New Container: vLLM

What is the new container you'd like to have?

A vLLM object container would serve testing and usage better.

Why not just use a generic container for this?

The Ollama object is not maintained against Ollama, which causes several issues:

1. Performance: Ollama performs poorly with batch sizes of 50+ requests.
2. Functionality: The new "search functionality" in Ollama requires ollama keys, which are not supported in the current object.
3. Coverage & Quality: Ollama’s coverage and performance are lower compared to vLLM.

Why vLLM?

The vLLM object allows development and testing in a highly configurable, programmatic way, improving the overall testing experience.

Other references

以现有的 Ollama 对象和仓库当前的容器对象组织方式作为起点进行审查。当提供一个适合测试的、可配置且可通过编程方式使用的 vLLM 对象，并解决 issue 中描述的覆盖率和性能问题时，即视为完成。