File size: 3,551 Bytes
ce5b5d6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
import os
import gradio as gr
from dotenv import load_dotenv
from openai import OpenAI
from prompts.initial_prompt import INITIAL_PROMPT
from prompts.main_prompt import MAIN_PROMPT

# .env ํŒŒ์ผ์—์„œ OPENAI_API_KEY ๋กœ๋“œ
if os.path.exists(".env"):
    load_dotenv(".env")

OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")

client = OpenAI(api_key=OPENAI_API_KEY)



def gpt_call(history, user_message,
             model="gpt-4o-mini",
             max_tokens=512,
             temperature=0.7,
             top_p=0.95):
    """
    OpenAI ChatCompletion API๋ฅผ ํ†ตํ•ด ๋‹ต๋ณ€์„ ์ƒ์„ฑํ•˜๋Š” ํ•จ์ˆ˜.
    - history: [(user_text, assistant_text), ...]
    - user_message: ์‚ฌ์šฉ์ž๊ฐ€ ๋ฐฉ๊ธˆ ์ž…๋ ฅํ•œ ๋ฉ”์‹œ์ง€
    """
    # 1) ์‹œ์Šคํ…œ ๋ฉ”์‹œ์ง€(=MAIN_PROMPT)๋ฅผ ๊ฐ€์žฅ ์•ž์— ์ถ”๊ฐ€
    messages = [{"role": "system", "content": MAIN_PROMPT}]
    
    # 2) ๊ธฐ์กด ๋Œ€ํ™” ๊ธฐ๋ก(history)์„ OpenAI ํ˜•์‹์œผ๋กœ ๋ณ€ํ™˜
    #    user_text -> 'user' / assistant_text -> 'assistant'
    for user_text, assistant_text in history:
        if user_text:
            messages.append({"role": "user", "content": user_text})
        if assistant_text:
            messages.append({"role": "assistant", "content": assistant_text})

    # 3) ๋งˆ์ง€๋ง‰์— ์ด๋ฒˆ ์‚ฌ์šฉ์ž์˜ ์ž…๋ ฅ์„ ์ถ”๊ฐ€
    messages.append({"role": "user", "content": user_message})

    # 4) OpenAI API ํ˜ธ์ถœ
    completion = client.chat.completions.create(
        model=model,
        messages=messages,
        max_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p
    )
    return completion.choices[0].message.content

def respond(user_message, history):
    """
    Gradio ์ƒ์—์„œ submitํ•  ๋•Œ ํ˜ธ์ถœ๋˜๋Š” ํ•จ์ˆ˜
    - user_message: ์‚ฌ์šฉ์ž๊ฐ€ ๋ฐฉ๊ธˆ ์นœ ๋ฉ”์‹œ์ง€
    - history: ๊ธฐ์กด (user, assistant) ํŠœํ”Œ ๋ฆฌ์ŠคํŠธ
    """
    # ์‚ฌ์šฉ์ž๊ฐ€ ๋นˆ ๋ฌธ์ž์—ด์„ ๋ณด๋ƒˆ๋‹ค๋ฉด ์•„๋ฌด ์ผ๋„ ํ•˜์ง€ ์•Š์Œ
    if not user_message:
        return "", history

    # GPT ๋ชจ๋ธ๋กœ๋ถ€ํ„ฐ ์‘๋‹ต์„ ๋ฐ›์Œ
    assistant_reply = gpt_call(history, user_message)

    # history์— (user, assistant) ์Œ ์ถ”๊ฐ€
    history.append((user_message, assistant_reply))

    # Gradio์—์„œ๋Š” (์ƒˆ๋กœ ๋น„์›Œ์งˆ ์ž…๋ ฅ์ฐฝ, ๊ฐฑ์‹ ๋œ history)๋ฅผ ๋ฐ˜ํ™˜
    return "", history

##############################
#  Gradio Blocks UI
##############################
with gr.Blocks() as demo:
    gr.Markdown("## Simple Chat Interface")

    # Chatbot ์ดˆ๊ธฐ ์ƒํƒœ๋ฅผ ์„ค์ •
    # ์ฒซ ๋ฒˆ์งธ ๋ฉ”์‹œ์ง€๋Š” (user="", assistant=INITIAL_PROMPT) ํ˜•ํƒœ๋กœ ๋„ฃ์–ด
    # ํ™”๋ฉด์ƒ์—์„œ 'assistant'๊ฐ€ INITIAL_PROMPT๋ฅผ ๋งํ•œ ๊ฒƒ์ฒ˜๋Ÿผ ๋ณด์ด๊ฒŒ ํ•จ
    chatbot = gr.Chatbot(
        value=[("", INITIAL_PROMPT)],  # (user, assistant)
        height=500
    )

    # (user, assistant) ์Œ์„ ์ €์žฅํ•  ํžˆ์Šคํ† ๋ฆฌ ์ƒํƒœ
    # ์—ฌ๊ธฐ์„œ๋„ ๋™์ผํ•œ ์ดˆ๊ธฐ ์ƒํƒœ๋ฅผ ๋„ฃ์–ด์คŒ
    state_history = gr.State([("", INITIAL_PROMPT)])

    # ์‚ฌ์šฉ์ž ์ž…๋ ฅ
    user_input = gr.Textbox(
        placeholder="Type your message here...",
        label="Your Input"
    )

    # ์ž…๋ ฅ์ด submit๋˜๋ฉด respond() ํ˜ธ์ถœ โ†’ ์ถœ๋ ฅ์€ (์ƒˆ ์ž…๋ ฅ์ฐฝ, ๊ฐฑ์‹ ๋œ chatbot)
    user_input.submit(
        respond,
        inputs=[user_input, state_history],
        outputs=[user_input, chatbot]
    ).then(
        # respond ๋๋‚œ ๋’ค, ์ตœ์‹  history๋ฅผ state_history์— ๋ฐ˜์˜
        fn=lambda _, h: h,
        inputs=[user_input, chatbot],
        outputs=[state_history]
    )

# ๋ฉ”์ธ ์‹คํ–‰
if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860, share=True)