provider 为 openai 时处理有问题

[aizimuji]

  def check_litellm_providers(self, provider: str) -> bool:
      # Judge whether the provider is supported by LiteLLM, and if provider name should be added to the model name
      if provider in ['azure', 'azure_ai', 'anthropic', 'deepseek', 'sagemaker', 'bedrock', 'vertex_ai', 'vertex_ai_beta', 'palm', 'gemini', 'mistral', 'cloudflare', 'huggingface', 'replicate', 'together_ai', 'openrouter', 'baseten', 'nlp_cloud', 'petals', 'ollama', 'perplexity', 'groq', 'anyscale', 'watsonx', 'voyage', 'xinference', 'nvidia_nim']:
          # provider name should be added to the model name
          return True, 1
      elif provider in ['openai', 'cohere', 'ai21', 'deepinfra', 'ai21', 'alpha_alpha']:
          # provider name should not be added to the model name
          return True, 2
      elif provider in ['xai', 'qwen']:
          # append OpenAI as procall as OpenAI campatible API
          return True, 3
      else:
          return False, 0

这个函数, openai 应该也放在第一批处理, return True, 1

否则当直接传入模型名称时, 比如gpt-4o 会报错 LLM Provider NOT provided

除非model传入openai/gpt-4o, 这样重复太麻烦

[quentinzhang]

应该不需要强制加上openai，litellm默认会识别openai的模型，调用方式类似这样：
unionchat(provider="openai", model="gpt-4o", messages=[{"content": "你的开发者是谁？", "role": "user"}], stream=True)

我这边测试没有出现问题。