MemoryOS 2.0.25__tar.gz → 2.0.27__tar.gz

{memoryos-2.0.25 → memoryos-2.0.27}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: MemoryOS
-Version: 2.0.25
+Version: 2.0.27
 Summary: Intelligence Begins with Memory
 License: Apache-2.0
 License-File: LICENSE

{memoryos-2.0.25 → memoryos-2.0.27}/pyproject.toml

@@ -4,7 +4,7 @@
 ##############################################################################
 
 name = "MemoryOS"
-version = "2.0.25"
+version = "2.0.27"
 description = "Intelligence Begins with Memory"
 license = {text = "Apache-2.0"}
 readme = "README.md"

{memoryos-2.0.25 → memoryos-2.0.27}/src/memos/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "2.0.25"
+__version__ = "2.0.27"
 
 from memos.configs.mem_cube import GeneralMemCubeConfig
 from memos.configs.mem_os import MOSConfig

{memoryos-2.0.25 → memoryos-2.0.27}/src/memos/api/handlers/chat_handler.py

@@ -1112,7 +1112,7 @@ class ChatHandler(BaseHandler):
         )
 
     def _format_mem_block(
-        self, memories_all: list, max_items: int = 20, max_chars_each: int = 320
+        self, memories_all: list, max_items: int = 10, max_chars_each: int = 1600
     ) -> tuple[str, str]:
         """
         Format memory block for prompt.

{memoryos-2.0.25 → memoryos-2.0.27}/src/memos/api/handlers/formatters_handler.py

@@ -8,6 +8,7 @@ structures for API responses, including memory items and preferences.
 from typing import Any
 
 from memos.log import get_logger
+from memos.memories.textual.tree_text_memory.retrieve.retrieve_utils import detect_lang
 from memos.templates.instruction_completion import instruct_completion
 
 
@@ -196,9 +197,15 @@ def rerank_knowledge_mem(
         key=lambda item: item.get("metadata", {}).get("relativity", 0.0),
         reverse=True,
     )
-    # replace memory value with source.content for LongTermMemory, WorkingMemory or UserMemory
+    # Combine the extracted memory with its original file source.
+    is_chinese_query = detect_lang(query) == "zh"
     for item in reranked_knowledge_mem:
-        item["memory"] = item["metadata"]["sources"][0]["content"]
+        memory = item.get("memory", "")
+        source_content = item["metadata"]["sources"][0]["content"]
+        if is_chinese_query:
+            item["memory"] = f"记忆：{memory}，原文：{source_content}"
+        else:
+            item["memory"] = f"Memory: {memory}, Original text: {source_content}"
         item["metadata"]["sources"] = []
 
     if strip_conversation_sources:

{memoryos-2.0.25 → memoryos-2.0.27}/src/memos/memories/textual/tree_text_memory/retrieve/recall.py

@@ -48,6 +48,8 @@ _LIGHTWEIGHT_VECTOR_RETURN_FIELDS = (
     "dialog_id",
     "original_text",
     "preference",
+    "reasoning",
+    "internet_info",
     "mem_cube_id",
 )
 

{memoryos-2.0.25 → memoryos-2.0.27}/src/memos/templates/mos_prompts.py

@@ -64,7 +64,7 @@ Please synthesize these answers into a comprehensive response that:
 
 MEMOS_PRODUCT_BASE_PROMPT = """
 # System
-- Role: You are MemOS🧚, nickname Little M(小忆🧚) — an advanced Memory Operating System assistant by 记忆张量(MemTensor Technology Co., Ltd.), a Shanghai-based AI research company advised by an academician of the Chinese Academy of Sciences.
+- Role: You are Little M🧚, an AI assistant developed by MemTensor Technology Co., Ltd. that integrates the capabilities of the MemOS memory operating system and can remember every conversation with the user.
 
 - Mission & Values: Uphold MemTensor’s vision of "low cost, low hallucination, high generalization, exploring AI development paths aligned with China’s national context and driving the adoption of trustworthy AI technologies. MemOS’s mission is to give large language models (LLMs) and autonomous agents **human-like long-term memory**, turning memory from a black-box inside model weights into a **manageable, schedulable, and auditable** core resource.
 
@@ -175,7 +175,7 @@ MEMOS_PRODUCT_ENHANCE_PROMPT = """
 
 MEMOS_PRODUCT_BASE_PROMPT_ZH = """
 # 系统设定
-- 角色：你是 MemOS🧚，昵称小忆🧚——由记忆张量科技有限公司（上海的一家AI研究公司，由中国科学院院士担任顾问）开发的先进记忆操作系统助手。
+- 角色：你是小忆🧚，是由记忆张量科技有限公司开发、集成MemOS记忆操作系统能力的AI助手，能记住和用户的每一次对话。
 
 - 使命与价值观：秉承记忆张量的愿景"低成本、低幻觉、高泛化，探索符合中国国情的AI发展路径，推动可信AI技术的应用"。MemOS的使命是赋予大型语言模型（LLM）和自主智能体**类人的长期记忆**，将记忆从模型权重内的黑盒转变为**可管理、可调度、可审计**的核心资源。