[dev] rename config worker name

config/demo_config.yaml

@@ -1,11 +1,13 @@
 global_config:
   language: cn
   max_workers: 5
+
 memory_chat:
   cli_memory_chat:
     class: chat.cli_memory_chat
     memory_service: memory_chat_service
     generation_model: dashscope_generation
+
 memory_service:
   memory_chat_service:
     class: memory.service.chat_memory_service
@@ -18,40 +20,42 @@ memory_service:
         description: "read session messages of the user"
       read_memory:
         class: memory.operation.read_memory
-        workflow: set_query_worker,retrieve_store_worker,[extract_time_worker|semantic_rank_worker],fuse_rerank_worker
+        workflow: set_query,retrieve_memory1,[extract_time|semantic_rank],fuse_rerank
         description: "read related memories of the user"
       list_memory:
         class: memory.operation.read_memory
-        workflow: set_query_worker,retrieve_store_worker,print_memory_worker
+        workflow: set_query,retrieve_memory2,print_memory
         description: "read all memories of the user"
       write_memory:
         class: memory.operation.write_memory
-        workflow: info_filter_worker,[get_observation_worker|get_observation_with_time_worker],contra_repeat_worker,store_memory_worker
+        workflow: info_filter,load_memory1,[get_observation|get_observation_with_time],contra_repeat,store_memory
         description: "write observation memories of the user"
         interval_time: 60
       summary_memory:
         class: memory.operation.summary_memory
-        workflow: load_memory_worker,get_reflection_subject_worker,update_insight_worker,long_contra_repeat_worker,summary_collect_worker
+        workflow: load_memory2,get_reflection_subject,update_insight,long_contra_repeat,store_memory
         description: "summary observation memories of the user"
         interval_time: 300
+
 worker:
-  dummy_worker:
+  dummy:
     class: memory.worker.dummy_worker
     generation_model: dashscope_generation
     embedding_model: dashscope_embedding
     rank_model: dashscope_rank
-  set_query_worker:
+  set_query:
     class: memory.worker.read.set_query_worker
-  retrieve_store_worker:
-    class: memory.worker.read.retrieve_store_worker
+  retrieve_memory1:
+    class: memory.worker.read.retrieve_memory_worker
     retrieve_obs_top_k: 100
     retrieve_ins_pf_top_k: 100
-  extract_time_worker:
+    retrieve_expired_top_k: 0
+  extract_time:
     class: memory.worker.read.extract_time_worker
     generation_model_top_k: 1
-  semantic_rank_worker:
+  semantic_rank:
     class: memory.worker.read.semantic_rank_worker
-  fuse_rerank_worker:
+  fuse_rerank:
     class: memory.worker.read.fuse_rerank_worker
     fuse_score_threshold: 0.1
     fuse_ratio_dict:
@@ -61,41 +65,63 @@ worker:
       insight: 2.0
     fuse_time_ratio: 2.0
     fuse_rerank_top_k: 10
-  print_memory_worker:
+  retrieve_memory2:
+    class: memory.worker.read.retrieve_memory_worker
+    retrieve_obs_top_k: 100
+    retrieve_ins_pf_top_k: 100
+    retrieve_expired_top_k: 100
+  print_memory:
     class: memory.worker.read.print_memory_worker
-  info_filter_worker:
+  info_filter:
     class: memory.worker.write.info_filter_worker
     generation_model: dashscope_generation
     info_filter_msg_max_size: 200
     generation_model_top_k: 1
-  get_observation_worker:
+  load_memory1:
+    class: memory.worker.write.load_memory_worker
+    retrieve_not_reflected_top_k: 0
+    retrieve_not_updated_top_k: 0
+    retrieve_insight_top_k: 0
+    today_obs_top_k: 100
+  get_observation:
     class: memory.worker.write.get_observation_worker
     generation_model: dashscope_generation
     generation_model_top_k: 1
-  get_observation_with_time_worker:
+  get_observation_with_time:
     class: memory.worker.write.get_observation_with_time_worker
     generation_model: dashscope_generation
     generation_model_top_k: 1
-  contra_repeat_worker:
+  contra_repeat:
     class: memory.worker.write.contra_repeat_worker
     generation_model: dashscope_generation
     generation_model_top_k: 1
     retrieve_top_k: 30
     contra_repeat_max_count: 50
-  store_memory_worker:
+  store_memory:
     class: memory.worker.write.store_memory_worker
-  load_memory_worker:
-    class: memory.worker.summary.load_memory_worker
-  get_reflection_subject_worker:
+    store_key: all
+  load_memory2:
+    class: memory.worker.write.load_memory_worker
+    retrieve_not_reflected_top_k: 100
+    retrieve_not_updated_top_k: 100
+    retrieve_insight_top_k: 100
+    today_obs_top_k: 0
+  get_reflection_subject:
     class: memory.worker.summary.get_reflection_subject_worker
     retrieve_top_k: 100
     reflect_obs_cnt_threshold: 32
-  update_insight_worker:
+    generation_model_top_k: 1
+  update_insight:
     class: memory.worker.summary.update_insight_worker
-  long_contra_repeat_worker:
+    update_insight_threshold: 0.1
+    generation_model_top_k: 1
+    update_insight_max_thread: 10
+  long_contra_repeat:
     class: memory.worker.summary.long_contra_repeat_worker
-  summary_collect_worker:
-    class: memory.worker.summary.summary_collect_worker
+    long_contra_repeat_top_k: 2
+    long_contra_repeat_threshold: 0.1
+    generation_model_top_k: 1
+
 models:
   dashscope_generation:
     class: models.llama_index_generation_model
@@ -109,11 +135,13 @@ models:
     class: models.llama_index_rank_model
     module_name: dashscope_rank
     model_name: gte-rerank
+
 memory_store:
   class: storage.llama_index_es_memory_store
   embedding_model: dashscope_embedding
   index_name: memory_index
   es_url: http://localhost:9200
   use_hybrid: false
+
 monitor:
   class: storage.dummy_monitor

memory_scope/memory/worker/read/set_query_worker.py

@@ -8,6 +8,8 @@ class SetQueryWorker(MemoryBaseWorker):
 
     def _run(self):
         if "query" in self.chat_kwargs:
+            """ cli test query
+            """
             query = self.chat_kwargs["query"]
             query_timestamp = int(datetime.datetime.now().timestamp())
         else:

memory_scope/storage/llama_index_es_memory_store.py

@@ -12,9 +12,7 @@ from memory_scope.utils.logger import Logger
 
 
 class _AsyncDenseVectorStrategy(AsyncDenseVectorStrategy):
-    def _hybrid(
-            self, query: str, knn: Dict[str, Any], filter: List[Dict[str, Any]], top_k: int,
-    ) -> Dict[str, Any]:
+    def _hybrid(self, query: str, knn: Dict[str, Any], filter: List[Dict[str, Any]], top_k: int) -> Dict[str, Any]:
         # Add a query to the knn query.
         # RRF is used to even the score from the knn query and text query
         # RRF has two optional parameters: {'rank_constant':int, 'window_size':int}
@@ -52,8 +50,11 @@ class _AsyncDenseVectorStrategy(AsyncDenseVectorStrategy):
             vector_field: str,
             k: int,
             num_candidates: int,
-            filter: List[Dict[str, Any]] = [],
+            filter: List[Dict[str, Any]] = None,
     ) -> Dict[str, Any]:
+        if filter is None:
+            filter = []
+
         knn = {
             "filter": filter,
             "field": vector_field,