App/Chat/utils/Summarize.py

import aiohttp
import asyncio, pprint
import google.generativeai as palm
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain import PromptTemplate
import os

# from poe_api_wrapper import PoeApi
import pprint

# client = PoeApi("sXvCnfYy8CHnXNTRlxhmVg==")
bot = "Assistant"
CHAT_CODE = ""

PALM_API = ""
API_KEY = os.environ.get("PALM_API", PALM_API)
palm.configure(api_key=API_KEY)

GroqAPIKey = os.environ.get("Groq_API", "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")
text_splitter = RecursiveCharacterTextSplitter(
    separators=["\n\n", "\n", "."],
    chunk_size=1500,
    length_function=len,
    chunk_overlap=100,
)


map_prompt = """
Write a verbose summary  like a masters student of the following:
"{text}"
CONCISE SUMMARY:
"""


combine_prompt = """
Write a concise summary of the following text delimited by triple backquotes.
Return your response in a detailed verbose paragraph which covers the text. Make it as insightful to the reader as possible, write like a masters student.

```{text}```

SUMMARY:
"""


def count_tokens(text):
    return palm.count_message_tokens(prompt=text)["token_count"]


async def PalmTextModel(text, candidates=1):
    # url = f"https://generativelanguage.googleapis.com/v1beta/models/gemini-pro:generateContent?key={API_KEY}"
    url = "https://api.groq.com/openai/v1/chat/completions"
    headers = {
        "Content-Type": "application/json",
        "Authorization": f"Bearer {GroqAPIKey}",
    }
    payload = {"messages": [{"role": "user", "content": text}], "model": "mixtral-8x7b-32768"}
    # payload = {
    #     "contents": [
    #         {
    #             "role": "user",
    #             "parts": [
    #                 {
    #                     "text": text
    #                 }
    #             ]
    #         }
    #     ],
    #     "generationConfig": {
    #         "temperature": 0.9,
    #         "topK": 1,
    #         "topP": 1,
    #         "maxOutputTokens": 2048,
    #         "stopSequences": []
    #     },
    #     "safetySettings": [
    #         {
    #             "category": "HARM_CATEGORY_HARASSMENT",
    #             "threshold": "BLOCK_ONLY_HIGH"
    #         },
    #         {
    #             "category": "HARM_CATEGORY_HATE_SPEECH",
    #             "threshold": "BLOCK_ONLY_HIGH"
    #         },
    #         {
    #             "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
    #             "threshold": "BLOCK_ONLY_HIGH"
    #         },
    #         {
    #             "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
    #             "threshold": "BLOCK_ONLY_HIGH"
    #         }
    #     ]
    # }

    async with aiohttp.ClientSession() as session:
        async with session.post(url, json=payload, headers=headers) as response:
            if response.status == 200:
                result = await response.json()
                # print(result)
                # if candidates > 1:
                #     temp = [
                #         candidate["content"]["parts"][0]["text"]
                #         for candidate in result["candidates"]
                #     ]
                #     return temp
                temp = result["choices"][0]["message"]["content"]
                # temp="hello"
                # print('lo ')
                return temp
            else:
                print(f"Error: {response.status}\n{await response.text()}")


# async def PalmTextModel(message):
#     global CHAT_CODE
#     if CHAT_CODE == "":
#         for chunk in client.send_message(bot, message):
#             pass
#         CHAT_CODE = chunk["chatCode"]
#     else:
#         for chunk in client.send_message(bot, message, chatCode=CHAT_CODE):
#             pass

#     return chunk["text"]


async def Summarizer(essay):
    docs = text_splitter.create_documents([essay])

    # for 1 large document
    if len(docs) == 1:
        tasks = [
            PalmTextModel(combine_prompt.format(text=doc.page_content)) for doc in docs
        ]
        # Gather and execute the tasks concurrently
        responses = await asyncio.gather(*tasks)
        ans = " ".join(responses)
        return ans

    tasks = [PalmTextModel(map_prompt.format(text=doc.page_content)) for doc in docs]
    # Gather and execute the tasks concurrently
    chunked_tasks = [tasks[i:i+20] for i in range(0, len(tasks), 20)]
    responses = []
    for chunk in chunked_tasks:
        chunk_responses = await asyncio.gather(*chunk)
        responses.extend(chunk_responses)
    main = " ".join(responses)
    ans = await PalmTextModel(combine_prompt.format(text=main))
    return ans

# asyncio.run(Summarizer("This is a test of the summarizer"))