File size: 12,454 Bytes
eca1723
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
import os
from dotenv import load_dotenv
import gradio as gr
from huggingface_hub import InferenceClient
import pandas as pd
from typing import List, Tuple
import json
from datetime import datetime

# ν™˜κ²½ λ³€μˆ˜ μ„€μ •
HF_TOKEN = os.getenv("HF_TOKEN")

# LLM Models Definition
LLM_MODELS = {
    "Cohere c4ai-crp-08-2024": "CohereForAI/c4ai-command-r-plus-08-2024",  # Default
    "Meta Llama3.3-70B": "meta-llama/Llama-3.3-70B-Instruct"    # Backup model
}

class ChatHistory:
    def __init__(self):
        self.history = []
        self.history_file = "/tmp/chat_history.json"
        self.load_history()

    def add_conversation(self, user_msg: str, assistant_msg: str):
        conversation = {
            "timestamp": datetime.now().isoformat(),
            "messages": [
                {"role": "user", "content": user_msg},
                {"role": "assistant", "content": assistant_msg}
            ]
        }
        self.history.append(conversation)
        self.save_history()

    def format_for_display(self):
        # Gradio Chatbot μ»΄ν¬λ„ŒνŠΈμ— λ§žλŠ” ν˜•μ‹μœΌλ‘œ λ³€ν™˜
        formatted = []
        for conv in self.history:
            formatted.append([
                conv["messages"][0]["content"],  # user message
                conv["messages"][1]["content"]   # assistant message
            ])
        return formatted

    def get_messages_for_api(self):
        # API ν˜ΈμΆœμ„ μœ„ν•œ λ©”μ‹œμ§€ ν˜•μ‹
        messages = []
        for conv in self.history:
            messages.extend([
                {"role": "user", "content": conv["messages"][0]["content"]},
                {"role": "assistant", "content": conv["messages"][1]["content"]}
            ])
        return messages

    def clear_history(self):
        self.history = []
        self.save_history()

    def save_history(self):
        try:
            with open(self.history_file, 'w', encoding='utf-8') as f:
                json.dump(self.history, f, ensure_ascii=False, indent=2)
        except Exception as e:
            print(f"νžˆμŠ€ν† λ¦¬ μ €μž₯ μ‹€νŒ¨: {e}")

    def load_history(self):
        try:
            if os.path.exists(self.history_file):
                with open(self.history_file, 'r', encoding='utf-8') as f:
                    self.history = json.load(f)
        except Exception as e:
            print(f"νžˆμŠ€ν† λ¦¬ λ‘œλ“œ μ‹€νŒ¨: {e}")
            self.history = []


# μ „μ—­ ChatHistory μΈμŠ€ν„΄μŠ€ 생성
chat_history = ChatHistory()

def get_client(model_name="Cohere c4ai-crp-08-2024"):
    try:
        return InferenceClient(LLM_MODELS[model_name], token=HF_TOKEN)
    except Exception:
        return InferenceClient(LLM_MODELS["Meta Llama3.3-70B"], token=HF_TOKEN)

def analyze_file_content(content, file_type):
    """Analyze file content and return structural summary"""
    if file_type in ['parquet', 'csv']:
        try:
            lines = content.split('\n')
            header = lines[0]
            columns = header.count('|') - 1
            rows = len(lines) - 3
            return f"πŸ“Š 데이터셋 ꡬ쑰: {columns}개 컬럼, {rows}개 데이터"
        except:
            return "❌ 데이터셋 ꡬ쑰 뢄석 μ‹€νŒ¨"
    
    lines = content.split('\n')
    total_lines = len(lines)
    non_empty_lines = len([line for line in lines if line.strip()])
    
    if any(keyword in content.lower() for keyword in ['def ', 'class ', 'import ', 'function']):
        functions = len([line for line in lines if 'def ' in line])
        classes = len([line for line in lines if 'class ' in line])
        imports = len([line for line in lines if 'import ' in line or 'from ' in line])
        return f"πŸ’» μ½”λ“œ ꡬ쑰: {total_lines}쀄 (ν•¨μˆ˜: {functions}, 클래슀: {classes}, μž„ν¬νŠΈ: {imports})"
    
    paragraphs = content.count('\n\n') + 1
    words = len(content.split())
    return f"πŸ“ λ¬Έμ„œ ꡬ쑰: {total_lines}쀄, {paragraphs}단락, μ•½ {words}단어"

def read_uploaded_file(file):
    if file is None:
        return "", ""
    try:
        file_ext = os.path.splitext(file.name)[1].lower()
        
        if file_ext == '.parquet':
            df = pd.read_parquet(file.name, engine='pyarrow')
            content = df.head(10).to_markdown(index=False)
            return content, "parquet"
        elif file_ext == '.csv':
            encodings = ['utf-8', 'cp949', 'euc-kr', 'latin1']
            for encoding in encodings:
                try:
                    df = pd.read_csv(file.name, encoding=encoding)
                    content = f"πŸ“Š 데이터 미리보기:\n{df.head(10).to_markdown(index=False)}\n\n"
                    content += f"\nπŸ“ˆ 데이터 정보:\n"
                    content += f"- 전체 ν–‰ 수: {len(df)}\n"
                    content += f"- 전체 μ—΄ 수: {len(df.columns)}\n"
                    content += f"- 컬럼 λͺ©λ‘: {', '.join(df.columns)}\n"
                    content += f"\nπŸ“‹ 컬럼 데이터 νƒ€μž…:\n"
                    for col, dtype in df.dtypes.items():
                        content += f"- {col}: {dtype}\n"
                    null_counts = df.isnull().sum()
                    if null_counts.any():
                        content += f"\n⚠️ 결츑치:\n"
                        for col, null_count in null_counts[null_counts > 0].items():
                            content += f"- {col}: {null_count}개 λˆ„λ½\n"
                    return content, "csv"
                except UnicodeDecodeError:
                    continue
            raise UnicodeDecodeError(f"❌ μ§€μ›λ˜λŠ” μΈμ½”λ”©μœΌλ‘œ νŒŒμΌμ„ 읽을 수 μ—†μŠ΅λ‹ˆλ‹€ ({', '.join(encodings)})")
        else:
            encodings = ['utf-8', 'cp949', 'euc-kr', 'latin1']
            for encoding in encodings:
                try:
                    with open(file.name, 'r', encoding=encoding) as f:
                        content = f.read()
                    return content, "text"
                except UnicodeDecodeError:
                    continue
            raise UnicodeDecodeError(f"❌ μ§€μ›λ˜λŠ” μΈμ½”λ”©μœΌλ‘œ νŒŒμΌμ„ 읽을 수 μ—†μŠ΅λ‹ˆλ‹€ ({', '.join(encodings)})")
    except Exception as e:
        return f"❌ 파일 읽기 였λ₯˜: {str(e)}", "error"

def chat(message, history, uploaded_file, system_message="", max_tokens=4000, temperature=0.7, top_p=0.9):
    if not message:
        return "", history

    system_prefix = """μ €λŠ” μ—¬λŸ¬λΆ„μ˜ μΉœκ·Όν•˜κ³  지적인 AI μ–΄μ‹œμŠ€ν„΄νŠΈ 'GiniGEN'μž…λ‹ˆλ‹€.. λ‹€μŒκ³Ό 같은 μ›μΉ™μœΌλ‘œ μ†Œν†΅ν•˜κ² μŠ΅λ‹ˆλ‹€:
1. 🀝 μΉœκ·Όν•˜κ³  곡감적인 νƒœλ„λ‘œ λŒ€ν™”
2. πŸ’‘ λͺ…ν™•ν•˜κ³  μ΄ν•΄ν•˜κΈ° μ‰¬μš΄ μ„€λͺ… 제곡
3. 🎯 질문의 μ˜λ„λ₯Ό μ •ν™•νžˆ νŒŒμ•…ν•˜μ—¬ λ§žμΆ€ν˜• λ‹΅λ³€
4. πŸ“š ν•„μš”ν•œ 경우 μ—…λ‘œλ“œλœ 파일 λ‚΄μš©μ„ μ°Έκ³ ν•˜μ—¬ ꡬ체적인 도움 제곡
5. ✨ 좔가적인 톡찰과 μ œμ•ˆμ„ ν†΅ν•œ κ°€μΉ˜ μžˆλŠ” λŒ€ν™”

항상 예의 λ°”λ₯΄κ³  μΉœμ ˆν•˜κ²Œ μ‘λ‹΅ν•˜λ©°, ν•„μš”ν•œ 경우 ꡬ체적인 μ˜ˆμ‹œλ‚˜ μ„€λͺ…을 μΆ”κ°€ν•˜μ—¬ 
이해λ₯Ό λ•κ² μŠ΅λ‹ˆλ‹€."""

    try:
        # 파일 μ—…λ‘œλ“œ 처리
        if uploaded_file:
            content, file_type = read_uploaded_file(uploaded_file)
            if file_type == "error":
                error_message = content
                chat_history.add_conversation(message, error_message)
                return "", history + [[message, error_message]]
            
            file_summary = analyze_file_content(content, file_type)
            
            if file_type in ['parquet', 'csv']:
                system_message += f"\n\n파일 λ‚΄μš©:\n```markdown\n{content}\n```"
            else:
                system_message += f"\n\n파일 λ‚΄μš©:\n```\n{content}\n```"
                
            if message == "파일 뢄석을 μ‹œμž‘ν•©λ‹ˆλ‹€...":
                message = f"""[파일 ꡬ쑰 뢄석] {file_summary}
λ‹€μŒ κ΄€μ μ—μ„œ 도움을 λ“œλ¦¬κ² μŠ΅λ‹ˆλ‹€:
1. πŸ“‹ μ „λ°˜μ μΈ λ‚΄μš© νŒŒμ•…
2. πŸ’‘ μ£Όμš” νŠΉμ§• μ„€λͺ…
3. 🎯 μ‹€μš©μ μΈ ν™œμš© λ°©μ•ˆ
4. ✨ κ°œμ„  μ œμ•ˆ
5. πŸ’¬ μΆ”κ°€ μ§ˆλ¬Έμ΄λ‚˜ ν•„μš”ν•œ μ„€λͺ…"""

        # λ©”μ‹œμ§€ 처리
        messages = [{"role": "system", "content": system_prefix + system_message}]
        
        # 이전 λŒ€ν™” νžˆμŠ€ν† λ¦¬ μΆ”κ°€
        if history:
            for user_msg, assistant_msg in history:
                messages.append({"role": "user", "content": user_msg})
                messages.append({"role": "assistant", "content": assistant_msg})
        
        messages.append({"role": "user", "content": message})

        # API 호좜 및 응닡 처리
        client = get_client()
        partial_message = ""
        
        for msg in client.chat_completion(
            messages,
            max_tokens=max_tokens,
            stream=True,
            temperature=temperature,
            top_p=top_p,
        ):
            token = msg.choices[0].delta.get('content', None)
            if token:
                partial_message += token
                current_history = history + [[message, partial_message]]
                yield "", current_history

        # μ™„μ„±λœ λŒ€ν™” μ €μž₯
        chat_history.add_conversation(message, partial_message)
        
    except Exception as e:
        error_msg = f"❌ 였λ₯˜κ°€ λ°œμƒν–ˆμŠ΅λ‹ˆλ‹€: {str(e)}"
        chat_history.add_conversation(message, error_msg)
        yield "", history + [[message, error_msg]]

with gr.Blocks(theme="Yntec/HaleyCH_Theme_Orange", title="GiniGEN πŸ€–") as demo:
    # κΈ°μ‘΄ νžˆμŠ€ν† λ¦¬ λ‘œλ“œ
    initial_history = chat_history.format_for_display()
    with gr.Row():
        with gr.Column(scale=2):
            chatbot = gr.Chatbot(
                value=initial_history,  # μ €μž₯된 νžˆμŠ€ν† λ¦¬λ‘œ μ΄ˆκΈ°ν™”
                height=600, 
                label="λŒ€ν™”μ°½ πŸ’¬",
                show_label=True
            )    


            msg = gr.Textbox(
                label="λ©”μ‹œμ§€ μž…λ ₯",
                show_label=False,
                placeholder="무엇이든 λ¬Όμ–΄λ³΄μ„Έμš”... πŸ’­",
                container=False
            )
            with gr.Row():
                clear = gr.ClearButton([msg, chatbot], value="λŒ€ν™”λ‚΄μš© μ§€μš°κΈ°")
                send = gr.Button("보내기 πŸ“€")
        
        with gr.Column(scale=1):
            gr.Markdown("### GiniGEN πŸ€– [파일 μ—…λ‘œλ“œ] πŸ“\n지원 ν˜•μ‹: ν…μŠ€νŠΈ, μ½”λ“œ, CSV, Parquet 파일")
            file_upload = gr.File(
                label="파일 선택",
                file_types=["text", ".csv", ".parquet"],
                type="filepath"
            )
            
            with gr.Accordion("κ³ κΈ‰ μ„€μ • βš™οΈ", open=False):
                system_message = gr.Textbox(label="μ‹œμŠ€ν…œ λ©”μ‹œμ§€ πŸ“", value="")
                max_tokens = gr.Slider(minimum=1, maximum=8000, value=4000, label="μ΅œλŒ€ 토큰 수 πŸ“Š")
                temperature = gr.Slider(minimum=0, maximum=1, value=0.7, label="μ°½μ˜μ„± μˆ˜μ€€ 🌑️")
                top_p = gr.Slider(minimum=0, maximum=1, value=0.9, label="응닡 λ‹€μ–‘μ„± πŸ“ˆ")

    # μ˜ˆμ‹œ 질문
    gr.Examples(
        examples=[
            ["μ•ˆλ…•ν•˜μ„Έμš”! μ–΄λ–€ 도움이 ν•„μš”ν•˜μ‹ κ°€μš”? 🀝"],
            ["μ œκ°€ μ΄ν•΄ν•˜κΈ° μ‰½κ²Œ μ„€λͺ…ν•΄ μ£Όμ‹œκ² μ–΄μš”? πŸ“š"],
            ["이 λ‚΄μš©μ„ μ‹€μ œλ‘œ μ–΄λ–»κ²Œ ν™œμš©ν•  수 μžˆμ„κΉŒμš”? 🎯"],
            ["μΆ”κ°€λ‘œ μ‘°μ–Έν•΄ μ£Όμ‹€ λ‚΄μš©μ΄ μžˆμœΌμ‹ κ°€μš”? ✨"],
            ["κΆκΈˆν•œ 점이 더 μžˆλŠ”λ° 여쭀봐도 λ κΉŒμš”? πŸ€”"],
        ],
        inputs=msg,
    )

    # λŒ€ν™”λ‚΄μš© μ§€μš°κΈ° λ²„νŠΌμ— νžˆμŠ€ν† λ¦¬ μ΄ˆκΈ°ν™” κΈ°λŠ₯ μΆ”κ°€
    def clear_chat():
        chat_history.clear_history()
        return None, None

    # 이벀트 바인딩
    msg.submit(
        chat,
        inputs=[msg, chatbot, file_upload, system_message, max_tokens, temperature, top_p],
        outputs=[msg, chatbot]
    )

    send.click(
        chat,
        inputs=[msg, chatbot, file_upload, system_message, max_tokens, temperature, top_p],
        outputs=[msg, chatbot]
    )

    clear.click(
        clear_chat,
        outputs=[msg, chatbot]
    )

    # 파일 μ—…λ‘œλ“œμ‹œ μžλ™ 뢄석
    file_upload.change(
        lambda: "파일 뢄석을 μ‹œμž‘ν•©λ‹ˆλ‹€...",
        outputs=msg
    ).then(
        chat,
        inputs=[msg, chatbot, file_upload, system_message, max_tokens, temperature, top_p],
        outputs=[msg, chatbot]
    )

if __name__ == "__main__":
    demo.launch()