File size: 7,665 Bytes
f9b9d56
83ee74c
574f73e
705c5b5
83ee74c
f9b9d56
1fd4ab2
ad9db85
1fd4ab2
bee778b
ad9db85
 
bee778b
 
 
 
 
 
f9b9d56
2af89cf
1fd4ab2
99d94e0
da20c1b
99d94e0
 
 
1fd4ab2
 
da20c1b
1fd4ab2
2af89cf
 
 
 
 
da20c1b
2af89cf
 
da20c1b
1fd4ab2
da20c1b
 
 
1fd4ab2
2af89cf
705c5b5
0997082
2af89cf
705c5b5
99d94e0
 
 
705c5b5
2af89cf
 
99d94e0
521288b
 
 
 
1fd4ab2
 
 
 
 
 
521288b
 
 
 
1fd4ab2
521288b
1fd4ab2
521288b
 
 
1fd4ab2
705c5b5
521288b
 
 
 
 
 
 
 
1fd4ab2
705c5b5
1fd4ab2
0997082
d57197f
 
 
 
 
 
 
 
4fec5a3
bee778b
4fec5a3
da20c1b
4fec5a3
 
 
 
 
705c5b5
4fec5a3
 
705c5b5
4fec5a3
 
1fd4ab2
 
 
4fec5a3
 
1fd4ab2
 
 
4fec5a3
63c5e29
 
b45e256
 
4fec5a3
a3b8f2f
b45e256
7ffca43
4fec5a3
7ffca43
4fec5a3
7ffca43
 
b45e256
4fec5a3
 
63c5e29
 
4fec5a3
7ffca43
4fec5a3
99d94e0
d57197f
7ffca43
63c5e29
4fec5a3
 
 
 
 
63c5e29
4fec5a3
63c5e29
 
4fec5a3
 
 
 
 
 
63c5e29
 
 
f9b9d56
 
b45e256
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
import gradio as gr
from huggingface_hub import InferenceClient
import os
import pandas as pd
from typing import List, Tuple

# 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
}

def get_client(model_name="Cohere c4ai-crp-08-2024"):
    try:
        return InferenceClient(LLM_MODELS[model_name], token=os.getenv("HF_TOKEN"))
    except Exception:
        # If primary model fails, try backup model
        return InferenceClient(LLM_MODELS["Meta Llama3.3-70B"], token=os.getenv("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 Dataset Structure: {columns} columns, {rows} data samples"
        except:
            return "❌ Dataset structure analysis failed"
    
    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’» Code Structure: {total_lines} lines (Functions: {functions}, Classes: {classes}, Imports: {imports})"
    
    paragraphs = content.count('\n\n') + 1
    words = len(content.split())
    return f"πŸ“ Document Structure: {total_lines} lines, {paragraphs} paragraphs, ~{words} words"

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 Data Preview:\n{df.head(10).to_markdown(index=False)}\n\n"
                    content += f"\nπŸ“ˆ Data Information:\n"
                    content += f"- Total Rows: {len(df)}\n"
                    content += f"- Total Columns: {len(df.columns)}\n"
                    content += f"- Column List: {', '.join(df.columns)}\n"
                    content += f"\nπŸ“‹ Column Data Types:\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⚠️ Missing Values:\n"
                        for col, null_count in null_counts[null_counts > 0].items():
                            content += f"- {col}: {null_count} missing\n"
                    return content, "csv"
                except UnicodeDecodeError:
                    continue
            raise UnicodeDecodeError(f"❌ Unable to read file with supported encodings ({', '.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"❌ Unable to read file with supported encodings ({', '.join(encodings)})")
    except Exception as e:
        return f"❌ Error reading file: {str(e)}", "error"

def format_history(history):
    formatted_history = []
    for user_msg, assistant_msg in history:
        formatted_history.append({"role": "user", "content": user_msg})
        if assistant_msg:
            formatted_history.append({"role": "assistant", "content": assistant_msg})
    return formatted_history

# μ‹œμŠ€ν…œ ν”„λ‘¬ν”„νŠΈ μˆ˜μ •
def chat(message, history, uploaded_file, system_message="", max_tokens=4000, temperature=0.7, top_p=0.9):
    system_prefix = """μ €λŠ” μ—¬λŸ¬λΆ„μ˜ μΉœκ·Όν•˜κ³  지적인 AI μ–΄μ‹œμŠ€ν„΄νŠΈμž…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‹€."""

# UI ν…μŠ€νŠΈ ν•œκΈ€ν™”
with gr.Blocks(theme="Yntec/HaleyCH_Theme_Orange", title="AI μ–΄μ‹œμŠ€ν„΄νŠΈ πŸ€–") as demo:
    gr.HTML(
        """
        <div style="text-align: center; max-width: 800px; margin: 0 auto;">
            <h1 style="font-size: 3em; font-weight: 600; margin: 0.5em;">AI μ–΄μ‹œμŠ€ν„΄νŠΈ πŸ€–</h1>
            <h3 style="font-size: 1.2em; margin: 1em;">λ‹Ήμ‹ μ˜ λ“ λ“ ν•œ λŒ€ν™” νŒŒνŠΈλ„ˆ πŸ’¬</h3>
        </div>
        """
    )

    with gr.Row():
        with gr.Column(scale=2):
            chatbot = gr.Chatbot(
                height=600, 
                label="λŒ€ν™”μ°½ πŸ’¬",
                type="messages"
            )
            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("### 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 μˆ˜μ •
    gr.Examples(
        examples=[
            ["μ•ˆλ…•ν•˜μ„Έμš”! μ–΄λ–€ 도움이 ν•„μš”ν•˜μ‹ κ°€μš”? 🀝"],
            ["이 λ‚΄μš©μ— λŒ€ν•΄ μ’€ 더 μžμ„Ένžˆ μ„€λͺ…ν•΄ μ£Όμ‹€ 수 μžˆλ‚˜μš”? πŸ’‘"],
            ["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,
    )

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