Make system content as variable

I am using keybert for a topic extraction task and I have the need of making the system content as a variable so I can change accordingly based on my task. Here below I represent what I mean.

Current implementation:

class OpenAI(BaseLLM):
     [...]
    def __init__(self,
                 client,
                 model: str = "gpt-3.5-turbo-instruct",
                 prompt: str = None,
                 generator_kwargs: Mapping[str, Any] = {},
                 delay_in_seconds: float = None,
                 exponential_backoff: bool = False,
                 chat: bool = False,
                 verbose: bool = False
                 ):
        self.client = client
        self.model = model

        if prompt is None:
            self.prompt = DEFAULT_CHAT_PROMPT if chat else DEFAULT_PROMPT
        else:
            self.prompt = prompt

        self.default_prompt_ = DEFAULT_CHAT_PROMPT if chat else DEFAULT_PROMPT
        self.delay_in_seconds = delay_in_seconds
        self.exponential_backoff = exponential_backoff
        self.chat = chat
        self.verbose = verbose

        self.generator_kwargs = generator_kwargs
        if self.generator_kwargs.get("model"):
            self.model = generator_kwargs.get("model")
        if self.generator_kwargs.get("prompt"):
            del self.generator_kwargs["prompt"]
        if not self.generator_kwargs.get("stop") and not chat:
            self.generator_kwargs["stop"] = "\n"

    def extract_keywords(self, documents: List[str], candidate_keywords: List[List[str]] = None):
        [..]
            # Use a chat model
            if self.chat:
                messages = [
                    {"role": "system", "content": "You are a helpful assistant."},
                    {"role": "user", "content": prompt}
                ]
       [..]

Proposed implementation:

class OpenAI(BaseLLM):
     [...]
    def __init__(self,
                 client,
                 model: str = "gpt-3.5-turbo-instruct",
                 prompt: str = None,
                 system_content: str = "You are a helpful assistant.", </strong>
                 generator_kwargs: Mapping[str, Any] = {},
                 delay_in_seconds: float = None,
                 exponential_backoff: bool = False,
                 chat: bool = False,
                 verbose: bool = False
                 ):
        self.client = client
        self.model = model

        if prompt is None:
            self.prompt = DEFAULT_CHAT_PROMPT if chat else DEFAULT_PROMPT
        else:
            self.prompt = prompt

        self.system_content = system_content
        self.default_prompt_ = DEFAULT_CHAT_PROMPT if chat else DEFAULT_PROMPT
        self.delay_in_seconds = delay_in_seconds
        self.exponential_backoff = exponential_backoff
        self.chat = chat
        self.verbose = verbose

        self.generator_kwargs = generator_kwargs
        if self.generator_kwargs.get("model"):
            self.model = generator_kwargs.get("model")
        if self.generator_kwargs.get("prompt"):
            del self.generator_kwargs["prompt"]
        if not self.generator_kwargs.get("stop") and not chat:
            self.generator_kwargs["stop"] = "\n"

    def extract_keywords(self, documents: List[str], candidate_keywords: List[List[str]] = None):
        [..]
            # Use a chat model
            if self.chat:
                messages = [
                    {"role": "system", "content": self.system_content},
                    {"role": "user", "content": prompt}
                ]
       [..]

It implies replacing the hard coded string "You are a helpful assistant." with a variable coming from init which default value is actually the existing content. Retrocompatibility is guaranteed.
This logic can be applied both on keybert/llm/_openai.py and keybert/llm/_litellm.py files.

Review keybert/llm/_openai.py and keybert/llm/_litellm.py, starting at each class's init and chat-message construction in extract_keywords. Confirm the system content is configurable while preserving the existing default and behavior when no value is supplied.