From 8157ca43f005b39c467dd3819050d4e40e743fb1 Mon Sep 17 00:00:00 2001 From: "jinli.yl" Date: Mon, 2 Sep 2024 14:24:10 +0800 Subject: [PATCH] update en readme * update en readme * feat: Resolve conflict, auto committed by CodeFlow --- README.md | 57 ++++++++++++++++++++--------- README_ZH.md | 16 ++++---- examples/advance/replacement.yaml | 4 +- examples/api/simple_usages_cn.ipynb | 12 ++---- memoryscope/__init__.py | 2 +- setup.py | 18 ++++++--- 6 files changed, 68 insertions(+), 41 deletions(-) diff --git a/README.md b/README.md index 4586e9f0..30000355 100644 --- a/README.md +++ b/README.md @@ -2,26 +2,37 @@ English | [**中文**](./README_ZH.md) # MemoryScope -![MemoryScope Logo](./docs/images/logo_1.png) +![MemoryScope Logo](./docs/images/logo.png) Equip your LLM chatbot with a powerful and flexible long term memory system. +[![](https://img.shields.io/badge/python-3.10+-blue)](https://pypi.org/project/memoryscope/) +[![](https://img.shields.io/badge/pypi-v0.1.1-blue?logo=pypi)](https://pypi.org/project/memoryscope/) +[![](https://img.shields.io/badge/Docs-English%7C%E4%B8%AD%E6%96%87-blue?logo=markdown)](https://modelscope.github.io/memoryscope/#welcome-to-memoryscope-tutorial-hub) +[![](https://img.shields.io/badge/Docs-API_Reference-blue?logo=markdown)](https://modelscope.github.io/memoryscope/) +[![](https://img.shields.io/badge/license-Apache--2.0-black)](./LICENSE) +[![](https://img.shields.io/badge/Contribute-Welcome-green)](https://modelscope.github.io/memoryscope/tutorial/contribute.html) + ---- ## News -- **[2024-07-29]** We release MemoryScope v0.1.0.2 now, which is also available in [PyPI](https://pypi.org/simple)! +- **[2024-09-06]** We release MemoryScope v0.1.1 now, which is also available in [PyPI](https://pypi.org/simple/memoryscope/)! ---- ## What is MemoryScope？ +MemoryScope provides LLM chatbots with powerful and flexible long-term memory capabilities, offering a framework for building such abilities. +It can be applied to scenarios like personal assistants and emotional companions, continuously learning through long-term memory to remember users' basic information as well as various habits and preferences. +This allows users to gradually experience a sense of "understanding" when using the LLM. -MemoryScope is a powerful and flexible long term memory system for LLM chatbots. It consists -of a memory database and three customizable system operations, which can be flexibly combined to provide -robust long term memory services for your LLM chatbot. +![Framework](./docs/images/framework.png) -💾 Memory Database: -- MemoryScope comes with an *ElasticSearch (ES)* vector database to store all the -memory pieces recorded in the system. +### Framework + +💾 Memory Database: MemoryScope is equipped with a vector database (default is *ElasticSearch*) to store all memory fragments recorded in the system. + +🔧 Worker Library: MemoryScope atomizes the capabilities of long-term memory into individual workers, including over 20 workers for tasks such as query information filtering, observation extraction, and insight updating. + +🛠️ Operation Library: Based on the worker pipeline, it constructs the operations for memory services, realizing key capabilities such as memory retrieval and memory consolidation. -🛠️ System operations: - Memory Retrieval: Upon arrival of a user query, this operation returns the semantically related memory pieces and/or those from the corresponding time if the query involves reference to time. - Memory Consolidation: This operation takes in a batch of user queries and returns important user information @@ -30,9 +41,11 @@ extracted from the queries as consolidated *observations* to be stored in the me to form and update *insights*. Then, memory re-consolidation is performed to ensure contradictions and repetitions among memory pieces are properly handled. -### Framework -![MemoryScope Logo](./docs/images/framework.png) +⚙️ Best Practices: + +- Based on the core capabilities of long-term memory, MemoryScope has implemented a dialogue interface (API) with long-term memory and a command-line dialogue practice (CLI) with long-term memory. +- MemoryScope combines currently popular agent frameworks (AutoGen, AgentScope) to provide best practices. ### Main Features @@ -52,19 +65,29 @@ from the aggregation of similarly-themed *observations*. - The system is time sensitive when performing both Memory Retrieval and Memory Consolidation. Therefore, it can retrieve accurate relevant information when the query involves reference to time. +---- + +## 💼 Supported Model API + +| Backend | Task | Some Supported Models | +|-------------------|------------|------------------------------------------------------------------------| +| openai_backend | Generation | gpt-4o, gpt-4o-mini, gpt-4, gpt-3.5-turbo | +| | Embedding | text-embedding-ada-002, text-embedding-3-large, text-embedding-3-small | +| dashscope_backend | Generation | qwen-max, qwen-plus, qwen-plus, qwen2-72b-instruct | +| | Embedding | text-embedding-v1, text-embedding-v2 | +| | Reranker | gte-rerank | + +In the future, we will support more model interfaces and local deployment of LLM and embedding services. ## 🚀 Installation -For installation, please refer to [Installation.md](docs/Installation.md). +For installation, please refer to [Installation.md](docs/installation.md). -### One-key Demo Run - -Run `sudo docker run -it --rm --net=host memoryscope/memoryscope` to launch memoryscope cli demo. ## Example Usages - [Simple Usages (Quick Start)](./examples/api/simple_usages_en.ipynb) -- [CLI with a MemoryScope Chatbot](./examples/cli/dash_cli_cn1.sh) -- [Advanced Customization](./examples/api/advanced_customization_en.ipynb) +- [CLI with a MemoryScope Chatbot](./examples/cli/README.md) +- [Advanced Customization](./examples/advance/custom_operator.md) ## 💡 Contribute diff --git a/README_ZH.md b/README_ZH.md index 009af6d5..42373be2 100644 --- a/README_ZH.md +++ b/README_ZH.md @@ -8,7 +8,7 @@ [![](https://img.shields.io/badge/python-3.10+-blue)](https://pypi.org/project/memoryscope/) [![](https://img.shields.io/badge/pypi-v0.1.1-blue?logo=pypi)](https://pypi.org/project/memoryscope/) -[![](https://img.shields.io/badge/Docs-English%7C%E4%B8%AD%E6%96%87-blue?logo=markdown)](https://modelscope.github.io/memoryscope/#welcome-to-agentscope-tutorial-hub) +[![](https://img.shields.io/badge/Docs-English%7C%E4%B8%AD%E6%96%87-blue?logo=markdown)](https://modelscope.github.io/memoryscope/#welcome-to-memoryscope-tutorial-hub) [![](https://img.shields.io/badge/Docs-API_Reference-blue?logo=markdown)](https://modelscope.github.io/memoryscope/) [![](https://img.shields.io/badge/license-Apache--2.0-black)](./LICENSE) [![](https://img.shields.io/badge/Contribute-Welcome-green)](https://modelscope.github.io/memoryscope/tutorial/contribute.html) @@ -21,7 +21,7 @@ ---- ## 新闻 -- **[2024-09-02]** 我们现在发布了 MemoryScope v0.1.1，该版本也可以在 [PyPI](https://pypi.org/simple/memoryscope/) 上获取！ +- **[2024-09-06]** 我们现在发布了 MemoryScope v0.1.1，该版本也可以在 [PyPI](https://pypi.org/simple/memoryscope/) 上获取！ ---- ## 什么是MemoryScope？ @@ -38,15 +38,15 @@ MemoryScope可以用于个人助理、情感陪伴等记忆场景，通过长期 🛠️ 核心Op库: 并基于worker的pipeline构建了memory服务的核心operation，实现了记忆检索，记忆巩固等核心能力。 -1. 记忆检索：当用户输入对话，此操作返回语义相关的记忆片段。如果输入对话包含对时间的指涉，则同时返回相应时间中的记忆片段。 -2. 记忆巩固：此操作接收一批用户的输入对话，并从对话中提取重要的用户信息，将其作为 *observation* 形式的记忆片段存储在记忆数据库中。 -3. 反思与再巩固：每隔一段时间，此操作对新记录的 *observations* 进行反思，以形成和更新 *insight* +- 记忆检索：当用户输入对话，此操作返回语义相关的记忆片段。如果输入对话包含对时间的指涉，则同时返回相应时间中的记忆片段。 +- 记忆巩固：此操作接收一批用户的输入对话，并从对话中提取重要的用户信息，将其作为 *observation* 形式的记忆片段存储在记忆数据库中。 +- 反思与再巩固：每隔一段时间，此操作对新记录的 *observations* 进行反思，以形成和更新 *insight* 形式的记忆片段。然后执行记忆再巩固，以确保记忆片段之间的矛盾和重复得到妥善处理。 ⚙️ 最佳实践: -1. MemoryScope在构建了长期记忆核心能力的基础上，实现了带长期记忆的对话接口(API)和带长期记忆的命令行对话实践(CLI)。 -2. MemoryScope结合了目前流行的Agent框架（AutoGen、AgentScope），给出了最佳实践。 +- MemoryScope在构建了长期记忆核心能力的基础上，实现了带长期记忆的对话接口(API)和带长期记忆的命令行对话实践(CLI)。 +- MemoryScope结合了目前流行的Agent框架（AutoGen、AgentScope），给出了最佳实践。 ### 🤝主要特点 @@ -89,7 +89,9 @@ MemoryScope可以用于个人助理、情感陪伴等记忆场景，通过长期 ### Docker方式一键运行Demo + 完整的安装方法请参考[安装指南](docs/installation_zh.md)。 diff --git a/examples/advance/replacement.yaml b/examples/advance/replacement.yaml index 66871336..4ef847f0 100644 --- a/examples/advance/replacement.yaml +++ b/examples/advance/replacement.yaml @@ -29,7 +29,7 @@ memory_service: retrieve_memory: class: core.operation.frontend_operation - workflow: set_query_meow,[extract_time|retrieve_obs_ins,semantic_rank],fuse_rerank + workflow: rewrite_query,[extract_time|retrieve_obs_ins,semantic_rank],fuse_rerank description: "retrieve long-term memory" list_memory: @@ -74,7 +74,7 @@ worker: class: core.worker.frontend.read_message_worker set_query: class: core.worker.frontend.set_query_worker - set_query_meow: + rewrite_query: class: contrib.example_query_worker generation_model: generation_model retrieve_obs_ins: diff --git a/examples/api/simple_usages_cn.ipynb b/examples/api/simple_usages_cn.ipynb index c590bcc4..79ef7580 100644 --- a/examples/api/simple_usages_cn.ipynb +++ b/examples/api/simple_usages_cn.ipynb @@ -26,8 +26,6 @@ "metadata": {}, "outputs": [], "source": [ - "import sys\n", - "sys.path.append(\".\")\n", "from memoryscope import MemoryScope, Arguments\n", "arguments = Arguments(\n", " language=\"cn\",\n", @@ -35,14 +33,12 @@ " assistant_name=\"AI\",\n", " memory_chat_class=\"api_memory_chat\",\n", " generation_backend=\"dashscope_generation\",\n", - " generation_model=\"qwen2-72b-instruct\",\n", + " generation_model=\"qwen-max\",\n", " embedding_backend=\"dashscope_embedding\",\n", " embedding_model=\"text-embedding-v2\",\n", " rank_backend=\"dashscope_rank\",\n", " rank_model=\"gte-rerank\",\n", - " enable_ranker=True,\n", - " worker_params={\"get_reflection_subject\": {\"reflect_num_questions\": 3}}\n", - ")\n", + " enable_ranker=True)\n", "\n", "ms = MemoryScope(arguments=arguments)\n" ] @@ -465,9 +461,9 @@ "metadata": {}, "source": [ "## 更多用法\n", - "我们建议读者参考[进阶自定义用法](./examples/api/advanced_customization_cn.ipynb)来对MemoryScope系统进行各种自定义设置。您还可以通过自定义**workflow**和对应的**worker**来创建或定制满足您特定需求的**operation**。\n", + "我们建议读者参考[进阶自定义用法](../advance/custom_operator.md)来对MemoryScope系统进行各种自定义设置。您还可以通过自定义**workflow**和对应的**worker**来创建或定制满足您特定需求的**operation**。\n", "\n", - "此外，您还可以尝试使用[在命令行与MemoryScope聊天机器人交互](../cli/README.md)。我们在这里实现了始终在后台异步运行**记忆巩固**和**反思与再巩固**这两个操作，从而使得它们不会增加聊天的响应时间。" + "此外，您还可以尝试使用[在命令行与MemoryScope聊天机器人交互](../cli/README_ZH.md)。我们在这里实现了始终在后台异步运行**记忆巩固**和**反思与再巩固**这两个操作，从而使得它们不会增加聊天的响应时间。" ] } ], diff --git a/memoryscope/__init__.py b/memoryscope/__init__.py index 61f9f76d..43596816 100644 --- a/memoryscope/__init__.py +++ b/memoryscope/__init__.py @@ -1,5 +1,5 @@ """ Version of MemoryScope.""" -__version__ = "0.1.1.1" +__version__ = "0.1.0.9" import fire from memoryscope.core.config.arguments import Arguments # noqa: F401 diff --git a/setup.py b/setup.py index 6e37bfd8..5a64f173 100644 --- a/setup.py +++ b/setup.py @@ -7,6 +7,8 @@ python setup.py sdist bdist_wheel # 3. finally, upload twine upload dist/* + +rm -rf dist build && python setup.py sdist bdist_wheel && twine upload dist/* """ import setuptools, glob, os @@ -14,6 +16,7 @@ import setuptools, glob, os with open("README.md", "r", encoding="utf-8") as fh: long_description = fh.read() + def _process_requirements(): packages = open('requirements.txt').read().strip().split('\n') requires = [] @@ -25,6 +28,7 @@ def _process_requirements(): requires.append(pkg) return requires + def package_files(directory): paths = [] for (path, directories, filenames) in os.walk(directory): @@ -33,8 +37,8 @@ def package_files(directory): paths.append(os.path.join('..', path, filename)) return paths -extra_files = package_files('memoryscope') +extra_files = package_files('memoryscope') authors = [ {"name": "Li Yu", "email": "jinli.yl@alibaba-inc.com"}, @@ -47,10 +51,12 @@ authors = [ setuptools.setup( name="memoryscope", - version="0.1.1.1", + version="0.1.0.9", author=', '.join([author['name'] for author in authors]), author_email=', '.join([author['email'] for author in authors]), - description="MemoryScope is a powerful and flexible long term memory system for LLM chatbots. It consists of a memory database and three customizable system operations, which can be flexibly combined to provide robust long term memory services for your LLM chatbot.", + description="MemoryScope is a powerful and flexible long term memory system for LLM chatbots. It consists of a " + "memory database and three customizable system operations, which can be flexibly combined to provide " + "robust long term memory services for your LLM chatbot.", long_description=long_description, long_description_content_type="text/markdown", url="https://github.com/modelscope/memoryscope", @@ -59,16 +65,16 @@ setuptools.setup( }, classifiers=[ "Programming Language :: Python :: 3", - "License :: Apache License", + "License :: OSI Approved :: Apache Software License", "Operating System :: OS Independent", ], package_dir={"": "."}, package_data={"": extra_files}, include_package_data=True, - entry_points = { + entry_points={ 'console_scripts': ['memoryscope=memoryscope:cli'], }, packages=setuptools.find_packages(where="."), python_requires=">=3.10", install_requires=_process_requirements(), -) \ No newline at end of file +)