import os
import pandas as pd
from dotenv import load_dotenv
load_dotenv()
import openai

# # set OPENAI_API_KEY environment variable from .env file
# openai.api_key = os.getenv("OPENAI_API_KEY")

# # read in llm-data-cleaner/prompts/gpt4-system-message.txt file into variable system_message
# system_message = open('../prompts/gpt4-system-message.txt', 'r').read()

class OpenAIChatCompletions:
    def __init__(self, model="gpt-4", system_message=None):
        self.model = model
        self.system_message = system_message


    # function to input args such as model, prompt, etc. and return completion
    def openai_chat_completion(self, prompt, n_shot=None):
        messages = [{"role": "system", "content": self.system_message}] if self.system_message else []
        
        # add n_shot number of samples to messages list ... if n_shot is None, then only system_message and prompt will be added to messages list
        if n_shot is not None:
            messages = self._add_samples(messages, n_samples=n_shot)

        messages.append({"role": "user", "content": prompt})

        # set up the API request parameters for OpenAI
        chat_request_kwargs = dict(
            model=self.model,
            messages=messages,
        )

        # make the API request to OpenAI
        response = openai.ChatCompletion.create(**chat_request_kwargs)

        # return only the completion text
        # return response['choices'][0]['message']['content']
        # return response
        return response


    # function to use test data to predict completions
    def predict_jsonl(
        self,
        # path_or_buf='../data/cookies_train.jsonl',
        path_or_buf='~/data/cookies_train.jsonl',
        n_samples=None,
        n_shot=None
        ):
        
        jsonObj = pd.read_json(path_or_buf=path_or_buf, lines=True)
        if n_samples is not None:
            jsonObj = jsonObj.sample(n_samples, random_state=42)

        iter_range = range(len(jsonObj))
        prompts = [jsonObj.iloc[i]['prompt'] for i in iter_range]
        completions = [jsonObj.iloc[i]['completion'] for i in iter_range]
        predictions = [self.openai_chat_completion(prompt, n_shot=n_shot) for prompt in prompts]

        return prompts, completions, predictions


    # a method that adds prompt and completion samples to messages
    @staticmethod
    def _add_samples(messages, n_samples=None):
        if n_samples is None:
            return messages

        samples = OpenAIChatCompletions._sample_jsonl(n_samples=n_samples)
        for i in range(n_samples):
            messages.append({"role": "user", "content": samples.iloc[i]['prompt']})
            messages.append({"role": "assistant", "content": samples.iloc[i]['completion']})

        return messages


    # a method that samples n rows from a jsonl file, returning a pandas dataframe
    @staticmethod
    def _sample_jsonl(
        # path_or_buf='../data/cookies_train.jsonl',
        path_or_buf='~/data/cookies_train.jsonl',
        n_samples=5
        ):
        
        jsonObj = pd.read_json(path_or_buf=path_or_buf, lines=True)
        return jsonObj.sample(n_samples, random_state=42)