Update app.py

app.py

@@ -2,12 +2,13 @@ import gradio as gr
 from huggingface_hub import InferenceClient
 import os
 import pandas as pd
+import pdfplumber
 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",    
+    "Meta Llama3.3-70B": "meta-llama/Llama-3.3-70B-Instruct",
     "Mistral Nemo 2407": "mistralai/Mistral-Nemo-Instruct-2407",
     "Alibaba Qwen QwQ-32B": "Qwen/QwQ-32B-Preview"
 }
@@ -17,11 +18,18 @@ def get_client(model_name):
 
 def analyze_file_content(content, file_type):
     """Analyze file content and return structural summary"""
-    if file_type in ['parquet', 'csv']:
+    if file_type in ['parquet', 'csv', 'pdf']:
         try:
-            lines = content.split('\n')
+            if file_type == 'pdf':
+                with pdfplumber.open(content) as pdf:
+                    pages = pdf.pages
+                    lines = []
+                    for page in pages:
+                        lines.extend(page.extract_text().split('\n'))
+            else:
+                lines = content.split('\n')
             header = lines[0]
-            columns = header.count('|') - 1
+            columns = len(header.split('|')) - 1
             rows = len(lines) - 3
             return f"📊 Dataset Structure: {columns} columns, {rows} data samples"
         except:
@@ -47,39 +55,36 @@ def read_uploaded_file(file):
     try:
         file_ext = os.path.splitext(file.name)[1].lower()
         
-        if file_ext == '.parquet':
-            df = pd.read_parquet(file.name, engine='pyarrow')
+        if file_ext in ['.parquet', '.pdf']:
+            if file_ext == '.parquet':
+                df = pd.read_parquet(file.name, engine='pyarrow')
+            else:
+                df = pd.read_csv(file.name, encoding='utf-8', engine='python')  # Use 'python' engine to handle PDF files
             content = df.head(10).to_markdown(index=False)
-            return content, "parquet"
+            return content, file_ext
         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"📊 Data Preview:\n{df.head(10).to_markdown(index=False)}\n\n"
-                    content += f"\n📈 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"\n📋 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)})")
+            df = pd.read_csv(file.name)
+            content = f"📊 Data Preview:\n{df.head(10).to_markdown(index=False)}\n\n"
+            content += f"\n📈 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"\n📋 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, file_ext
         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"
+                    return content, file_ext
                 except UnicodeDecodeError:
                     continue
             raise UnicodeDecodeError(f"❌ Unable to read file with supported encodings ({', '.join(encodings)})")
@@ -96,45 +101,55 @@ def format_history(history):
 
 def chat(message, history, uploaded_file, model_name, system_message="", max_tokens=4000, temperature=0.7, top_p=0.9):
     system_prefix = """You are a file analysis expert. Analyze the uploaded file in depth from the following perspectives:
-1. 📋 Overall structure and composition
-2. 📊 Key content and pattern analysis
-3. 📈 Data characteristics and meaning
-   - For datasets: Column meanings, data types, value distributions
-   - For text/code: Structural features, main patterns
-4. 💡 Potential applications
-5. ✨ Data quality and areas for improvement
-Provide detailed and structured analysis from an expert perspective, but explain in an easy-to-understand way. Format the analysis results in Markdown and include specific examples where possible."""
+   1. 📋 Overall structure and composition
+   2. 📊 Key content and pattern analysis
+   3. 📈 Data characteristics and meaning
+      - For datasets: Column meanings, data types, value distributions
+      - For text/code: Structural features, main patterns
+   4. 💡 Potential applications
+   5. ✨ Data quality and areas for improvement
+   Provide detailed and structured analysis from an expert perspective, but explain in an easy-to-understand way. Format the analysis results in Markdown and include specific examples where possible."""
 
     if uploaded_file:
         content, file_type = read_uploaded_file(uploaded_file)
         if file_type == "error":
-            yield "", history + [[message, content]]
-            return
+            return "", [{"role": "user", "content": message}, {"role": "assistant", "content": content}]
         
         file_summary = analyze_file_content(content, file_type)
         
-        if file_type in ['parquet', 'csv']:
+        if file_type in ['parquet', 'csv', 'pdf']:
             system_message += f"\n\nFile Content:\n```markdown\n{content}\n```"
         else:
             system_message += f"\n\nFile Content:\n```\n{content}\n```"
             
         if message == "Starting file analysis...":
-            message = f"""[Structure Analysis] {file_summary}
-Please provide detailed analysis from these perspectives:
-1. 📋 Overall file structure and format
-2. 📊 Key content and component analysis
-3. 📈 Data/content characteristics and patterns
-4. ⭐ Quality and completeness evaluation
-5. 💡 Suggested improvements
-6. 🎯 Practical applications and recommendations"""
+            message = f"""[구조 분석] {file_summary}
+자세히 분석해주세요:
+1. 📋 전체 구조 및 형식
+2. 📊 주요 내용 및 구성요소 분석
+3. 📈 데이터/내용의 특성 및 패턴
+4. ⭐ 품질 및 완전성 평가
+5. 💡 제안하는 개선점
+6. 🎯 실용적인 활용 및 권장사항"""
 
     messages = [{"role": "system", "content": f"{system_prefix} {system_message}"}]
-    messages.extend(format_history(history))
+    
+    # Convert history to message format
+    if history is not None:
+        for item in history:
+            if isinstance(item, dict):
+                messages.append(item)
+            elif isinstance(item, (list, tuple)) and len(item) == 2:
+                messages.append({"role": "user", "content": item[0]})
+                if item[1]:
+                    messages.append({"role": "assistant", "content": item[1]})
+
     messages.append({"role": "user", "content": message})
 
     try:
         client = get_client(model_name)
         partial_message = ""
+        current_history = []
         
         for msg in client.chat_completion(
             messages,
@@ -146,11 +161,19 @@ Please provide detailed analysis from these perspectives:
             token = msg.choices[0].delta.get('content', None)
             if token:
                 partial_message += token
-                yield "", history + [[message, partial_message]]
+                current_history = [
+                    {"role": "user", "content": message},
+                    {"role": "assistant", "content": partial_message}
+                ]
+                yield "", current_history
                 
     except Exception as e:
         error_msg = f"❌ Inference error: {str(e)}"
-        yield "", history + [[message, error_msg]]
+        error_history = [
+            {"role": "user", "content": message},
+            {"role": "assistant", "content": error_msg}
+        ]
+        yield "", error_history
 
 css = """
 footer {visibility: hidden}
@@ -172,38 +195,36 @@ with gr.Blocks(theme="Yntec/HaleyCH_Theme_Orange", css=css, title="EveryChat
         with gr.Column(scale=2):
             chatbot = gr.Chatbot(
                 height=600, 
-                label="Chat Interface 💬",
-                type="messages"  # 경고 해결을 위해 type 지정
+                label="채팅 인터페이스 💬",
+                type="messages"
             )
             msg = gr.Textbox(
-                label="Type your message",
+                label="메시지를 입력하세요",
                 show_label=False,
-                placeholder="Ask me anything about the uploaded file... 💭",
+                placeholder="업로드된 파일에 대해 물어보세요... 💭",
                 container=False
             )
-            with gr.Row():
-                clear = gr.ClearButton([msg, chatbot])  # label ��거
-                send = gr.Button("Send 📤")
+            send = gr.Button("전송 📤")
         
         with gr.Column(scale=1):
             model_name = gr.Radio(
                 choices=list(LLM_MODELS.keys()),
                 value="Cohere c4ai-crp-08-2024",
-                label="Select LLM Model 🤖",
-                info="Choose your preferred AI model"
+                label="LLM 모델 선택 🤖",
+                info="선호하는 AI 모델을 선택하세요"
             )
             
+            gr.Markdown("### 파일 업로드 📁\n지원: 텍스트, 코드, CSV, Parquet, PDF 파일")
             file_upload = gr.File(
-                label="Upload File 📁",
-                info="Support: Text, Code, CSV, Parquet files",
-                file_types=["text", ".csv", ".parquet"],
+                label="파일 업로드",
+                file_types=["text", ".csv", ".parquet", ".pdf"],
                 type="filepath"
             )
             
-            with gr.Accordion("Advanced Settings ⚙️", open=False):
-                system_message = gr.Textbox(label="System Message 📝", value="")
-                max_tokens = gr.Slider(minimum=1, maximum=8000, value=4000, label="Max Tokens 📊")
-                temperature = gr.Slider(minimum=0, maximum=1, value=0.7, label="Temperature 🌡️")
+            with gr.Accordion("고급 설정 ⚙️", open=False):
+                system_message = gr.Textbox(label="시스템 메시지 📝", value="")
+                max_tokens = gr.Slider(minimum=1, maximum=8000, value=4000, label="최대 토큰 📊")
+                temperature = gr.Slider(minimum=0, maximum=1, value=0.7, label="온도 🌡️")
                 top_p = gr.Slider(minimum=0, maximum=1, value=0.9, label="Top P 📈")
 
     # Event bindings
@@ -218,7 +239,7 @@ with gr.Blocks(theme="Yntec/HaleyCH_Theme_Orange", css=css, title="EveryChat
         [msg]
     )
 
-    send.click(  # 전송 버튼 이벤트 추가
+    send.click(
         chat,
         inputs=[msg, chatbot, file_upload, model_name, system_message, max_tokens, temperature, top_p],
         outputs=[msg, chatbot],
@@ -232,7 +253,7 @@ with gr.Blocks(theme="Yntec/HaleyCH_Theme_Orange", css=css, title="EveryChat
     # Auto-analysis on file upload
     file_upload.change(
         chat,
-        inputs=[gr.Textbox(value="Starting file analysis..."), chatbot, file_upload, model_name, system_message, max_tokens, temperature, top_p],
+        inputs=[gr.Textbox(value="파일 분석 시작..."), chatbot, file_upload, model_name, system_message, max_tokens, temperature, top_p],
         outputs=[msg, chatbot],
         queue=True
     )
@@ -240,12 +261,12 @@ with gr.Blocks(theme="Yntec/HaleyCH_Theme_Orange", css=css, title="EveryChat
     # Example queries
     gr.Examples(
         examples=[
-            ["Please explain the overall structure and features of the file in detail 📋"],
-            ["Analyze the main patterns and characteristics of this file 📊"],
-            ["Evaluate the file's quality and potential improvements 💡"],
-            ["How can we practically utilize this file? 🎯"],
-            ["Summarize the main content and derive key insights ✨"],
-            ["Please continue with more detailed analysis 📈"],
+            ["파일의 전체 구조와 특징을 자세히 설명해주세요 📋"],
+            ["파일의 주요 패턴과 특성을 분석해주세요 📊"],
+            ["파일의 품질과 개선점을 평가해주세요 💡"],
+            ["이 파일을 어떻게 실용적으로 활용할 수 있을까요? 🎯"],
+            ["주요 내용을 요약하고 핵심 통찰력을 도출해주세요 ✨"],
+            ["더 자세한 분석을 계속해주세요 📈"],
         ],
         inputs=msg,
     )