utils/call_llm.py

from langchain_ollama import ChatOllama
from langchain_core.messages import SystemMessage

def llm(
    model_name: str,
    messages: list[dict],
    temperature: float = 0.1,
    max_tokens: int = 1024,
    **kwargs,
) -> str:
    """
    Calls the Ollama Chat model and returns the generated response content.
    """
    try:
        print("[NODE] ----- Calling Ollama Chat -----")

        # Construct the prompt with explicit separation for SystemMessage
        prompt_parts = []
        for message in messages:
            if isinstance(message, dict):
                prompt_parts.append(f"{message['role'].capitalize()}: {message['content']}")
            elif isinstance(message, SystemMessage):
                prompt_parts.append(f"System: {message.content}")
            else:
                prompt_parts.append(message)

        prompt = "\n\n".join(prompt_parts)  # Add extra separation for clarity

        print(f"Constructed Prompt:\n{prompt}")

        chat = ChatOllama(
            model=model_name,
            temperature=temperature,
            max_tokens=max_tokens,
            **kwargs
        )

        response = chat.invoke(prompt)
        print("----- Ollama Chat response -----")
        print(response.content)

        if not response or not response.content:
            print("No content returned from the Ollama Chat model.")
            return "No content generated."

        return response.content

    except Exception as e:
        print("An error occurred while calling the Ollama Chat model: %s", str(e))
        raise