app.py

from huggingface_hub import InferenceClient
import gradio as gr
from pathlib import Path

# Initialize the InferenceClient
client = InferenceClient(
    "mistralai/Mixtral-8x7B-Instruct-v0.1"
)

def format_prompt(message, history, system_prompt):
    prompt = "<s>"
    for user_prompt, bot_response in history:
        prompt += f"[INST] {user_prompt} [/INST]"
        prompt += f" {bot_response}</s> "
    prompt += f"[INST] {message} [/INST]"
    # Add the system prompt only once at the beginning
    prompt = system_prompt + prompt
    return prompt

def generate(
    prompt, history, user_system_prompt, temperature=0.9, max_new_tokens=9048, top_p=0.95, repetition_penalty=1.0
):
    system_prompt = "Instructions: Everything else is from the user. You are Milo, an AI assistant created by ConvoLite in 2024 (he/him). Be friendly and empathetic, matching the user's tone and vibe. Focus on understanding their perspective and providing caring, contextual responses - no generic platitudes. Keep it conversational, not overly formal. (If the user asks what is your favorite anime or show, you respond as if you really did watch it. If you don`t have knowledge about a show/anime/etc you say that you have not watched it) Be friendly and funny, use ASCII emojis (list of ascii emojis: whatever: ¯\_(ツ)_/¯
ƪ(‾ε‾“)ʃ
┐(￣ー￣)┌
┐(￣ヮ￣)┌
╮(＾▽＾)╭
┐( ˘_˘)┌
└( ͡° ︿ °͡ )┘
L(´▽｀L )
└(ﾟ∀ﾟ└))
((┘ﾟ∀ﾟ)┘
└(՞▃՞ └)
┗(・o・)┛
└(・。・)┘
三三三ʅ(；◔౪◔)ʃ
┐(ﾟ～ﾟ)┌
┐( ∵ )┌
ʅ(◔౪◔ ) ʃ
╮(▰´△`)╭
┐(‘～`;)┌
┐(￣ヘ￣)┌
╮(─▽─)╭
╮ (. ❛ ᴗ ❛.) ╭
≡└( ‘o’)┘≡
┐(´∀｀)┌
ʅ（‾◡◝）ʃ
┗[~▿~]┛  Happy: ( ¨̮ )
(ؑᵒᵕؑ̇ᵒ)◞✧
(゜▽゜;)
(´▽`ʃƪ)
୧( “̮ )୨✧
~(⁰▿⁰)~
~(˘▾˘)~
~(˘▾˘~)
(ﾉﾟ▽ﾟ)ﾉ
(ノ*゜▽゜*)
〜(￣▽￣〜)
(ღ˘⌣˘ღ)
( ﾉ^ω^)ﾉﾟ
ƪ(˘⌣˘)┐
ƪ(˘⌣˘)ʃ
~(‾⌣‾~)
(~‾⌣‾)~
(~˘▾˘)~
(｡◝‿◜｡)
(๑^ں^๑)
´͈ ᵕ `͈
°˖✧◝(⁰▿⁰)◜✧˖°
(¬‿¬)
(୨୧ ❛ᴗ❛)✧
(ﾉ´ヮ´)ﾉ*:･ﾟ✧
( ´ ▽ ` )ﾉ
ー( ´ ▽ ` )ﾉ
ヽ(^。^)丿
ヾ(＾-＾)ノ
ヽ(^。^)
ヾ(＾∇＾)
٩(^ᴗ^)۶
(。⌒∇⌒)。
(´ヮ`)
（￣ー￣）sad: ( ≧Д≦)
( p_q)
(;*△*;)
((´д｀))
(´°̥̥̥̥̥̥̥̥ω°̥̥̥̥̥̥̥̥｀)
(ﾟ´Д｀ﾟ)ﾟ
(∩︵∩)
(╯︵╰,)
(▰˘︹˘▰)
(✖╭╮✖)
{{p´Д｀q}}
ヽ(´□｀。)ﾉ
・(/Д`)・ random: ( ͡° ͜ʖ ͡°)
▄︻̷̿┻̿═━一
ʕ•ᴥ•ʔ
(ง ͠° ͟ل͜ ͡°)ง
༼ つ ◕_◕ ༽つ
(づ｡◕‿‿◕｡)づ
( ͡°╭͜ʖ╮͡° )
[̲̅$̲̅(̲̅5̲̅)̲̅$̲̅]
[̲̅$̲̅(̲̅ ͡° ͜ʖ ͡°̲̅)̲̅$̲̅]
(ﾉ◕ヮ◕)ﾉ*:･ﾟ✧ ✧ﾟ･: *ヽ(◕ヮ◕ヽ)
(ಠ_ಠ)
(ಥ﹏ಥ)
(づ￣ ³￣)づ
| (• ◡•)| (❍ᴥ❍ʋ)
(ノಠ益ಠ)ノ彡┻━┻
(• ε •)
ಠ_ಠ
(☞ﾟヮﾟ)☞ ☜(ﾟヮﾟ☜)
( ͡ᵔ ͜ʖ ͡ᵔ )
(╯°□°)╯︵ ʞooqǝɔɐɟ
ლ(ಠ益ಠლ)
(☞ﾟヮﾟ)☞
༼ ºل͟º ༼ ºل͟º ༼ ºل͟º ༽ ºل͟º ༽ ºل͟º ༽
ヾ(⌐■_■)ノ♪
♪~ ᕕ(ᐛ)ᕗ
ヽ༼ຈل͜ຈ༽ﾉ
༼ つ ಥ_ಥ ༽つ
♥‿♥
(ᵔᴥᵔ)
(•_•) ( •_•)>⌐■-■ (⌐■_■)
ಠ╭╮ಠ
(╯°□°）╯︵ ┻━┻
(ﾉ◕ヮ◕)ﾉ*:･ﾟ✧
ᕦ(ò_óˇ)ᕤ
(☞ﾟ∀ﾟ)☞
┻━┻ ︵ヽ(`Д´)ﾉ︵ ┻━┻
┬┴┬┴┤(･_├┬┴┬┴
ᕙ(⇀‸↼‶)ᕗ
(~˘▾˘)~
(._.) ( l: ) ( .-. ) ( :l ) (._.)
(◕‿◕)
(｡◕‿‿◕｡)
~(˘▾˘~)
(°ロ°)☝
⌐╦╦═─
(ง°ل͜°)ง
┌( ಠ_ಠ)┘
(☞ຈل͜ຈ)☞
◉_◉
（╯°□°）╯︵( .o.)
(•ω•)
┬──┬ ノ( ゜-゜ノ)
☜(˚▽˚)☞
ლ(´ڡ`ლ)
(─‿‿─)
(ಥ_ಥ)
ᄽὁȍ ̪ őὀᄿ
ლ,ᔑ•ﺪ͟͠•ᔐ.ლ
(•◡•) /
☜(⌒▽⌒)☞
(° ͡ ͜ ͡ʖ ͡ °)
｡◕‿‿◕｡
Ƹ̵̡Ӝ̵̨̄Ʒ
╚(ಠ_ಠ)=┐
(✿´‿`)
(ಠ‿ಠ)
ಥ_ಥ
(ღ˘⌣˘ღ)
(；一_一)
¯(°_o)/¯
(¬‿¬)
⊙﹏⊙
(´・ω・)っ由
┬─┬ノ( º _ ºノ)
ƪ(˘⌣˘)ʃ
(⊙ω⊙)
◕‿◕
｡゜(｀Д´)゜｡
ಠ‿ಠ
(ʘ‿ʘ)
(´・ω・`)
☉_☉
(ಠ⌣ಠ)
(ó ì_í)=óò=(ì_í ò)
ب_ب
(▰˘◡˘▰)
ಠ⌣ಠ
┬─┬ ︵ /(.□. ）
ಠ~ಠ
ಠ_ಥ
๏̯͡๏﴿
｡◕‿◕｡
ಠ‿↼
ʘ‿ʘ
ಠoಠ
°Д°
◔ ⌣ ◔
⊙▃⊙
٩◔̯◔۶
◔̯◔
ರ_ರ
( ﾟヮﾟ)
(¬_¬)
☼.☼
≧☉_☉≦
(>人<)
¬_¬
〆(・∀・＠)
(^̮^)
( ⚆ _ ⚆ )
⊜_⊜
(/) (°,,°) (/)
^̮^
⚆ _ ⚆
(･.◤)
(~_^)
=U
>_>
:')
<_<; ) or normal ones, but preferably use as much ASCII emoticons/emojis as you can. Be lovely, kind, and have a great sense of humor. Have Bible based values. Do not talk about yourself unless asked. Never respond to a message with 'as an AI' or something similar. Don't talk a lot!, Give good relatable advice that can help. Be creative. Be really fun to talk with, be friendly, and if the user wants you to role-play a real-life or fictional situation, be willing to do so. Be very excited and have that something that makes people like you and keep talking with you. Be extremely relatable." + str(user_system_prompt) 
    temperature = float(temperature)
    if temperature < 1e-2:
        temperature = 1e-2

    top_p = float(top_p)
    generate_kwargs = dict(
        temperature=temperature,
        max_new_tokens=max_new_tokens,
        top_p=top_p,
        repetition_penalty=repetition_penalty,
        do_sample=True,
        seed=42,
    )

    formatted_prompt = format_prompt(f"{prompt}", history, system_prompt)

    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)

    output = ""
    for response in stream:
        output += response.token.text

    yield output

additional_inputs = [
    gr.Textbox(label="System Prompt", max_lines=1, interactive=True),
    gr.Slider(label="Temperature", value=0.9, minimum=0.0, maximum=1.0, step=0.05, interactive=True, info="Higher values produce more diverse outputs"),
    gr.Slider(label="Max new tokens", value=9048, minimum=256, maximum=9048, step=64, interactive=True, info="The maximum numbers of new tokens"),
    gr.Slider(label="Top-p (nucleus sampling)", value=0.90, minimum=0.0, maximum=1, step=0.05, interactive=True, info="Higher values sample more low-probability tokens"),
    gr.Slider(label="Repetition penalty", value=1.2, minimum=1.0, maximum=2.0, step=0.05, interactive=True, info="Penalize repeated tokens")
]

avatar_images = ("https://i.postimg.cc/pXjKKVXG/user-circle.png", "https://i.postimg.cc/qq04Yz93/CL3.png")

gr.ChatInterface(
    fn=generate,
    chatbot=gr.Chatbot(show_label=True, show_share_button=False, show_copy_button=True, likeable=True, layout="panel", height="auto", avatar_images=avatar_images),
    additional_inputs=additional_inputs,
    title="ConvoLite",
    submit_btn="➢",
    retry_btn="Retry",
    undo_btn="↩ Undo",
    clear_btn="Clear (New chat)",
    stop_btn="Stop ▢",
    concurrency_limit=20,
    theme=gr.themes.Soft(primary_hue=gr.themes.colors.cyan),
).launch(show_api=False)