base = self.retriever.retriever.vectorstore
+                if knowledge_base.is_local():
+                    to_close.append(knowledge_base)
+                if child_knowledge_base.is_local():
+                    to_close.append(child_knowledge_base)
+            for kb in to_close:
+                kb.client.close()
+
         return prompt, links

intel_extension_for_transformers/neural_chat/tests/requirements.txt

@@ -63,7 +63,7 @@ pypinyin
 python-docx
 python-multipart
 pyyaml
-qdrant-client==1.8.2
+qdrant-client==1.9.0
 rank_bm25
 resampy==0.3.1
 rouge_score