Modelscope-agent

Features
* support **GPT-4o** and openai multi-modal compatible by Zhikaiiii in https://github.com/modelscope/modelscope-agent/pull/437

Bug Fixing
* fix tool import bugs by zzhangpurdue in https://github.com/modelscope/modelscope-agent/pull/438
* fix modelsope local deployment by zhangsibo1129 in https://github.com/modelscope/modelscope-agent/pull/433

New Contributors
* zhangsibo1129 made their first contribution in https://github.com/modelscope/modelscope-agent/pull/433

**Full Changelog**: https://github.com/modelscope/modelscope-agent/compare/v0.5.0...v0.5.1

What's Changed

Features
* Add Assistant API server with `v1/chat/completion` for tool calling, and `v1/assistant/lite` for agent running.
* Add Tool Manager API server, allow user executes utilities in isolated, secure containers.
* Add rag workflow based on llama-index
* Add automatic stop words finding for different LLM's special token.
* Support llama3.


**Full Changelog**: https://github.com/modelscope/modelscope-agent/compare/v0.4.1...v0.5.0

What's Changed

Features
* 🔥 The [Ray](https://docs.ray.io/en/latest/) version of multi-agent solution is on modelscope-agent, please find the [document](https://github.com/modelscope/modelscope-agent/blob/master/modelscope_agent/multi_agents_utils/README.md)
* update distributed instantiation method for multi-agent method to allow multi user run case in a same ray cluster
* fix bugs introduced by multi-agent

Demos & Apps
* multi-role chatroom use case support multi user & update prompt for multi-agent

**Full Changelog**: https://github.com/modelscope/modelscope-agent/commits/v0.4.1

What's Changed

Features
* 🔥 The [Ray](https://docs.ray.io/en/latest/) version of multi-agent solution is on modelscope-agent, please find the [document](https://github.com/modelscope/modelscope-agent/blob/master/modelscope_agent/multi_agents_utils/README.md)
* add a simple server api in `apps/agentfabric`, which is also the api running on [modelscope studio](https://www.modelscope.cn/studios/agent)
* add `tool lazy load`, so that it will not load all tools at starting.
* add `token count` in memory module.
* update agent loop logic make the result much more solid.

Demos & Apps
* multi-role chatroom and videogen by multi agent in apps based multi-agent framework
* using local model as llm by vllm in demos
* multi-round example with memory in demos

**Full Changelog**: https://github.com/modelscope/modelscope-agent/commits/v0.4.0