From 75d01299092be1577cb196b8cd03f9085010a162 Mon Sep 17 00:00:00 2001 From: hs Date: Tue, 18 Jun 2024 19:25:00 +0800 Subject: [PATCH] fix api --- config/config.json | 6 +- config/worker.json | 74 ++++++++++------------- memory_scope/cli.py | 19 +++--- memory_scope/config.py | 2 - memory_scope/pipeline/memory.py | 6 +- memory_scope/utils/tool_functions.py | 7 ++- memory_scope/worker/memory_base_worker.py | 25 ++++++-- tests/test_memory.py | 18 +++--- 8 files changed, 81 insertions(+), 76 deletions(-) diff --git a/config/config.json b/config/config.json index 9f4d9316..899a84bd 100644 --- a/config/config.json +++ b/config/config.json @@ -2,9 +2,9 @@ "thread_pool_max_count": "", "worker": "config/worker.json", "pipeline": { - "summary_short": "", - "summary_long": "", - "retrieve": "" + "summary_short": "parse_params,[load_profile|es_new_obs|es_insight],[update_insight|get_reflection,get_insight|update_profile],summary_collect,memory_store", + "summary_long": "parse_params,info_filter,[es_today_obs|get_observation|get_observation_with_time],contra_repeat,memory_store", + "retrieve": "parse_params,load_profile,[extract_time|es_similar|es_keyword],semantic_rank,fuse_rerank" }, "model_embedding": "config/model/dash_embedding.json", "model_rerank": "config/model/dash_rerank.json", diff --git a/config/worker.json b/config/worker.json index bcce1b54..87724ef3 100644 --- a/config/worker.json +++ b/config/worker.json @@ -1,52 +1,45 @@ { - "EsInsightWorker": { - "module_name": "EsInsightWorker", - "module_path": "memory_scope/worker", - "kwargs": { - "es_insight_top_k": 128 - } + "es_insight": { + "name": "EsInsightWorker", + "path": "memory_scope/worker", + "es_insight_top_k": 128 }, "es_keyword": { "name": "EsKeywordWorker", "path": "memory_scope/worker", "es_keyword_top_k": 10 }, - "es_keyword2": { - "module_name": "EsKeywordWorker", - "module_path": "memory_scope/worker", - "es_keyword_top_k": 10 - }, "EsNewObsWorker": { - "module_name": "EsNewObsWorker", - "module_path": "memory_scope/worker", + "name": "EsNewObsWorker", + "path": "memory_scope/worker", "kwargs": { "es_new_obs_top_k": 256 } }, "EsNotReflectedWorker": { - "module_name": "EsNotReflectedWorker", - "module_path": "memory_scope/worker", + "name": "EsNotReflectedWorker", + "path": "memory_scope/worker", "kwargs": { "es_not_reflected_top_k": 256 } }, "EsSimilarWorker": { - "module_name": "EsSimilarWorker", - "module_path": "memory_scope/worker", + "name": "EsSimilarWorker", + "path": "memory_scope/worker", "kwargs": { "es_similar_top_k": 128 } }, "EsTodayObsWorker": { - "module_name": "EsTodayObsWorker", - "module_path": "memory_scope/worker", + "name": "EsTodayObsWorker", + "path": "memory_scope/worker", "kwargs": { "es_today_obs_top_k": 128 } }, "GetInsightWorker": { - "module_name": "GetInsightWorker", - "module_path": "memory_scope/worker", + "name": "GetInsightWorker", + "path": "memory_scope/worker", "kwargs": { "es_insight_similar_top_k": 128, "insight_obs_max_cnt": 10, @@ -57,17 +50,16 @@ } }, "ExtractTimeWorker": { - "module_name": "ExtractTimeWorker", - "module_path": "memory_scope/worker", - "parse_time_model": "", + "name": "ExtractTimeWorker", + "path": "memory_scope/worker", "parse_time_model": "qwen_1_8_parse_time_service", "parse_time_max_token": 100, "parse_time_temperature": 0.6, - "parse_time_top_k": 1, + "parse_time_top_k": 1 }, "InfoFilterWorker": { - "module_name": "InfoFilterWorker", - "module_path": "memory_scope/worker", + "name": "InfoFilterWorker", + "path": "memory_scope/worker", "kwargs": { "info_filter_msg_max_size": 200, "info_filter_model": "qwen_max", @@ -77,8 +69,8 @@ } }, "GetObservationWithTimeWorker": { - "module_name": "GetObservationWithTimeWorker", - "module_path": "memory_scope/worker", + "name": "GetObservationWithTimeWorker", + "path": "memory_scope/worker", "kwargs": { "summary_messages_model": "qwen_max", "summary_messages_max_token": 500, @@ -87,8 +79,8 @@ } }, "GetObservationWorker": { - "module_name": "GetObservationWorker", - "module_path": "memory_scope/worker", + "name": "GetObservationWorker", + "path": "memory_scope/worker", "kwargs": { "summary_messages_model": "qwen_max", "summary_messages_max_token": 500, @@ -97,8 +89,8 @@ } }, "ContraRepeatWorker": { - "module_name": "ContraRepeatWorker", - "module_path": "memory_scope/worker", + "name": "ContraRepeatWorker", + "path": "memory_scope/worker", "kwargs": { "merge_obs_model": "qwen_max", "merge_obs_max_token": 500, @@ -107,8 +99,8 @@ } }, "FuseRerankWorker": { - "module_name": "FuseRerankWorker", - "module_path": "memory_scope/worker", + "name": "FuseRerankWorker", + "path": "memory_scope/worker", "kwargs": { "fuse_score_threshold": 0.1, "fuse_ratio_dict": { @@ -123,8 +115,8 @@ } }, "UpdateProfileWorker": { - "module_name": "UpdateProfileWorker", - "module_path": "memory_scope/worker", + "name": "UpdateProfileWorker", + "path": "memory_scope/worker", "kwargs": { "update_profile_threshold": 0.1, "update_profile_model": "qwen_max", @@ -135,8 +127,8 @@ } }, "GetReflectionWorker": { - "module_name": "GetReflectionWorker", - "module_path": "memory_scope/worker", + "name": "GetReflectionWorker", + "path": "memory_scope/worker", "kwargs": { "reflect_obs_cnt_threshold": 40, "reflect_num_questions": 3, @@ -147,8 +139,8 @@ } }, "UpdateInsightWorker": { - "module_name": "UpdateInsightWorker", - "module_path": "memory_scope/worker", + "name": "UpdateInsightWorker", + "path": "memory_scope/worker", "kwargs": { "update_insight_threshold": 0.1, "update_insight_model": "qwen_max", diff --git a/memory_scope/cli.py b/memory_scope/cli.py index fa024a08..208339c3 100644 --- a/memory_scope/cli.py +++ b/memory_scope/cli.py @@ -1,15 +1,14 @@ # 使用argparse库的示例 import argparse +import fire +from config import C, init +from chat.memory_chat import MemoryChat - -def main(): - parser = argparse.ArgumentParser(description="示例CLI程序") - parser.add_argument('--echo', help="输出传入的消息") - - args = parser.parse_args() - if args.echo: - print(f"收到的消息: {args.echo}") - +def main(config_path:str): + init(config_path) + + agent = MemoryChat() + agent.run() if __name__ == "__main__": - main() + fire.Fire(main) diff --git a/memory_scope/config.py b/memory_scope/config.py index dfca733b..d69c08c5 100644 --- a/memory_scope/config.py +++ b/memory_scope/config.py @@ -19,10 +19,8 @@ class Wrapper: raise AttributeError("Please run __init__ first!") return getattr(self._provider, key) - C = Wrapper() - def init(config_path: str): config = json.loads(config_path) C.register(config) diff --git a/memory_scope/pipeline/memory.py b/memory_scope/pipeline/memory.py index 769dab4b..062a9786 100644 --- a/memory_scope/pipeline/memory.py +++ b/memory_scope/pipeline/memory.py @@ -11,4 +11,8 @@ class MemoryServiceRequestModel(BaseModel): messages: List[Message] = Field(..., description="summary: 多轮对话的list,默认按照时间正序，最后一条是最新的; retrieve: 最后一条是query") - messages_pick_n: int = Field(1, description="summary：传需要总结的msg的个数;retrieve：不传") + user_profile: List[UserAttribute] = Field([], description="user_profile") + + ext_info: Dict[str, str] = Field({}, description="extra information") + + extra_user_attrs: List = [] \ No newline at end of file diff --git a/memory_scope/utils/tool_functions.py b/memory_scope/utils/tool_functions.py index 849c9e0d..fdf0d8ef 100644 --- a/memory_scope/utils/tool_functions.py +++ b/memory_scope/utils/tool_functions.py @@ -144,10 +144,13 @@ def time_to_formatted_str(time: datetime | str | int | float = None, return return_str -def init_instance_by_config(config: dict, default_module_path=None, try_kwargs={}): +def init_instance_by_config(config: dict|object, default_module_path: str = None, try_kwargs: dict = {}, accept_types: type = None): + if isinstance(config, accept_types): + return config + import_module(config.pop("path", default_module_path)) clazz = getattr(module, config.pop("name")) try: - return clazz(**config,**try_kwargs) + return clazz(**config, **try_kwargs) except: return clazz(**config) \ No newline at end of file diff --git a/memory_scope/worker/memory_base_worker.py b/memory_scope/worker/memory_base_worker.py index 850b0d3d..0abef156 100644 --- a/memory_scope/worker/memory_base_worker.py +++ b/memory_scope/worker/memory_base_worker.py @@ -8,6 +8,7 @@ from node.user_attribute import UserAttribute from pipeline.memory import MemoryServiceRequestModel from worker.base_worker import BaseWorker from cli.cli_config import C +from utils.tool_functions import init_instance_by_config class MemoryBaseWorker(BaseWorker): def __init__(self, **kwargs): @@ -21,7 +22,7 @@ class MemoryBaseWorker(BaseWorker): def messages(self) -> List[Message]: messages: List[Message] = self.context_handler.get_context(MESSAGES) if messages is None: - messages = self.request.messages[-self.request.messages_pick_n:] + messages = self.request.messages self.context_handler.set_context(MESSAGES, messages) return messages @@ -51,20 +52,32 @@ class MemoryBaseWorker(BaseWorker): return self.request.user.prompt @property - def emb_client(self): - return C.model_embedding + def client(self, model_type: str, model_name:str): + models = C.get(model_type) + models["model_name"] = init_instance_by_config( + config = models.get(model_name), + try_kwargs={ + "is_multi_thread": is_multi_thread, + "thread_pool": self.thread_pool + } + ) + return models["model_name"] + + @property + def emb_client(self, model_name: str): + self.client("model_embedding", model_name) @property def gene_client(self): - return C.model_generate + self.client("model_generate", model_name) @property def rerank_client(self): - return C.model_rerank + self.client("model_rerank", model_name) @property def es_client(self): - return C.db + self.client("db", model_name) @property def tenant_id(self): diff --git a/tests/test_memory.py b/tests/test_memory.py index a6ef0235..29a4cf53 100644 --- a/tests/test_memory.py +++ b/tests/test_memory.py @@ -3,13 +3,13 @@ import os import time from typing import List, Dict -from utils.logger import Logger -from constants.common_constants import NEW_USER_PROFILE, MODIFIED_MEMORIES, RELATED_MEMORIES -from enumeration.memory_method_enum import MemoryMethodEnum -from node.memory_node import MemoryNode -from node.user_attribute import UserAttribute -from pipeline.memory import MemoryServiceRequestModel -from service.memory_service_bailian import MemoryServiceBailian +from memory_scope.utils.logger import Logger +from memory_scope.constants.common_constants import NEW_USER_PROFILE, MODIFIED_MEMORIES, RELATED_MEMORIES +from memory_scope.enumeration.memory_method_enum import MemoryMethodEnum +from memory_scope.node.memory_node import MemoryNode +from memory_scope.node.user_attribute import UserAttribute +from memory_scope.pipeline.memory import MemoryServiceRequestModel +from memory_scope.pipeline.memory_service import MemoryService """ 任务：随机生成一个用户的画像，随机种子0，并根据用户的画像虚拟一段用户和AI的对话。 @@ -139,10 +139,8 @@ for i, msg in enumerate(messages3): def summary_short(messages): - messages_pick_n = len(messages) request: MemoryServiceRequestModel = MemoryServiceRequestModel( messages=messages, - messages_pick_n=messages_pick_n, memory_id=memory_id, workspace_id=workspace_id, api_key=api_key, @@ -171,7 +169,6 @@ def summary_short(messages): def summary_long(): request: MemoryServiceRequestModel = MemoryServiceRequestModel( messages=[], - messages_pick_n=0, memory_id=memory_id, workspace_id=workspace_id, api_key=api_key, @@ -204,7 +201,6 @@ def summary_long(): def retrieve(messages): request: MemoryServiceRequestModel = MemoryServiceRequestModel( messages=messages, - messages_pick_n=1, memory_id=memory_id, workspace_id=workspace_id, api_key=api_key,