app.py

import os
import gradio as gr
import openai
#from numpy._core.defchararray import endswith, isdecimal, startswith
from openai import OpenAI
from dotenv import load_dotenv
from pathlib import Path
from time import sleep
import audioread
import queue
import threading
from glob import glob
import copy
import base64
import json
from PIL import Image
from io import BytesIO
from pydantic import BaseModel
import pprint
import pandas as pd
import yfinance as yf
from datetime import datetime, timedelta
import pytz
import math
import numpy as np
# import matplotlib.pyplot as plt
from pylatexenc.latex2text import LatexNodes2Text


load_dotenv(override=True)
key = os.getenv('OPENAI_API_KEY')
users = os.getenv('LOGNAME')
unames = users.split(',')
pwds = os.getenv('PASSWORD')
pwdList = pwds.split(',')
DEEPSEEK_KEY=os.getenv('DEEPSEEK_KEY')
GROQ_KEY=os.getenv('GROQ_KEY')

site = os.getenv('SITE')
if site == 'local':
    dp = Path('./data')
    dp.mkdir(exist_ok=True)
    dataDir = './data/'
else:
    dp = Path('/data')
    dp.mkdir(exist_ok=True)
    dataDir = '/data/'
stock_data_path = dataDir + 'Stocks.txt'


speak_file = dataDir + "speek.wav"

# client = OpenAI(api_key = key)

#digits = ['zero: ','one: ','two: ','three: ','four: ','five: ','six: ','seven: ','eight: ','nine: ']

abbrevs = {'St. ' : 'Saint ', 'Mr. ': 'mister ', 'Mrs. ':'mussus ', 'Mr. ':'mister ', 'Ms. ':'mizz '}

special_chat_types = ['math', 'logic']

class Step(BaseModel):
    explanation: str
    output: str

class MathReasoning(BaseModel):
    steps: list[Step]
    final_answer: str

def Client():
    return OpenAI(api_key = key)

def test_plot_df():
    data = {
        "month": ['2024-01','2024-02','2024-03'],
        "value": [22.4, 30.1, 25.6]
        }
    return pd.DataFrame(data)

def md(txt):
    return str(txt).replace('```', ' ').replace('  ', '&nbsp;&nbsp;').replace('  ', '&nbsp;&nbsp;').replace('  ', '&nbsp;&nbsp;').replace('\n','<br>')
    # return txt

def etz_now():
    eastern = pytz.timezone('US/Eastern')
    ltime = datetime.now(eastern)
    return ltime

def date_from_utime(utime):
    ts = int(utime)
    dt = datetime.utcfromtimestamp(ts)
    eastern = pytz.timezone('US/Eastern')
    return dt.astimezone(eastern).strftime('%Y-%m-%d')

def convert_latex_math(text):
    lines = text.split('\n')
    start_line = False
    out_txt = ''
    for line in lines:
        if len(line) == 0:
            out_txt += '\n'
            continue
        else:
            if line == r'\]':
                continue
            if line == r'\[':
                start_line = True
                continue
            if start_line:
                line = '\n' + LatexNodes2Text().latex_to_text(line.strip())
                start_line = False
            if line.startswith(r'\['):
                loc = line.find(r'\]')
                if loc > 0:
                    latex_code = line[2:loc]
                    line = '\n' + LatexNodes2Text().latex_to_text(latex_code)
        out_txt += (line + '\n')
    return out_txt

def stock_list():
    rv = ''
    with open(stock_data_path, 'rt') as fp:
        lines = fp.readlines()
        for line in lines:
            (name, symbol, shares) = line.rstrip().split(',')
            name = name.strip()
            symbol = symbol.strip()
            rv += f'{symbol}  {name}\n'
    return rv

def get_stock_list():
    stock_list = {}
    with open(stock_data_path, 'rt') as fp:
        lines = fp.readlines()
        for line in lines:
            (name, symbol, shares) = line.rstrip().split(',')
            stock_list[symbol.strip()] = (name.strip(),shares.strip())
    return stock_list

def get_stock_news(search_symbol):
    fuzzy = True
    have_symbol = False
    search_symbol = search_symbol.strip().upper()
    stock_list = get_stock_list()
    search_term = search_symbol
    if search_symbol in stock_list.keys():
        have_symbol = True
        (search_term, shares) = stock_list[search_symbol]
    try:
        news = yf.Search(search_term, news_count=5, enable_fuzzy_query=fuzzy).news
    except:
        return (f'No results for search term {search_term}, check spelling', None)
    rv = ''
    for item in news:
        rv += f'Title: {item["title"]}\n'
        rv += f'Publisher: {item["publisher"]}\n'
        rv += f'Date published: {date_from_utime(item["providerPublishTime"])}\n'
        rv += f'Link: [URL]({item["link"]})\n\n'

    if have_symbol:
        (plot_df, ymax, deltas) = stock_week_df(search_symbol)
    else:
        (plot_df, ymax, deltas) = (pd.DataFrame(), 0.0, (0.0, 0.0, 0.0))

    return (rv, plot_df, ymax, deltas)

def stock_history_df(num_weeks):
    values = []
    dates = []
    xmax = 0
    for offset in range(num_weeks+1):
        (value, date) = get_stock_report(False, offset)
        # date = date[5:]
        values.append(value)
        dates.append(date)
        if float(value) > xmax:
            xmax = float(value)
    values.reverse()
    dates.reverse()
    data = {
        "date": dates,
        "value" : values
        }
    return (pd.DataFrame(data), f'{int(xmax + 10000)}')

# def make_mp_figure(dates, values, fit_values, ymax = 0.0):
#     npdates = np.asarray(dates)
#     npvals = np.asarray(values)
#     npfits = np.asarray(fit_values)
#     plt_format = '-'
#     fig = plt.figure(layout="constrained", figsize=(6,2))
#     ax = fig.add_subplot(111)
#     tics = []
#     labels = []
#     i = len(dates) - 1
#     while i >= 0:
#         tics.append(i)
#         labels.append(dates[i])
#         i -= 5
#     # tics = list(range(0,len(dates),5))
#     # labels = dates[0::5]
#     ax.set_xticks(tics, labels = labels)
#     ax.plot(npdates, npvals, plt_format)
#     ax.plot(npdates, npfits, plt_format)
#     ax.set_ylim(0.0, ymax*1.05)
#     # ax.set_xlim(dates[0], dates[-1:])
#     return fig

# def lms_fit_trend(dates, values):
#     # days = []
#     # fit_data = []
#     days = list(range(0, len(dates)))
#     fit = np.polyfit(days, values, 1)
#     delta = len(dates) * float(fit[0])
#     avg = float(fit[1]) + 0.5 * delta
#     pct_delta = 100 * delta / avg
#     # for day in days:
#     #     fit_data.append(float(fit[0]) * day + float(fit[1]) )
#     # return fit_data
#     return pct_delta

def stock_deltas(values):
    num = len(values)
    month_end_avg = float(np.average(np.array(values[-3:])))
    month_start_avg = float(np.average(np.array(values[0:4])))
    week_start_avg = float(np.average(np.array(values[-7:-4])))
    week_end_avg = float(np.average(np.array(values[-2:])))
    month_delta = 100 * (month_end_avg - month_start_avg)/month_start_avg
    week_delta = 100 * (week_end_avg - week_start_avg)/week_start_avg
    daily_delta = 100 * ((float(values[num-1])/float(values[num-2])) - 1.0)
    # avg = np.average(npa)
    return (month_delta, week_delta, daily_delta)

def stock_week_df(symbol):
    try:
        dates = []
        values = []
        ymax = 0
        etime = etz_now()
        if etime.hour >= 16:
            etime = etime + timedelta(days=1)
        week_ago = etime - timedelta(days=40) # was 8
        end = etime.strftime('%Y-%m-%d')
        start = week_ago.strftime('%Y-%m-%d')
        df = yf.download(symbol.upper(),
                        start = start,
                        end = end,
                        progress = False,
                        )
        vals2d = df.values.tolist()
        valsTxt = []
        numDays = len(vals2d)
        for i in range(numDays):
            valsTxt.append(vals2d[i][0])
        for val in valsTxt:
            v = round(float(val),2)
            values.append(v)
            if v > ymax:
                ymax = v
        for row in df.index:
            dates.append(row.strftime('%Y-%m-%d'))
        # fit_data = lms_fit(dates, values)
        # pct_delta = lms_fit_trend(dates, values)
        deltas = stock_deltas(values)
        data = {
            "date": dates,
            "value" : values,
            # "fit" : fit_data
            }
        # fig = make_mp_figure(dates, values, fit_data, ymax)
        return (pd.DataFrame(data), ymax, deltas)
    except:
        return (pd.DataFrame(), ymax, (0.0, 0.0, 0.0))

def get_stock_report(verbose = True, offset = 0):
    try:
        stock_data = {}
        global stock_data_path
        error_msg = ''
        with open(stock_data_path, 'rt') as fp:
            lines = fp.readlines()
        for line in lines:
            (name, symbol, shares) = line.rstrip().split(',')
            name = name.strip()
            symbol = symbol.strip()
            shares = shares.strip()
            stock_data[symbol] = {"symbol": symbol, "name": name, "shares": shares, "closing": '0'}
        for symbol in stock_data.keys():
            (closing_price, closing_date) = get_last_closing(symbol, offset)
            if closing_price == 0:
                error_msg += f'Error getting closing for {symbol}\n'
            stock_data[symbol]['closing'] = f'{closing_price:.2f}'
        total_value = 0.0
        if verbose:
            rv = f'At closing on {closing_date}:\n'
            for item in stock_data.values():
                rv += str(item) + '\n'
                total_value += float(item['closing']) * float(item['shares'])
            rv += (f'\nTotal value = {total_value:.2f}\n')
            if len(error_msg) > 0:
                rv += error_msg
            rv += f'Eastern time is: {etz_now()}'
        else:
            for item in stock_data.values():
                total_value += float(item['closing']) * float(item['shares'])
            return (total_value, closing_date)
    except:
        rv = 'Error getting stock report'
    return rv

def get_last_closing(symbol, offset=0, timeout=10):
    try:
        etime = etz_now()
        if etime.hour >= 16:
            etime = etime + timedelta(days=1)
        if offset > 0:
            etime = etime - timedelta(weeks=offset)
        five_days_ago = etime - timedelta(days=6)
        end = etime.strftime('%Y-%m-%d')
        start = five_days_ago.strftime('%Y-%m-%d')
        df = yf.download(symbol,
                        start = start,
                        end = end,
                        progress = False,
                        timeout=timeout,
                        )
        # print(df)
        closing_date = 'unknown'
        data_top = df.tail(1)
        for row in data_top.index:
            closing_date = row.strftime('%Y-%m-%d')
            # print(closing_date)
        return (df.iat[-1,0], closing_date)
    except:
        return (0.0, "0000-00-00")

def get_total_daily_closing_sequence(num_days):
    try:
        first_loop = True
        max_val = 0.0
        stock_list = get_stock_list()
        symbols = [s for s in stock_list.keys()]
        # symbols = symbols[8:10]
        etime = etz_now()
        if etime.hour >= 16:
            etime = etime + timedelta(days=1)
        end = etime.strftime('%Y-%m-%d')
        start_time = etime - timedelta(days = num_days)
        start = start_time.strftime('%Y-%m-%d')
        df = yf.download(symbols,
                start = start,
                end = end,
                progress = False,
                )
        # val2d = df.values.tolist()
        dates = []
        for row in df.index:
            dates.append(row.strftime('%Y-%m-%d'))
        # columns = list(df.columns.values)
        # cvals = df[columns[0]].tolist()

        for sym in symbols:
            (name, shares) = stock_list[sym]
            values = df[('Close', sym)].tolist()
            n = len(values)
            for i in range(n):
                if math.isnan(float(values[i])):
                    if i == 0:
                        values[0] = values[1]
                    else:
                        values[i] = values[i-1]
            if first_loop:
                first_loop = False
                total_values = values.copy()
                for i in range(n):
                    total_values[i] = float(total_values[i]) * float(shares)
            else:
                for i in range(n):
                    total_values[i] += (float(values[i]) * float(shares))
        for i in range(n):
            total_values[i] = round(total_values[i], 2)
            if total_values[i] > max_val:
                max_val = total_values[i]
        data = {
            "date": dates,
            "value" : total_values
            }
        return (pd.DataFrame(data), max_val)
    except:
        return (pd.DataFrame(), 0.0)

def get_daily_closing_sequence(symbol, num_days):
    try:
        dates = []
        values = []
        etime = etz_now()
        if etime.hour >= 16:
            etime = etime + timedelta(days=1)
        end = etime.strftime('%Y-%m-%d')
        start_time = etime - timedelta(days = num_days)
        start = start_time.strftime('%Y-%m-%d')
        df = yf.download(symbol,
                        start = start,
                        end = end,
                        progress = False,
                        )
        vals2d = df.values.tolist()
        valsTxt = []
        values = [round(float(vals2d[i][0]),2) for i in range(len(vals2d))]
        for row in df.index:
            dates.append(row.strftime('%Y-%m-%d'))
        return(dates, values)
    except:
        return([],[])

def create_stock_data_file(txt):
    with open(stock_data_path, 'wt') as fp:
        fp.write(txt)

def solve(prompt, chatType):
    tokens_in = 0
    tokens_out = 0
    tokens = 0
    if chatType == 'math':
        instruction = "You are a helpful math tutor. Guide the user through the solution step by step."
    elif chatType == "logic":
        instruction = "you are an expert in logic and reasoning.  Guide the user through the solution step by step"
    try:
        completion = Client().beta.chat.completions.parse(
            model = 'gpt-4o-2024-08-06',
            messages = [
                {"role": "system", "content": instruction},
                {"role": "user", "content": prompt}
            ],
            response_format=MathReasoning,
            max_tokens = 2000
        )

        tokens_in = completion.usage.prompt_tokens
        tokens_out = completion.usage.completion_tokens
        tokens = completion.usage.total_tokens
        msg = completion.choices[0].message 
        if msg.parsed:
            dr = msg.parsed.model_dump()
            response = pprint.pformat(dr)
        elif msg.refusal:
            response = msg.refusal

    except Exception as e:
        if type(e) == openai.LengthFinishReasonError:
            response = 'Too many tokens' 
        else:
            response = str(e)
    return (response, tokens_in, tokens_out, tokens)

def genUsageStats(do_reset=False):
    result = []
    ttotal4o_in = 0
    ttotal4o_out = 0
    ttotal4mini_in = 0
    ttotal4mini_out = 0
    totalAudio = 0
    totalSpeech = 0
    totalImages = 0
    totalHdImages = 0
    if do_reset:
        dudPath = dataDir + '_speech.txt'
        if os.path.exists(dudPath):
            os.remove(dudPath)
    for user in unames:
        tokens4o_in = 0
        tokens4o_out = 0
        tokens4mini_in = 0
        tokens4mini_out = 0
        fp = dataDir + user + '_log.txt'
        if os.path.exists(fp):
            accessOk = False
            for i in range(3):
                try:
                    with open(fp) as f:
                        dataList = f.readlines()
                    if do_reset:
                        os.remove(fp)
                    else:
                        for line in dataList:
                            (u, t) = line.split(':')
                            (t, m) = t.split('-')
                            (tin, tout) = t.split('/')
                            incount = int(tin)
                            outcount = int(tout)
                            if 'mini' in m:
                                tokens4mini_in += incount
                                tokens4mini_out += outcount
                                ttotal4mini_in += incount
                                ttotal4mini_out += outcount
                            else:
                                tokens4o_in += incount
                                tokens4o_out += outcount
                                ttotal4o_in += incount
                                ttotal4o_out += outcount
                    accessOk = True
                    break
                except:
                    sleep(3)
            if not accessOk:
                return f'File access failed reading stats for user: {user}'
        userAudio = 0
        fp = dataDir + user + '_audio.txt'
        if os.path.exists(fp):
            accessOk = False
            for i in range(3):
                try:
                    with open(fp) as f:
                        dataList = f.readlines()
                    if do_reset:
                        os.remove(fp)
                    else:
                        for line in dataList:
                            (dud, len) = line.split(':')
                            userAudio += int(len)
                        totalAudio += int(userAudio)
                    accessOk = True
                    break
                except:
                    sleep(3)
            if not accessOk:
                return f'File access failed reading audio stats for user: {user}'
        userSpeech = 0
        fp = dataDir + user + '_speech.txt'
        if os.path.exists(fp):
            accessOk = False
            for i in range(3):
                try:
                    with open(fp) as f:
                        dataList = f.readlines()
                    if do_reset:
                        os.remove(fp)
                    else:
                        for line in dataList:
                            (dud, len) = line.split(':')
                            userSpeech += int(len)
                        totalSpeech += int(userSpeech)
                    accessOk = True
                    break
                except:
                    sleep(3)
            if not accessOk:
                return f'File access failed reading speech stats for user: {user}'
        user_images = 0
        user_hd_images = 0
        fp = image_count_path(user)
        if os.path.exists(fp):
            accessOk = False
            for i in range(3):
                try:
                    with open(fp) as f:
                        dataList = f.readlines()
                    if do_reset:
                        os.remove(fp)
                    else:
                        for line in dataList:
                            x = line.strip()
                            if x == 'hd':
                                user_hd_images += 1
                                totalHdImages += 1
                            else:
                                user_images += 1
                                totalImages += 1
                    accessOk = True
                    break
                except:
                    sleep(3)
            if not accessOk:
                return f'File access failed reading image gen stats for user: {user}'
        result.append([user, f'{tokens4mini_in}/{tokens4mini_out}', f'{tokens4o_in}/{tokens4o_out}', f'audio:{userAudio}',f'speech:{userSpeech}', f'images:{user_images}/{user_hd_images}'])
    result.append(['totals', f'{ttotal4mini_in}/{ttotal4mini_out}', f'{ttotal4o_in}/{ttotal4o_out}', f'audio:{totalAudio}',f'speech:{totalSpeech}', f'images:{totalImages}/{totalHdImages}'])
    return result       

def new_conversation(user):
    clean_up(user)  # .wav files
    flist = glob(f'{dataDir}{user}.png')
    flist.extend(glob(f'{dataDir}{user}_image.b64'))
    for fpath in flist:
        if os.path.exists(fpath):
            os.remove(fpath)
    return [None, [], gr.Markdown(value='', label='Dialog', container=True), gr.Image(visible=False, value=None),  gr.Image(visible=False, value=None), '',
            gr.LinePlot(visible=False)]

def updatePassword(txt):
    password = txt.lower().strip()
    return [password, "*********"]

# def parse_math(txt):
#     ref = 0
#     loc = txt.find(r'\(')
#     if loc == -1:
#         return txt
#     while (True):
#         loc2 = txt[ref:].find(r'\)')
#         if loc2 == -1:
#             break
#         loc = txt[ref:].find(r'\(')
#         if loc > -1:
#             loc2 += 2
#             slice = txt[ref:][loc:loc2]
#             frag = lconv.convert(slice)
#             txt = txt[:loc+ref] + frag + txt[loc2+ref:]
#             ref = len(txt[ref:loc]) + len(frag)
#     return txt

def chat(prompt, user_window, pwd_window, past, response, gptModel, uploaded_image_file='', plot=None):
    image_gen_model = 'gpt-4o-2024-08-06'
    user_window = user_window.lower().strip()
    isBoss = False
    if not response:
        response = ''
    plot = gr.LinePlot(visible=False)
    # plot = gr.Plot(visible=False)
    if user_window == unames[0] and pwd_window == pwdList[0]:
        isBoss = True
        if prompt == 'stats':
            response = genUsageStats()
            return [past, md(response), None, gptModel, uploaded_image_file, plot]
        if prompt == 'reset':
            response = genUsageStats(True)
            return [past, md(response), None, gptModel, uploaded_image_file, plot]
        if prompt.startswith('gpt4'):
            gptModel = 'gpt-4o-2024-08-06'
            prompt = prompt[5:]
        if prompt.startswith("clean"):
            user = prompt[6:]
            response = f'cleaned all .wav and .b64 files for {user}'
            final_clean_up(user, True)
            return [past, response, None, gptModel, uploaded_image_file, plot]
        if prompt.startswith('files'):
            (log_cnt, wav_cnt, other_cnt, others, log_list) = list_permanent_files()
            response = f'{log_cnt} log files\n{wav_cnt} .wav files\n{other_cnt} Other files:\n{others}\nlogs: {str(log_list)}'
            return [past, response, None, gptModel, uploaded_image_file, plot]
        if prompt.startswith('stock'):
            args = prompt.split(' ')
            num = len(args)
            if num == 1:
                response = stock_list()
                return [past, md(response), None, gptModel, uploaded_image_file, plot]
            elif num == 2:
                response = get_stock_report()
                if args[1] == 'value':
                    return [past, md(response), None, gptModel, uploaded_image_file, plot]
                elif args[1] == 'history':
                    (plot_df, ymax) = get_total_daily_closing_sequence(40)   #stock_history_df(12)
                    # ymax = float(ymax)
                    return [past, md(response), None, gptModel, uploaded_image_file, # plot]
                           gr.LinePlot(plot_df, x="date", y="value", visible=True, x_label_angle=270,
                                      y_lim=[500000, 700000], label="Portfolio Value History")]
            elif num >= 3:
                if args[1] == 'news':
                    symbol = ' '.join(args[2:])
                    (response, plot_df, ymax, (dm, dw, dd)) = get_stock_news(symbol)
                    ymax *= 1.1
                    mdtxt = md(f'News for {symbol}:\nTrends: Month = {dm:.1f}%, Week = {dw:.1f}%, Day = {dd:.1f}%\n\n' + response)
                    if plot_df.empty:
                        return [past, mdtxt, None, gptModel, uploaded_image_file, plot]
                    else:
                        return [past, mdtxt, None, gptModel, uploaded_image_file, #gr.Plot(plot_df, visible=True)]
                                    gr.LinePlot(plot_df, x="date", y="value", visible=True, x_label_angle=270,
                                                    y_lim=[0, ymax],label=f"{symbol.upper()} Recent Prices",
                                                    color_map={''})]
                # elif arg[1] == 'history':
                #     symbol = arg[2]
                #     response = 'ok' # get_


        # if prompt.startswith('stock values'):
        #     response = get_stock_report()
        #     if 'history' in prompt:
        #         (plot_df, ymax) = stock_history_df(12)
        #         ymax = float(ymax)
        #         return [past, response, None, gptModel, uploaded_image_file,
        #                gr.LinePlot(plot_df, x="date", y="value", visible=True,
        #                           y_lim=[500000, 700000], label="Portfolio Value History")]
        #     else:
        #         return [past, response, None, gptModel, uploaded_image_file, plot]
        # if prompt.startswith('stock news'):
        #         symbol = prompt[11:]
        #         response = get_stock_news(symbol)
        #         return [past, response, None, gptModel, uploaded_image_file, plot]
        if prompt.startswith('stockload'):
            create_stock_data_file(prompt[9:].lstrip())
            return [past, 'Stock data file created', None, gptModel, uploaded_image_file, plot]
    if user_window in unames and pwd_window == pwdList[unames.index(user_window)]:
        chatType = 'normal'
        deepseek = False
        using_groq = False
        reasoning = False
        prompt = prompt.strip()
        if prompt.lower().startswith('dsr1 '):
            deepseek = True
            ds_model = 'deepseek-ai/DeepSeek-R1'
            prompt = prompt[5:]
        elif prompt.lower().startswith('ds1.5 '):
            deepseek = True
            ds_model = 'deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B'
            prompt = prompt[6:]
        elif prompt.lower().startswith('ds14 '):
            deepseek = True
            ds_model = 'deepseek-ai/DeepSeek-R1-Distill-Qwen-14B'
            prompt = prompt[5:]
        elif prompt.lower().startswith('ds70 '):
            deepseek = True
            ds_model = 'deepseek-ai/DeepSeek-R1-Distill-Llama-70B'
            prompt = prompt[5:]
        elif prompt.lower().startswith('ds70g '):
            deepseek = True
            using_groq = True
            ds_model = 'deepseek-r1-distill-llama-70b'
            prompt = prompt[6:]
        elif prompt.lower().startswith('o1m '):
            reasoning = True
            gptModel = 'o1-mini'
            prompt = prompt[4:] + \
            '. Provide a detailed step-by-step description of your reasoning. Do not use Latex for math expressions.'
        elif prompt.lower().startswith('solve'):
            prompt = 'How do I solve ' + prompt[5:] + ' Do not use Latex for math expressions.'
            chatType = 'math'
        elif prompt.lower().startswith('puzzle'):
            chatType = 'logic'
            prompt = prompt[6:]
        if deepseek:
            prompt = prompt + '. Do not use Latex for math expressions.'
        past.append({"role":"user", "content":prompt})
        gen_image = (uploaded_image_file != '')
        if chatType in special_chat_types:
           (reply, tokens_in, tokens_out, tokens) = solve(prompt, chatType)
           final_text = reply
           reporting_model = image_gen_model
        elif not gen_image:
            if deepseek:
                if using_groq:
                    client = OpenAI(api_key=GROQ_KEY, base_url='https://api.groq.com/openai/v1')
                    completion = client.chat.completions.create(
                        temperature=0.6,
                        model= ds_model,
                        messages=past)
                    reporting_model='deepseek70-groq'
                else:
                    client = OpenAI(api_key=DEEPSEEK_KEY, base_url='https://api.together.xyz/v1')
                    completion = client.chat.completions.create(
                        temperature=0.6,
                        model= ds_model,
                        messages=past)
                    reporting_model='deepseek-together-' + ds_model[-3:].replace('.5B','1.5B')
            else:
                completion = Client().chat.completions.create(model=gptModel,
                                            messages=past)
                reporting_model = gptModel
        else:
            (completion, msg) = analyze_image(user_window, image_gen_model, prompt)
            uploaded_image_file= ''
            reporting_model = image_gen_model
            if not msg == 'ok':
                return [past, msg, None, gptModel, uploaded_image_file, plot]
        if not chatType in special_chat_types:
            reply = completion.choices[0].message.content
            # if 'groq' in reporting_model:
            if deepseek:
                reply = convert_latex_math(reply)
            final_text = reply
            if deepseek:
                loc1 = reply.find('<think>')
                if loc1 > -1:
                    loc2 = reply.find('</think>')
                    if loc2 > loc1:
                        final_text = reply[loc2 + 8:]
                reply = reply.replace('<think>','\n***Thinking***\n').replace('</think>','\n***Done thinking***\n')
            tokens_in = completion.usage.prompt_tokens
            tokens_out = completion.usage.completion_tokens
            tokens = completion.usage.total_tokens
        response += md("\n\n***YOU***: " + prompt + "\n***GPT***: " + reply)
        if isBoss:
            response += md(f"\n{reporting_model}: tokens in/out = {tokens_in}/{tokens_out}")
        if tokens > 40000:
            response += "\n\nTHIS DIALOG IS GETTING TOO LONG. PLEASE RESTART CONVERSATION SOON."
        past.append({"role":"assistant", "content": final_text})
        if not deepseek and not reasoning:
            accessOk = False
            for i in range(3):
                try:
                    dataFile = new_func(user_window)
                    with open(dataFile, 'a') as f:
                        m = '4o'
                        if 'mini' in reporting_model:
                            m = '4omini'
                        f.write(f'{user_window}:{tokens_in}/{tokens_out}-{m}\n')
                    accessOk = True
                    break
                except Exception as e:
                    sleep(3)
            if not accessOk:
                response += f"\nDATA LOG FAILED, path = {dataFile}"
        return [past, response , None, gptModel, uploaded_image_file, plot]
    else:
        return [[], "User name and/or password are incorrect", prompt, gptModel, uploaded_image_file, plot]

def new_func(user):
    dataFile = dataDir + user + '_log.txt'
    return dataFile

def image_count_path(user):
    fpath = dataDir + user + '_image_count.txt'
    return fpath

def transcribe(user, pwd, fpath):
    user = user.lower().strip()
    pwd = pwd.lower().strip()
    if not (user in unames and pwd in pwdList):
        return 'Bad credentials'
    with audioread.audio_open(fpath) as audio:
        duration = int(audio.duration)
        if duration > 0:
            with open(dataDir + user + '_audio.txt','a') as f:
                f.write(f'audio:{str(duration)}\n')
    with open(fpath,'rb') as audio_file:
        transcript = Client().audio.transcriptions.create(
            model='whisper-1', file = audio_file ,response_format = 'text' )
    reply = transcript
    return str(reply)

def pause_message():
    return "Audio input is paused.  Resume or Stop as desired"

# def gen_output_audio(txt):
#     if len(txt) < 10:
#         txt = "This dialog is too short to mess with!"
#     response = Client().audio.speech.create(model="tts-1", voice="fable", input=txt)
#     with open(speak_file, 'wb') as fp:
#         fp.write(response.content)
#     return speak_file


# def set_speak_button(txt):
#     vis = False
#     if txt and len(txt) > 2:
#         vis = True
#     return gr.Button(visible=vis)

def update_user(user_win):
    user_win = user_win.lower().strip()
    user = 'unknown'
    for s in unames:
        if user_win == s:
            user = s
            break
    return [user, user]

def speech_worker(chunks=[],q=[]):
    for chunk in chunks:
        fpath = q.pop(0)
        response = Client().audio.speech.create(model="tts-1", voice="fable", input=chunk, speed=0.85, response_format='wav')
        with open(fpath, 'wb') as fp:
            fp.write(response.content)

def gen_speech_file_names(user, cnt):
    rv = []
    for i in range(0, cnt):
        rv.append(dataDir + f'{user}_speech{i}.wav')
    return rv

def final_clean_up(user, do_b64 = False):
    user = user.strip().lower()
    if user == 'kill':
        flist = glob(dataDir + '*')
    elif user == 'all':
        flist = glob(dataDir + '*_speech*.wav')
        if do_b64:
            flist.extend(glob(dataDir + '*.b64'))
    else:
        flist = glob(dataDir + f'{user}_speech*.wav')
        if do_b64:
            flist.append(dataDir + user + '_image.b64')
    for fpath in flist:
        try:
            os.remove(fpath)
        except:
            continue

def delete_image(user):
    fpath = dataDir + user + '.png'
    if os.path.exists(fpath):
        os.remove(fpath)

def list_permanent_files():
    flist = os.listdir(dataDir)
    others = []
    log_cnt = 0
    wav_cnt = 0
    other_cnt = 0
    list_logs = []
    for fpath in flist:
        if fpath.endswith('.txt'):
            log_cnt += 1
            list_logs.append(fpath)
        elif fpath.endswith('.wav'):
            wav_cnt += 1
        else:
            others.append(fpath)
    other_cnt = len(others)
    if log_cnt > 5:
        list_logs = []
    return (str(log_cnt), str(wav_cnt), str(other_cnt), str(others), list_logs)

def make_image(prompt, user, pwd):
    user = user.lower().strip()
    msg = 'Error: unable to create image.'
    fpath = None
    model = 'dall-e-2'
    size = '512x512'
    quality = 'standard'
    if user in unames and pwd == pwdList[unames.index(user)]:
        if len(prompt.strip()) == 0:
            return [gr.Image(value=None, visible=False), 'You must provide a prompt describing image you desire']
        if prompt.startswith('hd '):
            prompt = prompt[3:]
            model = 'dall-e-3'
            size = '1024x1024'
            quality = 'hd'
        try:
            response = Client().images.generate(model=model, prompt=prompt,size=size,
               quality=quality, response_format='b64_json')
        except Exception as ex:
            msg = ex.message
            return [gr.Image(visible=False, value=None), msg]
        if len(response.data) == 0:
            msg = "OpenAI returned no image data"
            return [gr.Image(visible=False, value=None), msg]
        try:
            image_data = response.data[0].b64_json 
            with Image.open(BytesIO(base64.b64decode(image_data))) as image:
                fpath = dataDir + user + '.png'
                image.save(fpath)
            with open(image_count_path(user), 'at') as fp:
                if quality == 'hd':
                    fp.write('hd\n')
                else:
                    fp.write('1\n')
            msg = 'Image created!'
        except:
            return [gr.Image(visible=False, value=None), msg]
    else:
        msg = 'Incorrect user name or password'
        return [gr.Image(visible=False, value=None), msg]
    return [gr.Image(visible=True, value=fpath), msg]

def show_help():
    txt = '''
    1.  Gemeral:
        1.1 Login with user name and password (not case-sensitive)
        1.2 Type prompts (questions, instructions) into "Prompt or Question" window (OR) you can speak prompts by
           tapping the audio "Record" button, saying your prompt, then tapping the "Stop" button.
           Your prompt will appear in the Prompt window, and you can edit it there if needed.
        1.3 Text in the "Dialog" window can be spoken by tapping the "Speak Dialog" button.
    2.  Chat:
        2.1 Enter prompt and tap the "Submit Prompt/Question" button.  The responses appear in the Dialog window.
        2.2 Enter follow-up questions in the Prompt window either by typing or speaking. Tap the voice
              entry "Reset Voice Entry" button to enable additional voice entry. Then tap "Submit Prompt/Question".
        2.3 If topic changes or when done chatting, tap the "Restart Conversation" button.
    3.  Solve math equations or logic problems providing step-by-step analysis:
        3.1 Math:  Make "solve" the first word in your prompt, followed by the equation, e.g., x^2 - x + 1 = 0
        3.2 Logic: Make "puzzle" the first word in your prompt, followed by a detailed description of a logic
           problem with the answer(s) you desire.
    4.  Make Image:
          4.1 Enter description of desired image in prompt window via either typing or voice entry
          4.2 Tap the "Make Image" button.  This can take a few seconds.
          4.3 There is a download button on the image display if your system supports file downloads.
          4.4 When done viewing image, tap the "Restart Conversation" button
    5.  Analyze an Image you provide:
          5.1 Enter what you want to know about the image in the prompt window. You can include instructions
               to write a poem about something in the image, for example.  Or just say "what's in this image?"
          5.2 Tap the "Upload Image to Analyze" button.
          5.3 An empty image box will appear lower left. Drag or upload image into it. It offers web cam or camera
               input also.
          5.4 The image should appear. This can take some time with a slow internet connection and large image.
          5.5 Tap the "Submit Prompt/Question" button to start the analysis.  This initiates a chat dialog and
               you can ask follow-up questions. However, the image is not re-analyzed for follow-up dialog.
    Hints:
        1. Better chat and image results are obtained by including detailed descriptions and instructions
            in the prompt.
        2. Always tap "Restart Conversation" before requesting an image or changing chat topics.
        3. Audio input and output functions depend on the hardware capability of your device.
        4. "Speak Dialog" will voice whatever is currently in the Dialog window.  You can repeat it and you
             can edit what's to be spoken.  Except:  In a chat conversation, spoken dialog will only include
             the latest prompt/response ("YOU:/GPT:") sequence.'''
    return md(txt)

def upload_image(prompt, user, password):
    if not (user in unames and password == pwdList[unames.index(user)]):
        return [gr.Image(visible=False, interactive=True), "Incorrect user name and/or password"]
    if len(prompt) < 3:
        return [gr.Image(visible=False, interactive=True), "You must provide prompt/instructions (what to do with the image)"]
    return [gr.Image(visible=True, interactive=True), '']

def load_image(image, user):
    status = 'OK, image is ready! Tap "Submit Prompt/Question" to start analyzing'
    try:
        with open(image, 'rb') as image_file:
            base64_image = base64.b64encode(image_file.read()).decode('utf-8')
        fpath = dataDir + user + '_image.b64'
        with open(fpath, 'wt') as fp:
            fp.write(base64_image)
    except:
        status = 'Unable to upload image'
    return [fpath, status]

def analyze_image(user, model, prompt):
    status = 'ok'
    try:
        with open(dataDir + user + '_image.b64', 'rt') as fp:
            base64_image = fp.read()
    except:
        status = "base64 image file not found"
        return [None, status]

    completion = Client().chat.completions.create(
        model=model,
        messages=[
            { "role": "user",
               "content": [
                   {
                       "type": "text",
                       "text": prompt
                    },
                    {
                        "type": "image_url",
                        "image_url": {
                            "url": f"data:image/jpeg;base64,{base64_image}",
                            "detail": "high"
                            }
                    }
                   ]
               }
            ],
        max_tokens= 500
    )
    # response = completion.choices[0].message.content
    return [completion, status]


with gr.Blocks(theme=gr.themes.Soft()) as demo:
    history = gr.State([])
    password = gr.State("")
    user = gr.State("unknown")
    model = gr.State("gpt-4o-mini")
    q = gr.State([])
    qsave = gr.State([])
    uploaded_image_file = gr.State('')

    def clean_up(user):
        flist = glob(dataDir + f'{user}_speech*.wav')
        for fpath in flist:
            try:
                os.remove(fpath)
            except:
                continue

    def initial_audio_output(txt, user):
        global digits
        global abbrevs
        if not user in unames:
            return [gr.Audio(sources=None), []]
        clean_up(user)
        q = []
        if len(txt.strip()) < 5:
            return ['None', q]
        try:
            loc = txt.rindex('YOU:')
            txt = txt[loc:]
        except:
            pass
        for s,x in abbrevs.items():
            txt = txt.replace(s, x)
        words_in = txt.replace('**', '').replace('&nbsp;','').split('<br>')
        words_out = []
        for s in words_in:
            s = s.lstrip('- *@#$%^&_=+-')
            if len(s) > 0:
                loc = s.find(' ')
                if loc > 1:
                    val = s[0:loc]
                    isnum = val.replace('.','0').isdecimal()
                    if isnum:
                        if val.endswith('.'):
                            val = val[:-1].replace('.',' point ') + '., '
                        else:
                            val = val.replace('.', ' point ') + ', '
                        s = 'num'+ val + s[loc:]
                words_out.append(s)
        chunklist = []
        for chunk in words_out:
            if chunk.strip() == '':
                continue
            isnumbered = chunk.startswith('num')
            number = ''
            loc = 0
            if isnumbered:
                chunk = chunk[3:]
                loc = chunk.index(',')
                number = chunk[0:loc]
                chunk = chunk[loc:]
            locs = []
            for i in range(1,len(chunk)-1):
                (a, b, c) = chunk[i-1:i+2]
                if a.isdecimal() and b == '.' and c.isdecimal():
                    locs.append(i)
            for i in locs:
                chunk = chunk[:i] + ' point ' + chunk[i+1:]
            if len(chunk) > 50:
                finechunks = chunk.split('.')
                for fchunk in finechunks:
                    if isnumbered:
                        fchunk = number + fchunk
                        isnumbered = False
                    if len(fchunk) > 0:
                        if fchunk != '"':
                            chunklist.append(fchunk)
            else:
                line = number + chunk
                if line != '"':
                    chunklist.append(line)
        total_speech = 0
        for chunk in chunklist:
            total_speech += len(chunk)
        with open(dataDir + user + '_speech.txt','a') as f:
            f.write(f'speech:{str(total_speech)}\n')
        chunk = chunklist[0]
        if chunk.strip() == '':
            return gr.Audio(sources=None)
        fname_list = gen_speech_file_names(user, len(chunklist))
        q = fname_list.copy()
        qsave = fname_list.copy()
        fname = q.pop(0)
        if len(chunklist) > 0:
            threading.Thread(target=speech_worker, daemon=True, args=(chunklist[1:],fname_list[1:])).start()
        response = Client().audio.speech.create(model="tts-1", voice="fable", input=chunk, speed=0.85, response_format='wav')
        with open(fname, 'wb') as fp:
            fp.write(response.content)
        return [fname, q]

    def gen_output_audio(q, user):
        try:
            fname = q.pop(0)
        except:
            final_clean_up(user)
            return [None, gr.Audio(sources=None)]
        if not os.path.exists(fname):
            sleep(3)
            if not os.path.exists(fname):
                response = Client().audio.speech.create(model="tts-1", voice="fable",
                    input='Sorry, text-to-speech is responding too slow right now', speed=0.85, response_format='wav')
                with open(fname, 'wb') as fp:
                    fp.write(response.content)
                q = []
        return [fname, q]


    gr.Markdown('# GPT Chat')
    gr.Markdown('Enter user name & password.  Tap "Help & Hints" button for more instructions.')
    with gr.Row():
        user_window = gr.Textbox(label = "User Name")
        user_window.blur(fn=update_user, inputs=user_window, outputs=[user, user_window])
        pwd_window = gr.Textbox(label = "Password")
        pwd_window.blur(updatePassword, inputs = pwd_window, outputs = [password, pwd_window])
        help_button = gr.Button(value='Help & Hints')
    with gr.Row():
        audio_widget = gr.Audio(type='filepath', format='wav',waveform_options=gr.WaveformOptions(
           show_recording_waveform=True), sources=['microphone'], scale = 3, label="Prompt/Question Voice Entry", max_length=120)
        reset_button = gr.ClearButton(value="Reset Voice Entry", scale=1) #new_func1()
    with gr.Row():
        clear_button = gr.Button(value="Restart Conversation")
        # gpt_chooser=gr.Radio(choices=[("GPT-3.5","gpt-3.5-turbo"),("GPT-4o","gpt-4o-mini")],
        #                      value="gpt-3.5-turbo", label="GPT Model", interactive=True)
        button_do_image = gr.Button(value='Make Image')
        button_get_image = gr.Button(value='Upload Image to Analyze')
        speak_output = gr.Button(value="Speak Dialog", visible=True)
        submit_button = gr.Button(value="Submit Prompt/Question")
    prompt_window = gr.Textbox(label = "Prompt or Question")
    gr.Markdown('### **Dialog:**')
    output_window = gr.Markdown(container=True)
    with gr.Row():
        with gr.Column():
            image_window2 = gr.Image(visible=False, interactive=True, label='Image to Analyze', type='filepath')
        with gr.Column():
            image_window = gr.Image(visible=False, label='Generated Image')
    with gr.Row():
        # plot = gr.Plot(visible=False)
        plot = gr.LinePlot(test_plot_df(), x="month", y="value", visible=False, label="Portfolio Value History")
    submit_button.click(chat,
             inputs=[prompt_window, user_window, password, history, output_window, model, uploaded_image_file],
             outputs=[history, output_window, prompt_window, model, uploaded_image_file, plot])
    clear_button.click(fn=new_conversation, inputs=user_window,
                      outputs=[prompt_window, history, output_window, image_window, image_window2, uploaded_image_file, plot])
    audio_widget.stop_recording(fn=transcribe, inputs=[user_window, password, audio_widget],
                                outputs=[prompt_window])
    audio_widget.pause_recording(fn=pause_message, outputs=[prompt_window])
    reset_button.add(audio_widget)
    audio_out = gr.Audio(autoplay=True, visible=False)
    audio_out.stop(fn=gen_output_audio, inputs=[q, user_window], outputs = [audio_out, q])
    speak_output.click(fn=initial_audio_output, inputs=[output_window, user_window], outputs=[audio_out, q])
    # output_window.change(fn=set_speak_button, inputs=output_window,outputs=speak_output)
    button_do_image.click(fn=make_image, inputs=[prompt_window,user_window, password],outputs=[image_window, output_window])
    image_window.change(fn=delete_image, inputs=[user])
    help_button.click(fn=show_help, outputs=output_window)
    button_get_image.click(fn=upload_image,inputs = [prompt_window, user, password], outputs = [image_window2, output_window])
    image_window2.upload(fn=load_image, inputs=[image_window2, user], outputs=[uploaded_image_file, output_window])
    # demo.unload(final_clean_up(user))
demo.launch(share=True, allowed_paths=[dataDir], ssr_mode=False)