app/helpers/chat.py

from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from helpers.generate_embbedings import vector_store
from helpers.f5_model import f5_model

def make_prompt(history, prompt, context=None):
    formatted_history = ""

    if context:
        formatted_history += f"[CONTEXT] {context} [/CONTEXT]\n"

    for history_item in history:
        if history_item.from_ == 'user':
            formatted_history += f"[INST] {history_item.message} [/INST]\n"
        else:
            formatted_history += f"{history_item.message}\n"

    formatted_history += f"[INST] {prompt} [/INST]\n"

    return formatted_history

async def ask_question(question: str, history: list = [], project_id=None):
    """
    Generate a response using F5 model based on history and project-specific context.
    """
    try:
        context = ""
        if project_id is not None:
            context = vector_store.similarity_search(
                query=question, k=4, filter={"project_id": project_id}
            )
        
        prompt = make_prompt(history, question, context)

        async for chunk in f5_model.stream_response(prompt):
            yield chunk
    except Exception as e:
        raise RuntimeError(f"Error generating response: {str(e)}")