File size: 13,902 Bytes
2317674
 
 
 
 
 
 
1cfe513
 
2317674
 
 
0cdbe8f
2317674
f4a0f87
 
1cfe513
 
f4a0f87
 
a188372
f4a0f87
1cfe513
f4a0f87
 
 
 
1cfe513
 
0b984ea
1cfe513
f4a0f87
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1cfe513
f4a0f87
2317674
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1cfe513
 
2317674
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1cfe513
2317674
1cfe513
2317674
1cfe513
 
2317674
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1cfe513
2317674
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1cfe513
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
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
import os
from dotenv import load_dotenv
import gradio as gr
from huggingface_hub import InferenceClient
import pandas as pd
import json
from datetime import datetime
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

# ํ™˜๊ฒฝ ๋ณ€์ˆ˜ ์„ค์ •
HF_TOKEN = os.getenv("HF_TOKEN")
MODEL_ID = "CohereForAI/c4ai-command-r7b-12-2024"

from transformers import pipeline

class ModelManager:
    def __init__(self):
        self.pipe = None
        self.setup_pipeline()
    
    def setup_pipeline(self):
        try:
            print("ํŒŒ์ดํ”„๋ผ์ธ ์ดˆ๊ธฐํ™” ์‹œ์ž‘...")
            self.pipe = pipeline(
                "text-generation",
                model=MODEL_ID,
                token=HF_TOKEN,
                device_map="auto",
                torch_dtype=torch.float16
            )
            print("ํŒŒ์ดํ”„๋ผ์ธ ์ดˆ๊ธฐํ™” ์™„๋ฃŒ")
        except Exception as e:
            print(f"ํŒŒ์ดํ”„๋ผ์ธ ์ดˆ๊ธฐํ™” ์ค‘ ์˜ค๋ฅ˜ ๋ฐœ์ƒ: {e}")
            raise Exception(f"ํŒŒ์ดํ”„๋ผ์ธ ์ดˆ๊ธฐํ™” ์‹คํŒจ: {e}")

    def generate_response(self, messages, max_tokens=4000, temperature=0.7, top_p=0.9):
        try:
            # ๋ฉ”์‹œ์ง€ ํ˜•์‹ ๋ณ€ํ™˜
            prompt = ""
            for msg in messages:
                role = msg["role"]
                content = msg["content"]
                if role == "system":
                    prompt += f"System: {content}\n"
                elif role == "user":
                    prompt += f"User: {content}\n"
                elif role == "assistant":
                    prompt += f"Assistant: {content}\n"
            
            # ์‘๋‹ต ์ƒ์„ฑ
            response = self.pipe(
                prompt,
                max_new_tokens=max_tokens,
                temperature=temperature,
                top_p=top_p,
                do_sample=True,
                num_return_sequences=1,
                pad_token_id=self.pipe.tokenizer.eos_token_id
            )
            
            # ์‘๋‹ต ํ…์ŠคํŠธ ์ถ”์ถœ ๋ฐ ์ŠคํŠธ๋ฆฌ๋ฐ ์‹œ๋ฎฌ๋ ˆ์ด์…˜
            generated_text = response[0]['generated_text'][len(prompt):].strip()
            words = generated_text.split()
            
            # ๋‹จ์–ด ๋‹จ์œ„๋กœ ์ŠคํŠธ๋ฆฌ๋ฐ
            partial_response = ""
            for word in words:
                partial_response += word + " "
                yield type('Response', (), {
                    'choices': [type('Choice', (), {
                        'delta': {'content': word + " "}
                    })()]
                })()
                
        except Exception as e:
            raise Exception(f"์‘๋‹ต ์ƒ์„ฑ ์‹คํŒจ: {e}")

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):
        formatted = []
        for conv in self.history:
            formatted.append([
                conv["messages"][0]["content"],
                conv["messages"][1]["content"]
            ])
        return formatted

    def get_messages_for_api(self):
        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"ํžˆ์Šคํ† ๋ฆฌ ์ €์žฅ ์‹คํŒจ: {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"ํžˆ์Šคํ† ๋ฆฌ ๋กœ๋“œ ์‹คํŒจ: {e}")
            self.history = []

# ์ „์—ญ ์ธ์Šคํ„ด์Šค ์ƒ์„ฑ
chat_history = ChatHistory()
model_manager = ModelManager()

def get_client():
    return InferenceClient(MODEL_ID, 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"๐Ÿ“Š ๋ฐ์ดํ„ฐ์…‹ ๊ตฌ์กฐ: {columns}๊ฐœ ์ปฌ๋Ÿผ, {rows}๊ฐœ ๋ฐ์ดํ„ฐ"
        except:
            return "โŒ ๋ฐ์ดํ„ฐ์…‹ ๊ตฌ์กฐ ๋ถ„์„ ์‹คํŒจ"
    
    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"๐Ÿ’ป ์ฝ”๋“œ ๊ตฌ์กฐ: {total_lines}์ค„ (ํ•จ์ˆ˜: {functions}, ํด๋ž˜์Šค: {classes}, ์ž„ํฌํŠธ: {imports})"
    
    paragraphs = content.count('\n\n') + 1
    words = len(content.split())
    return f"๐Ÿ“ ๋ฌธ์„œ ๊ตฌ์กฐ: {total_lines}์ค„, {paragraphs}๋‹จ๋ฝ, ์•ฝ {words}๋‹จ์–ด"

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"๐Ÿ“Š ๋ฐ์ดํ„ฐ ๋ฏธ๋ฆฌ๋ณด๊ธฐ:\n{df.head(10).to_markdown(index=False)}\n\n"
                    content += f"\n๐Ÿ“ˆ ๋ฐ์ดํ„ฐ ์ •๋ณด:\n"
                    content += f"- ์ „์ฒด ํ–‰ ์ˆ˜: {len(df)}\n"
                    content += f"- ์ „์ฒด ์—ด ์ˆ˜: {len(df.columns)}\n"
                    content += f"- ์ปฌ๋Ÿผ ๋ชฉ๋ก: {', '.join(df.columns)}\n"
                    content += f"\n๐Ÿ“‹ ์ปฌ๋Ÿผ ๋ฐ์ดํ„ฐ ํƒ€์ž…:\n"
                    for col, dtype in df.dtypes.items():
                        content += f"- {col}: {dtype}\n"
                    null_counts = df.isnull().sum()
                    if null_counts.any():
                        content += f"\nโš ๏ธ ๊ฒฐ์ธก์น˜:\n"
                        for col, null_count in null_counts[null_counts > 0].items():
                            content += f"- {col}: {null_count}๊ฐœ ๋ˆ„๋ฝ\n"
                    return content, "csv"
                except UnicodeDecodeError:
                    continue
            raise UnicodeDecodeError(f"โŒ ์ง€์›๋˜๋Š” ์ธ์ฝ”๋”ฉ์œผ๋กœ ํŒŒ์ผ์„ ์ฝ์„ ์ˆ˜ ์—†์Šต๋‹ˆ๋‹ค ({', '.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"โŒ ์ง€์›๋˜๋Š” ์ธ์ฝ”๋”ฉ์œผ๋กœ ํŒŒ์ผ์„ ์ฝ์„ ์ˆ˜ ์—†์Šต๋‹ˆ๋‹ค ({', '.join(encodings)})")
    except Exception as e:
        return f"โŒ ํŒŒ์ผ ์ฝ๊ธฐ ์˜ค๋ฅ˜: {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 = """์ €๋Š” ์—ฌ๋Ÿฌ๋ถ„์˜ ์นœ๊ทผํ•˜๊ณ  ์ง€์ ์ธ AI ์–ด์‹œ์Šคํ„ดํŠธ 'GiniGEN'์ž…๋‹ˆ๋‹ค.. ๋‹ค์Œ๊ณผ ๊ฐ™์€ ์›์น™์œผ๋กœ ์†Œํ†ตํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค:
1. ๐Ÿค ์นœ๊ทผํ•˜๊ณ  ๊ณต๊ฐ์ ์ธ ํƒœ๋„๋กœ ๋Œ€ํ™”
2. ๐Ÿ’ก ๋ช…ํ™•ํ•˜๊ณ  ์ดํ•ดํ•˜๊ธฐ ์‰ฌ์šด ์„ค๋ช… ์ œ๊ณต
3. ๐ŸŽฏ ์งˆ๋ฌธ์˜ ์˜๋„๋ฅผ ์ •ํ™•ํžˆ ํŒŒ์•…ํ•˜์—ฌ ๋งž์ถคํ˜• ๋‹ต๋ณ€
4. ๐Ÿ“š ํ•„์š”ํ•œ ๊ฒฝ์šฐ ์—…๋กœ๋“œ๋œ ํŒŒ์ผ ๋‚ด์šฉ์„ ์ฐธ๊ณ ํ•˜์—ฌ ๊ตฌ์ฒด์ ์ธ ๋„์›€ ์ œ๊ณต
5. โœจ ์ถ”๊ฐ€์ ์ธ ํ†ต์ฐฐ๊ณผ ์ œ์•ˆ์„ ํ†ตํ•œ ๊ฐ€์น˜ ์žˆ๋Š” ๋Œ€ํ™”
ํ•ญ์ƒ ์˜ˆ์˜ ๋ฐ”๋ฅด๊ณ  ์นœ์ ˆํ•˜๊ฒŒ ์‘๋‹ตํ•˜๋ฉฐ, ํ•„์š”ํ•œ ๊ฒฝ์šฐ ๊ตฌ์ฒด์ ์ธ ์˜ˆ์‹œ๋‚˜ ์„ค๋ช…์„ ์ถ”๊ฐ€ํ•˜์—ฌ 
์ดํ•ด๋ฅผ ๋•๊ฒ ์Šต๋‹ˆ๋‹ค."""

    try:
        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\nํŒŒ์ผ ๋‚ด์šฉ:\n```markdown\n{content}\n```"
            else:
                system_message += f"\n\nํŒŒ์ผ ๋‚ด์šฉ:\n```\n{content}\n```"
                
            if message == "ํŒŒ์ผ ๋ถ„์„์„ ์‹œ์ž‘ํ•ฉ๋‹ˆ๋‹ค...":
                message = f"""[ํŒŒ์ผ ๊ตฌ์กฐ ๋ถ„์„] {file_summary}
๋‹ค์Œ ๊ด€์ ์—์„œ ๋„์›€์„ ๋“œ๋ฆฌ๊ฒ ์Šต๋‹ˆ๋‹ค:
1. ๐Ÿ“‹ ์ „๋ฐ˜์ ์ธ ๋‚ด์šฉ ํŒŒ์•…
2. ๐Ÿ’ก ์ฃผ์š” ํŠน์ง• ์„ค๋ช…
3. ๐ŸŽฏ ์‹ค์šฉ์ ์ธ ํ™œ์šฉ ๋ฐฉ์•ˆ
4. โœจ ๊ฐœ์„  ์ œ์•ˆ
5. ๐Ÿ’ฌ ์ถ”๊ฐ€ ์งˆ๋ฌธ์ด๋‚˜ ํ•„์š”ํ•œ ์„ค๋ช…"""

        messages = [{"role": "system", "content": system_prefix + system_message}]
        
        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})

        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

        chat_history.add_conversation(message, partial_message)
        
    except Exception as e:
        error_msg = f"โŒ ์˜ค๋ฅ˜๊ฐ€ ๋ฐœ์ƒํ–ˆ์Šต๋‹ˆ๋‹ค: {str(e)}"
        chat_history.add_conversation(message, error_msg)
        yield "", history + [[message, error_msg]]

with gr.Blocks(theme="Yntec/HaleyCH_Theme_Orange", title="GiniGEN ๐Ÿค–") as demo:
    initial_history = chat_history.format_for_display()
    with gr.Row():
        with gr.Column(scale=2):
            chatbot = gr.Chatbot(
                value=initial_history,
                height=600, 
                label="๋Œ€ํ™”์ฐฝ ๐Ÿ’ฌ",
                show_label=True
            )    

            msg = gr.Textbox(
                label="๋ฉ”์‹œ์ง€ ์ž…๋ ฅ",
                show_label=False,
                placeholder="๋ฌด์—‡์ด๋“  ๋ฌผ์–ด๋ณด์„ธ์š”... ๐Ÿ’ญ",
                container=False
            )
            with gr.Row():
                clear = gr.ClearButton([msg, chatbot], value="๋Œ€ํ™”๋‚ด์šฉ ์ง€์šฐ๊ธฐ")
                send = gr.Button("๋ณด๋‚ด๊ธฐ ๐Ÿ“ค")
        
        with gr.Column(scale=1):
            gr.Markdown("### GiniGEN ๐Ÿค– [ํŒŒ์ผ ์—…๋กœ๋“œ] ๐Ÿ“\n์ง€์› ํ˜•์‹: ํ…์ŠคํŠธ, ์ฝ”๋“œ, CSV, Parquet ํŒŒ์ผ")
            file_upload = gr.File(
                label="ํŒŒ์ผ ์„ ํƒ",
                file_types=["text", ".csv", ".parquet"],
                type="filepath"
            )
            
            with gr.Accordion("๊ณ ๊ธ‰ ์„ค์ • โš™๏ธ", open=False):
                system_message = gr.Textbox(label="์‹œ์Šคํ…œ ๋ฉ”์‹œ์ง€ ๐Ÿ“", value="")
                max_tokens = gr.Slider(minimum=1, maximum=8000, value=4000, label="์ตœ๋Œ€ ํ† ํฐ ์ˆ˜ ๐Ÿ“Š")
                temperature = gr.Slider(minimum=0, maximum=1, value=0.7, label="์ฐฝ์˜์„ฑ ์ˆ˜์ค€ ๐ŸŒก๏ธ")
                top_p = gr.Slider(minimum=0, maximum=1, value=0.9, label="์‘๋‹ต ๋‹ค์–‘์„ฑ ๐Ÿ“ˆ")

    gr.Examples(
        examples=[
            ["์•ˆ๋…•ํ•˜์„ธ์š”! ์–ด๋–ค ๋„์›€์ด ํ•„์š”ํ•˜์‹ ๊ฐ€์š”? ๐Ÿค"],
            ["์ œ๊ฐ€ ์ดํ•ดํ•˜๊ธฐ ์‰ฝ๊ฒŒ ์„ค๋ช…ํ•ด ์ฃผ์‹œ๊ฒ ์–ด์š”? ๐Ÿ“š"],
            ["์ด ๋‚ด์šฉ์„ ์‹ค์ œ๋กœ ์–ด๋–ป๊ฒŒ ํ™œ์šฉํ•  ์ˆ˜ ์žˆ์„๊นŒ์š”? ๐ŸŽฏ"],
            ["์ถ”๊ฐ€๋กœ ์กฐ์–ธํ•ด ์ฃผ์‹ค ๋‚ด์šฉ์ด ์žˆ์œผ์‹ ๊ฐ€์š”? โœจ"],
            ["๊ถ๊ธˆํ•œ ์ ์ด ๋” ์žˆ๋Š”๋ฐ ์—ฌ์ญค๋ด๋„ ๋ ๊นŒ์š”? ๐Ÿค”"],
        ],
        inputs=msg,
    )

    def clear_chat():
        chat_history.clear_history()
        return None, None

    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]
    )

    file_upload.change(
        lambda: "ํŒŒ์ผ ๋ถ„์„์„ ์‹œ์ž‘ํ•ฉ๋‹ˆ๋‹ค...",
        outputs=msg
    ).then(
        chat,
        inputs=[msg, chatbot, file_upload, system_message, max_tokens, temperature, top_p],
        outputs=[msg, chatbot]
    )

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