Language Version: 简体中文 | English
This document records the version update history of the project.
The format of the version number is MAJOR.MINOR.PATCH, and the version number increment rule is as follows:
- MAJOR version when you make incompatible API changes,
- MINOR version when you add functionality in a backwards compatible manner,
- PATCH version when you make backwards compatible bug fixes.
- For more details, please refer to Semantic Versioning 2.0.0
Init - Project initialization. Added - Newly added features. Changed - Changes to existing functionalities. Deprecated - Soon to be deprecated features. Removed - Features removed in this version. Fixed - Any bug fixes. Security - Patches and security improvements. Note - Additional remarks regarding the version.
- Added standard integration for the DeepSeek model in the LLM module.
- Added a new OpenAI general protocol wrapper class, OpenAIStyleLLM.
- Models using the OpenAI protocol can be configured directly.
- Added a new LangChain tool wrapper class, LangChainTool, with several example tools for search and execution.
- LangChain tools can be configured directly.
- Added Agent information collection capability in the monitor module.
- Added use cases.
- Supplemented documentation with a financial event analysis case study using PEER collaborative mode.
- Added several new documents for LLM components, tool components, and the Monitor module.
- Updated the new README.
- Some code optimizations and documentation updates.
- Added standard integration for Claude and Ollama LLM components
- Added new Qwen embedding module
- Added default agents for ReAct-Type and NL2API-Type
-
Added new use cases
- RAG-Type Agent Examples: Legal Consultation Agent
- ReAct-Type Agent Examples: Python Code Generation and Execution Agent
- Multi-Agent: Discussion Group Based on Multi-Turn Multi-Agent Mode
For more details, please refer to the use case section in the user documentation.
-
Some code optimizations and documentation updates.
- Introduced a new monitor module
- Data running in any agentUniverse can be collected and observed
- Added webserver post_fork functionality
- Provides multi-node process intervention capabilities after starting the webserver in agentUniverse
- Introduced SQLDB_WRAPPER wrapper class, offering typical database connection methods
- Through the SQLDB_WRAPPER wrapper class, you can conveniently connect to various databases and storage technologies including SQLServer, MySQL, Oracle, PostgreSQL, SQLite and others
- Added connection support for Milvus vector database component
For more usage of the above features, please pay attention to the agentUniverse guidebook.
- Flask is set as the default webserver startup method across all platforms, with gunicorn and gRPC capabilities disabled by default
- In the previous version, we found slight compatibility differences with gunicorn and gRPC across different operating systems. Thus, we have made Flask the primary startup method for all platforms. You can enable gunicorn and gRPC in the configuration as needed.
- Some aU dependencies were identified to have security vulnerabilities in third-party packages. For security reasons, we have upgraded their versions, with the main changes including:
- requests (^2.31.0 -> ^2.32.0)
- flask (^2.2 -> ^2.3.2)
- werkzeug (^2.2.2 -> ^3.0.3)
- langchain (0.0.352 -> 0.1.20)
- langchain-core (0.1.3 -> 0.1.52)
- langchain-community (no version lock -> 0.0.38)
- gunicorn (21.2.0 -> ^22.0.0)
- Jinja2 (no version lock -> ^3.1.4)
- tqdm (no version lock -> ^4.66.3) If your system has external access, we strongly recommend installing version v0.0.8 of agentUniverse to mitigate the security risks posed by these third-party packages. For more detailed information, you can visit https://security.snyk.io.
- Some code optimizations and documentation updates.
- LLM component supports multimodal parameter invocation.
- Added LLM integration methods for Qwen, WenXin, Kimi, Baichuan, etc.
- Added a multimodal example agent, see the invocation details in
sample_standard_app.app.test.test_multimodal_agent.MultimodalAgentTest. - Some code optimizations and documentation updates.
- Support for the GPT-4o model, with updates to related examples.
- Support for the RPC component gRPC, providing a standard method for service startup.
- Provide standard Docker images and K8S deployment solutions.
- Some code optimizations and documentation updates.
- The LLM component supports streaming calls.
- The Knowledge component adds an update definition.
- Fixed potential concurrency safety issues in the peer planner.
- Fixed the issue in version 0.0.4 of the PyPI package where the packaging method forced users to enter an AK upon startup.
- Some code optimizations and documentation updates.
- Add version management capability to the prompt.
- Fixed compatibility issues on Windows
- Due to compatibility issues of Gunicorn with Windows systems, automatically identify the kernel version to select the web startup method.
- Specified YAML reading as UTF-8 encoding method.
- [2024-05-08] Please be aware that the PyPI package version 0.0.4 includes the sample_standard_app example project by default. This will reference additional components from sample_standard_app at startup and require users to input an AK. If you are not using the corresponding components, you can bypass this restriction by using a mock AK. This issue has been fixed in version 0.0.5.
- The official release version of AgentUniverse has been initialized. Enjoy using it!
- Fixed an issue where associated dependencies were not being automatically installed when installing package versions.
- Project initialization commit. This framework is a large model multi-agent framework. Enjoy using it!