utils.py

# from SoundScribe.speakerID import find_user
import datetime
import requests
import torch
import json
import cv2
import os


API_URL = 'https://bceb7f41087d-7754001953109090881.ngrok-free.app/'


def get_time():
    return datetime.datetime.now().strftime('%a %d %b %Y %I:%M %p')


def load_chat():
    full_history = []
    sorted_list = []
    prev_id = ""
    with open('./database/chat_history.jsonl', 'r') as history:
        for line in history:
            chat_message = json.loads(line)
            id = chat_message['ID']
            message = chat_message['message']
            if id != prev_id:
                full_history.append(f"<|im_start|>{id}\n{message}<|im_end|>\n")
            else:
                full_history[-1] += message+"\n"

            prev_id = id

    for chat in full_history:
        if chat.startswith("<|im_start|>CRYSTAL") or chat.startswith("<|im_start|>Helper"):
            sorted_list[-1] += "\n"+chat
        else:
            sorted_list.append(chat)

    return sorted_list


def record_chat(role, message):
    new_message = {
        "ID": role,
        "message": message
    }

    with open('./database/chat_history.jsonl', 'a') as history:
        history.write(json.dumps(new_message) + '\n')


def check_api_usage():
    USE_CLOUD_API = False
    if os.path.isdir("models"):
        if requests.get(API_URL).ok:
            choice = input(
                "CRYSTAL CLOUD API HAS BEEN DETECTED.\n"
                "Would you like to:\n"
                "\t1. Use Cloud API Computing\n"
                "\t2. Use On-Device Calculations\n"
                "Enter your choice (1/2): ")
            USE_CLOUD_API = choice == "1"
            if USE_CLOUD_API:
                print("RUNNING ON CLOUD")
            else:
                print("RUNNING LOCALLY")

        else:
            print("CRYSTAL Cloud API not reachable.")
    else:
        raise RuntimeError(
            "Unauthorized access! This action will be reported immediately!")

    return USE_CLOUD_API


def perceptrix(prompt):
    url = API_URL+"perceptrix"

    payload = {'prompt': prompt}
    headers = {'Content-Type': 'application/json'}
    response = requests.post(url, json=payload, headers=headers)

    return response.json()["message"]


def robotix(prompt):
    url = API_URL+"robotix"

    payload = {'prompt': prompt}
    headers = {'Content-Type': 'application/json'}
    response = requests.post(url, json=payload, headers=headers)

    return response.json()["message"]


def identify_objects_from_text(prompt):
    url = API_URL+"identify_objects_from_text"

    payload = {'prompt': prompt}
    headers = {'Content-Type': 'application/json'}
    response = requests.post(url, json=payload, headers=headers)

    return response.json()["message"]


def search_keyword(prompt):
    url = API_URL+"search_keyword"

    payload = {'prompt': prompt}
    headers = {'Content-Type': 'application/json'}
    response = requests.post(url, json=payload, headers=headers)

    return response.json()["message"]


def answer_question(prompt, frame):
    url = API_URL+"vqa"
    frame = cv2.imread(frame)
    
    _, image_data = cv2.imencode('.jpg', frame)

    image = image_data.tolist()

    payload = {'image': image,
               'prompt': prompt}
    headers = {'Content-Type': 'application/json'}
    response = requests.post(url, json=payload, headers=headers)

    return response.json()["message"]


def find_object_description(prompt, frame):
    url = API_URL+"object_description"
    if type(frame) == str:
        frame = cv2.imread(frame)
    _, image_data = cv2.imencode('.jpg', frame)
    image = image_data.tolist()

    payload = {'image': image,
               'prompt': prompt}
    headers = {'Content-Type': 'application/json'}
    response = requests.post(url, json=payload, headers=headers)

    return response.json()["message"]


def locate_object(prompt, frame):
    url = API_URL+"locate_object"
    if type(frame) == str:
        frame = cv2.imread(frame)
    _, image_data = cv2.imencode('.jpg', frame)
    image = image_data.tolist()

    payload = {'image': image,
               'prompt': prompt}
    headers = {'Content-Type': 'application/json'}
    response = requests.post(url, json=payload, headers=headers)

    return response.json()["annotated_image"], response.json()["message"]


def setup_device():
    if torch.backends.mps.is_available():
        device = torch.device("mps")
    elif torch.cuda.is_available():
        device = torch.device("cuda")
    else:
        device = torch.device("cpu")
    return device


def transcribe(audio):
    url = API_URL + "transcribe"
    with open(audio, 'rb') as audio_file:
        files = {'audio': (audio, audio_file)}
        response = requests.post(url, files=files)

    transcription = response.json()["message"]
    print(transcription)
    # user = find_user("database/recording.wav")
    user = "Vatsal"
    return transcription, user