Polaroid / app (35).py
aiqcamp's picture
Upload 3 files
a2cc18f verified
raw
history blame
70.5 kB
import os, json, re, logging, requests, markdown, time, io
from datetime import datetime
import random
import base64
from io import BytesIO
from PIL import Image
import streamlit as st
from openai import OpenAI
from gradio_client import Client
import pandas as pd
import PyPDF2 # For handling PDF files
import kagglehub
# ──────────────────────────────── Environment Variables / Constants ─────────────────────────
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY", "")
BRAVE_KEY = os.getenv("SERPHOUSE_API_KEY", "") # Keep this name
BRAVE_ENDPOINT = "https://api.search.brave.com/res/v1/web/search"
BRAVE_VIDEO_ENDPOINT = "https://api.search.brave.com/res/v1/videos/search"
BRAVE_NEWS_ENDPOINT = "https://api.search.brave.com/res/v1/news/search"
IMAGE_API_URL = "http://211.233.58.201:7896"
MAX_TOKENS = 7999
KAGGLE_API_KEY = os.getenv("KDATA_API", "")
# Set Kaggle API key
os.environ["KAGGLE_KEY"] = KAGGLE_API_KEY
# Analysis modes and style definitions
ANALYSIS_MODES = {
"price_forecast": "농산물 가격 예츑과 μ‹œμž₯ 뢄석",
"market_trend": "μ‹œμž₯ 동ν–₯ 및 μˆ˜μš” νŒ¨ν„΄ 뢄석",
"production_analysis": "μƒμ‚°λŸ‰ 뢄석 및 μ‹λŸ‰ μ•ˆλ³΄ 전망",
"agricultural_policy": "농업 μ •μ±… 및 규제 영ν–₯ 뢄석",
"climate_impact": "κΈ°ν›„ λ³€ν™”κ°€ 농업에 λ―ΈμΉ˜λŠ” 영ν–₯ 뢄석"
}
RESPONSE_STYLES = {
"professional": "전문적이고 ν•™μˆ μ μΈ 뢄석",
"simple": "μ‰½κ²Œ 이해할 수 μžˆλŠ” κ°„κ²°ν•œ μ„€λͺ…",
"detailed": "μƒμ„Έν•œ 톡계 기반 깊이 μžˆλŠ” 뢄석",
"action_oriented": "μ‹€ν–‰ κ°€λŠ₯ν•œ μ‘°μ–Έκ³Ό μΆ”μ²œ 쀑심"
}
# Example search queries
EXAMPLE_QUERIES = {
"example1": "μŒ€ 가격 μΆ”μ„Έ 및 ν–₯ν›„ 6κ°œμ›” 전망을 λΆ„μ„ν•΄μ£Όμ„Έμš”",
"example2": "κΈ°ν›„ λ³€ν™”λ‘œ ν•œκ΅­ 과일 생산 μ „λž΅κ³Ό μˆ˜μš” 예츑 λ³΄κ³ μ„œλ₯Ό μž‘μ„±ν•˜λΌ.",
"example3": "2025λ…„λΆ€ν„° 2030λ…„κΉŒμ§€ 좩뢁 μ¦ν‰κ΅°μ—μ„œ μž¬λ°°ν•˜λ©΄ μœ λ§ν•œ μž‘λ¬Όμ€? μˆ˜μ΅μ„±κ³Ό 관리성이 μ’‹μ•„μ•Όν•œλ‹€"
}
# ──────────────────────────────── Logging ────────────────────────────────
logging.basicConfig(level=logging.INFO,
format="%(asctime)s - %(levelname)s - %(message)s")
# ──────────────────────────────── OpenAI Client ──────────────────────────
@st.cache_resource
def get_openai_client():
"""Create an OpenAI client with timeout and retry settings."""
if not OPENAI_API_KEY:
raise RuntimeError("⚠️ OPENAI_API_KEY ν™˜κ²½ λ³€μˆ˜κ°€ μ„€μ •λ˜μ§€ μ•Šμ•˜μŠ΅λ‹ˆλ‹€.")
return OpenAI(
api_key=OPENAI_API_KEY,
timeout=60.0,
max_retries=3
)
# ────────────────────────────── Kaggle Dataset Access ──────────────────────
@st.cache_resource
def load_agriculture_dataset():
"""Download and load the UN agriculture dataset from Kaggle"""
try:
path = kagglehub.dataset_download("unitednations/global-food-agriculture-statistics")
logging.info(f"Kaggle dataset downloaded to: {path}")
# Load metadata about available files
available_files = []
for root, dirs, files in os.walk(path):
for file in files:
if file.endswith('.csv'):
file_path = os.path.join(root, file)
file_size = os.path.getsize(file_path) / (1024 * 1024) # Size in MB
available_files.append({
'name': file,
'path': file_path,
'size_mb': round(file_size, 2)
})
return {
'base_path': path,
'files': available_files
}
except Exception as e:
logging.error(f"Error loading Kaggle dataset: {e}")
return None
# New function to load Advanced Soybean Agricultural Dataset
@st.cache_resource
def load_soybean_dataset():
"""Download and load the Advanced Soybean Agricultural Dataset from Kaggle"""
try:
path = kagglehub.dataset_download("wisam1985/advanced-soybean-agricultural-dataset-2025")
logging.info(f"Soybean dataset downloaded to: {path}")
available_files = []
for root, dirs, files in os.walk(path):
for file in files:
if file.endswith(('.csv', '.xlsx')):
file_path = os.path.join(root, file)
file_size = os.path.getsize(file_path) / (1024 * 1024) # Size in MB
available_files.append({
'name': file,
'path': file_path,
'size_mb': round(file_size, 2)
})
return {
'base_path': path,
'files': available_files
}
except Exception as e:
logging.error(f"Error loading Soybean dataset: {e}")
return None
# Function to load Crop Recommendation Dataset
@st.cache_resource
def load_crop_recommendation_dataset():
"""Download and load the Soil and Environmental Variables Crop Recommendation Dataset"""
try:
path = kagglehub.dataset_download("agriinnovate/agricultural-crop-dataset")
logging.info(f"Crop recommendation dataset downloaded to: {path}")
available_files = []
for root, dirs, files in os.walk(path):
for file in files:
if file.endswith(('.csv', '.xlsx')):
file_path = os.path.join(root, file)
file_size = os.path.getsize(file_path) / (1024 * 1024) # Size in MB
available_files.append({
'name': file,
'path': file_path,
'size_mb': round(file_size, 2)
})
return {
'base_path': path,
'files': available_files
}
except Exception as e:
logging.error(f"Error loading Crop recommendation dataset: {e}")
return None
# Function to load Climate Change Impact Dataset
@st.cache_resource
def load_climate_impact_dataset():
"""Download and load the Climate Change Impact on Agriculture Dataset"""
try:
path = kagglehub.dataset_download("waqi786/climate-change-impact-on-agriculture")
logging.info(f"Climate impact dataset downloaded to: {path}")
available_files = []
for root, dirs, files in os.walk(path):
for file in files:
if file.endswith(('.csv', '.xlsx')):
file_path = os.path.join(root, file)
file_size = os.path.getsize(file_path) / (1024 * 1024) # Size in MB
available_files.append({
'name': file,
'path': file_path,
'size_mb': round(file_size, 2)
})
return {
'base_path': path,
'files': available_files
}
except Exception as e:
logging.error(f"Error loading Climate impact dataset: {e}")
return None
def get_dataset_summary():
"""Generate a summary of the available agriculture datasets"""
dataset_info = load_agriculture_dataset()
if not dataset_info:
return "Failed to load the UN global food and agriculture statistics dataset."
summary = "# UN κΈ€λ‘œλ²Œ μ‹λŸ‰ 및 농업 톡계 데이터셋\n\n"
summary += f"총 {len(dataset_info['files'])}개의 CSV 파일이 ν¬ν•¨λ˜μ–΄ μžˆμŠ΅λ‹ˆλ‹€.\n\n"
# List files with sizes
summary += "## μ‚¬μš© κ°€λŠ₯ν•œ 데이터 파일:\n\n"
for i, file_info in enumerate(dataset_info['files'][:10], 1): # Limit to first 10 files
summary += f"{i}. **{file_info['name']}** ({file_info['size_mb']} MB)\n"
if len(dataset_info['files']) > 10:
summary += f"\n...μ™Έ {len(dataset_info['files']) - 10}개 파일\n"
# Add example of data structure
try:
if dataset_info['files']:
sample_file = dataset_info['files'][0]['path']
df = pd.read_csv(sample_file, nrows=5)
summary += "\n## 데이터 μƒ˜ν”Œ ꡬ쑰:\n\n"
summary += df.head(5).to_markdown() + "\n\n"
summary += "## 데이터셋 λ³€μˆ˜ μ„€λͺ…:\n\n"
for col in df.columns:
summary += f"- **{col}**: [λ³€μˆ˜ μ„€λͺ… ν•„μš”]\n"
except Exception as e:
logging.error(f"Error generating dataset sample: {e}")
summary += "\n데이터 μƒ˜ν”Œμ„ μƒμ„±ν•˜λŠ” 쀑 였λ₯˜κ°€ λ°œμƒν–ˆμŠ΅λ‹ˆλ‹€.\n"
return summary
def analyze_dataset_for_query(query):
"""Find and analyze relevant data from the dataset based on the query"""
dataset_info = load_agriculture_dataset()
if not dataset_info:
return "데이터셋을 뢈러올 수 μ—†μŠ΅λ‹ˆλ‹€. Kaggle API 연결을 ν™•μΈν•΄μ£Όμ„Έμš”."
# Extract key terms from the query
query_lower = query.lower()
# Define keywords to look for in the dataset files
keywords = {
"μŒ€": ["rice", "grain"],
"λ°€": ["wheat", "grain"],
"μ˜₯수수": ["corn", "maize", "grain"],
"μ±„μ†Œ": ["vegetable", "produce"],
"과일": ["fruit", "produce"],
"가격": ["price", "cost", "value"],
"생산": ["production", "yield", "harvest"],
"수좜": ["export", "trade"],
"μˆ˜μž…": ["import", "trade"],
"μ†ŒλΉ„": ["consumption", "demand"]
}
# Find relevant files based on the query
relevant_files = []
# First check for Korean keywords in the query
found_keywords = []
for k_term, e_terms in keywords.items():
if k_term in query_lower:
found_keywords.extend([k_term] + e_terms)
# If no Korean keywords found, check for English terms in the filenames
if not found_keywords:
# Generic search through all files
relevant_files = dataset_info['files'][:5] # Take first 5 files as default
else:
# Search for files related to the found keywords
for file_info in dataset_info['files']:
file_name_lower = file_info['name'].lower()
for keyword in found_keywords:
if keyword.lower() in file_name_lower:
relevant_files.append(file_info)
break
# If still no relevant files, take the first 5 files
if not relevant_files:
relevant_files = dataset_info['files'][:5]
# Read and analyze the relevant files
analysis_result = "# 농업 데이터 뢄석 κ²°κ³Ό\n\n"
analysis_result += f"쿼리: '{query}'에 λŒ€ν•œ 뢄석을 μˆ˜ν–‰ν–ˆμŠ΅λ‹ˆλ‹€.\n\n"
if found_keywords:
analysis_result += f"## 뢄석 ν‚€μ›Œλ“œ: {', '.join(set(found_keywords))}\n\n"
# Process each relevant file
for file_info in relevant_files[:3]: # Limit to 3 files for performance
try:
analysis_result += f"## 파일: {file_info['name']}\n\n"
# Read the CSV file
df = pd.read_csv(file_info['path'])
# Basic file stats
analysis_result += f"- ν–‰ 수: {len(df)}\n"
analysis_result += f"- μ—΄ 수: {len(df.columns)}\n"
analysis_result += f"- μ—΄ λͺ©λ‘: {', '.join(df.columns.tolist())}\n\n"
# Sample data
analysis_result += "### 데이터 μƒ˜ν”Œ:\n\n"
analysis_result += df.head(5).to_markdown() + "\n\n"
# Statistical summary of numeric columns
numeric_cols = df.select_dtypes(include=['number']).columns
if len(numeric_cols) > 0:
analysis_result += "### κΈ°λ³Έ 톡계:\n\n"
stats_df = df[numeric_cols].describe()
analysis_result += stats_df.to_markdown() + "\n\n"
# Time series analysis if possible
time_cols = [col for col in df.columns if 'year' in col.lower() or 'date' in col.lower()]
if time_cols:
analysis_result += "### μ‹œκ³„μ—΄ νŒ¨ν„΄:\n\n"
analysis_result += "데이터셋에 μ‹œκ°„ κ΄€λ ¨ 열이 μžˆμ–΄ μ‹œκ³„μ—΄ 뢄석이 κ°€λŠ₯ν•©λ‹ˆλ‹€.\n\n"
except Exception as e:
logging.error(f"Error analyzing file {file_info['name']}: {e}")
analysis_result += f"이 파일 뢄석 쀑 였λ₯˜κ°€ λ°œμƒν–ˆμŠ΅λ‹ˆλ‹€: {str(e)}\n\n"
analysis_result += "## 농산물 가격 예츑 및 μˆ˜μš” 뢄석에 λŒ€ν•œ μΈμ‚¬μ΄νŠΈ\n\n"
analysis_result += "λ°μ΄ν„°μ…‹μ—μ„œ μΆ”μΆœν•œ 정보λ₯Ό λ°”νƒ•μœΌλ‘œ λ‹€μŒ μΈμ‚¬μ΄νŠΈλ₯Ό μ œκ³΅ν•©λ‹ˆλ‹€:\n\n"
analysis_result += "1. 데이터 기반 뢄석 (기본적인 μš”μ•½)\n"
analysis_result += "2. μ£Όμš” 가격 및 μˆ˜μš” 동ν–₯\n"
analysis_result += "3. μƒμ‚°λŸ‰ 및 무역 νŒ¨ν„΄\n\n"
analysis_result += "이 뢄석은 UN κΈ€λ‘œλ²Œ μ‹λŸ‰ 및 농업 톡계 데이터셋을 기반으둜 ν•©λ‹ˆλ‹€.\n\n"
return analysis_result
# Function to analyze crop recommendation dataset
def analyze_crop_recommendation_dataset(query):
"""Find and analyze crop recommendation data based on the query"""
try:
dataset_info = load_crop_recommendation_dataset()
if not dataset_info or not dataset_info['files']:
return "μž‘λ¬Ό μΆ”μ²œ 데이터셋을 뢈러올 수 μ—†μŠ΅λ‹ˆλ‹€."
analysis_result = "# ν† μ–‘ 및 ν™˜κ²½ λ³€μˆ˜ 기반 μž‘λ¬Ό μΆ”μ²œ 데이터 뢄석\n\n"
# Process main files
for file_info in dataset_info['files'][:2]: # Limit to the first 2 files
try:
analysis_result += f"## 파일: {file_info['name']}\n\n"
if file_info['name'].endswith('.csv'):
df = pd.read_csv(file_info['path'])
elif file_info['name'].endswith('.xlsx'):
df = pd.read_excel(file_info['path'])
else:
continue
# Basic dataset info
analysis_result += f"- 데이터 크기: {len(df)} ν–‰ Γ— {len(df.columns)} μ—΄\n"
analysis_result += f"- ν¬ν•¨λœ μž‘λ¬Ό μ’…λ₯˜: "
# Check if crop column exists
crop_cols = [col for col in df.columns if 'crop' in col.lower() or 'μž‘λ¬Ό' in col.lower()]
if crop_cols:
main_crop_col = crop_cols[0]
unique_crops = df[main_crop_col].unique()
analysis_result += f"{len(unique_crops)}μ’… ({', '.join(str(c) for c in unique_crops[:10])})\n\n"
else:
analysis_result += "μž‘λ¬Ό 정보 열을 찾을 수 μ—†μŒ\n\n"
# Extract environmental factors
env_factors = [col for col in df.columns if col.lower() not in ['crop', 'label', 'id', 'index']]
if env_factors:
analysis_result += f"- 고렀된 ν™˜κ²½ μš”μ†Œ: {', '.join(env_factors)}\n\n"
# Sample data
analysis_result += "### 데이터 μƒ˜ν”Œ:\n\n"
analysis_result += df.head(5).to_markdown() + "\n\n"
# Summary statistics for environmental factors
if env_factors:
numeric_factors = df[env_factors].select_dtypes(include=['number']).columns
if len(numeric_factors) > 0:
analysis_result += "### ν™˜κ²½ μš”μ†Œ 톡계:\n\n"
stats_df = df[numeric_factors].describe().round(2)
analysis_result += stats_df.to_markdown() + "\n\n"
# Check for query-specific crops
query_terms = query.lower().split()
relevant_crops = []
if crop_cols:
for crop in df[main_crop_col].unique():
crop_str = str(crop).lower()
if any(term in crop_str for term in query_terms):
relevant_crops.append(crop)
if relevant_crops:
analysis_result += f"### 쿼리 κ΄€λ ¨ μž‘λ¬Ό 뢄석: {', '.join(str(c) for c in relevant_crops)}\n\n"
for crop in relevant_crops[:3]: # Limit to 3 crops
crop_data = df[df[main_crop_col] == crop]
analysis_result += f"#### {crop} μž‘λ¬Ό μš”μ•½:\n\n"
analysis_result += f"- μƒ˜ν”Œ 수: {len(crop_data)}개\n"
if len(numeric_factors) > 0:
crop_stats = crop_data[numeric_factors].describe().round(2)
analysis_result += f"- 평균 ν™˜κ²½ 쑰건:\n"
for factor in numeric_factors[:5]: # Limit to 5 factors
analysis_result += f" * {factor}: {crop_stats.loc['mean', factor]}\n"
analysis_result += "\n"
except Exception as e:
logging.error(f"Error analyzing crop recommendation file {file_info['name']}: {e}")
analysis_result += f"뢄석 였λ₯˜: {str(e)}\n\n"
analysis_result += "## μž‘λ¬Ό μΆ”μ²œ μΈμ‚¬μ΄νŠΈ\n\n"
analysis_result += "ν† μ–‘ 및 ν™˜κ²½ λ³€μˆ˜ 데이터셋 뢄석 κ²°κ³Ό, λ‹€μŒκ³Ό 같은 μ£Όμš” μΈμ‚¬μ΄νŠΈλ₯Ό μ œκ³΅ν•©λ‹ˆλ‹€:\n\n"
analysis_result += "1. μ§€μ—­ ν™˜κ²½μ— μ ν•©ν•œ μž‘λ¬Ό μΆ”μ²œ\n"
analysis_result += "2. μž‘λ¬Ό 생산성에 영ν–₯을 λ―ΈμΉ˜λŠ” μ£Όμš” ν™˜κ²½ μš”μΈ\n"
analysis_result += "3. 지속 κ°€λŠ₯ν•œ 농업을 μœ„ν•œ 졜적의 μž‘λ¬Ό 선택 κΈ°μ€€\n\n"
return analysis_result
except Exception as e:
logging.error(f"Crop recommendation dataset analysis error: {e}")
return "μž‘λ¬Ό μΆ”μ²œ 데이터셋 뢄석 쀑 였λ₯˜κ°€ λ°œμƒν–ˆμŠ΅λ‹ˆλ‹€."
# Function to analyze climate impact dataset
def analyze_climate_impact_dataset(query):
"""Find and analyze climate impact on agriculture data based on the query"""
try:
dataset_info = load_climate_impact_dataset()
if not dataset_info or not dataset_info['files']:
return "κΈ°ν›„ λ³€ν™” 영ν–₯ 데이터셋을 뢈러올 수 μ—†μŠ΅λ‹ˆλ‹€."
analysis_result = "# κΈ°ν›„ λ³€ν™”κ°€ 농업에 λ―ΈμΉ˜λŠ” 영ν–₯ 데이터 뢄석\n\n"
# Process main files
for file_info in dataset_info['files'][:2]: # Limit to first 2 files
try:
analysis_result += f"## 파일: {file_info['name']}\n\n"
if file_info['name'].endswith('.csv'):
df = pd.read_csv(file_info['path'])
elif file_info['name'].endswith('.xlsx'):
df = pd.read_excel(file_info['path'])
else:
continue
# Basic dataset info
analysis_result += f"- 데이터 크기: {len(df)} ν–‰ Γ— {len(df.columns)} μ—΄\n"
# Check for region column
region_cols = [col for col in df.columns if 'region' in col.lower() or 'country' in col.lower() or 'μ§€μ—­' in col.lower()]
if region_cols:
main_region_col = region_cols[0]
regions = df[main_region_col].unique()
analysis_result += f"- ν¬ν•¨λœ μ§€μ—­: {len(regions)}개 ({', '.join(str(r) for r in regions[:5])})\n"
# Identify climate and crop related columns
climate_cols = [col for col in df.columns if any(term in col.lower() for term in
['temp', 'rainfall', 'precipitation', 'climate', 'weather', '기온', 'κ°•μˆ˜λŸ‰'])]
crop_cols = [col for col in df.columns if any(term in col.lower() for term in
['yield', 'production', 'crop', 'harvest', 'μˆ˜ν™•λŸ‰', 'μƒμ‚°λŸ‰'])]
if climate_cols:
analysis_result += f"- κΈ°ν›„ κ΄€λ ¨ λ³€μˆ˜: {', '.join(climate_cols)}\n"
if crop_cols:
analysis_result += f"- μž‘λ¬Ό κ΄€λ ¨ λ³€μˆ˜: {', '.join(crop_cols)}\n\n"
# Sample data
analysis_result += "### 데이터 μƒ˜ν”Œ:\n\n"
analysis_result += df.head(5).to_markdown() + "\n\n"
# Time series pattern if available
year_cols = [col for col in df.columns if 'year' in col.lower() or 'date' in col.lower() or '연도' in col.lower()]
if year_cols:
analysis_result += "### μ‹œκ³„μ—΄ κΈ°ν›„ 영ν–₯ νŒ¨ν„΄:\n\n"
analysis_result += "이 데이터셋은 μ‹œκ°„μ— λ”°λ₯Έ κΈ°ν›„ 변화와 농업 생산성 κ°„μ˜ 관계λ₯Ό 뢄석할 수 μžˆμŠ΅λ‹ˆλ‹€.\n\n"
# Statistical summary of key variables
key_vars = climate_cols + crop_cols
numeric_vars = df[key_vars].select_dtypes(include=['number']).columns
if len(numeric_vars) > 0:
analysis_result += "### μ£Όμš” λ³€μˆ˜ 톡계:\n\n"
stats_df = df[numeric_vars].describe().round(2)
analysis_result += stats_df.to_markdown() + "\n\n"
# Check for correlations between climate and crop variables
if len(climate_cols) > 0 and len(crop_cols) > 0:
numeric_climate = df[climate_cols].select_dtypes(include=['number']).columns
numeric_crop = df[crop_cols].select_dtypes(include=['number']).columns
if len(numeric_climate) > 0 and len(numeric_crop) > 0:
analysis_result += "### 기후와 μž‘λ¬Ό 생산 κ°„μ˜ 상관관계:\n\n"
try:
corr_vars = list(numeric_climate)[:2] + list(numeric_crop)[:2] # Limit to 2 of each type
corr_df = df[corr_vars].corr().round(3)
analysis_result += corr_df.to_markdown() + "\n\n"
analysis_result += "μœ„ 상관관계 ν‘œλŠ” κΈ°ν›„ λ³€μˆ˜μ™€ μž‘λ¬Ό 생산성 κ°„μ˜ 관계 강도λ₯Ό λ³΄μ—¬μ€λ‹ˆλ‹€.\n\n"
except:
analysis_result += "상관관계 계산 쀑 였λ₯˜κ°€ λ°œμƒν–ˆμŠ΅λ‹ˆλ‹€.\n\n"
except Exception as e:
logging.error(f"Error analyzing climate impact file {file_info['name']}: {e}")
analysis_result += f"뢄석 였λ₯˜: {str(e)}\n\n"
analysis_result += "## κΈ°ν›„ λ³€ν™” 영ν–₯ μΈμ‚¬μ΄νŠΈ\n\n"
analysis_result += "κΈ°ν›„ λ³€ν™”κ°€ 농업에 λ―ΈμΉ˜λŠ” 영ν–₯ 데이터 뢄석 κ²°κ³Ό, λ‹€μŒκ³Ό 같은 μΈμ‚¬μ΄νŠΈλ₯Ό μ œκ³΅ν•©λ‹ˆλ‹€:\n\n"
analysis_result += "1. 기온 변화에 λ”°λ₯Έ μž‘λ¬Ό 생산성 변동 νŒ¨ν„΄\n"
analysis_result += "2. κ°•μˆ˜λŸ‰ λ³€ν™”κ°€ 농업 μˆ˜ν™•λŸ‰μ— λ―ΈμΉ˜λŠ” 영ν–₯\n"
analysis_result += "3. κΈ°ν›„ 변화에 λŒ€μ‘ν•˜κΈ° μœ„ν•œ 농업 μ „λž΅ μ œμ•ˆ\n"
analysis_result += "4. 지역별 κΈ°ν›„ μ·¨μ•½μ„± 및 적응 λ°©μ•ˆ\n\n"
return analysis_result
except Exception as e:
logging.error(f"Climate impact dataset analysis error: {e}")
return "κΈ°ν›„ λ³€ν™” 영ν–₯ 데이터셋 뢄석 쀑 였λ₯˜κ°€ λ°œμƒν–ˆμŠ΅λ‹ˆλ‹€."
# Function to analyze soybean dataset if selected
def analyze_soybean_dataset(query):
"""Find and analyze soybean agriculture data based on the query"""
try:
dataset_info = load_soybean_dataset()
if not dataset_info or not dataset_info['files']:
return "λŒ€λ‘ 농업 데이터셋을 뢈러올 수 μ—†μŠ΅λ‹ˆλ‹€."
analysis_result = "# κ³ κΈ‰ λŒ€λ‘ 농업 데이터 뢄석\n\n"
# Process main files
for file_info in dataset_info['files'][:2]: # Limit to the first 2 files
try:
analysis_result += f"## 파일: {file_info['name']}\n\n"
if file_info['name'].endswith('.csv'):
df = pd.read_csv(file_info['path'])
elif file_info['name'].endswith('.xlsx'):
df = pd.read_excel(file_info['path'])
else:
continue
# Basic file stats
analysis_result += f"- 데이터 크기: {len(df)} ν–‰ Γ— {len(df.columns)} μ—΄\n"
# Check for region/location columns
location_cols = [col for col in df.columns if any(term in col.lower() for term in
['region', 'location', 'area', 'country', 'μ§€μ—­'])]
if location_cols:
main_loc_col = location_cols[0]
locations = df[main_loc_col].unique()
analysis_result += f"- ν¬ν•¨λœ μ§€μ—­: {len(locations)}개 ({', '.join(str(loc) for loc in locations[:5])})\n"
# Identify yield and production columns
yield_cols = [col for col in df.columns if any(term in col.lower() for term in
['yield', 'production', 'harvest', 'μˆ˜ν™•λŸ‰', 'μƒμ‚°λŸ‰'])]
if yield_cols:
analysis_result += f"- 생산성 κ΄€λ ¨ λ³€μˆ˜: {', '.join(yield_cols)}\n"
# Identify environmental factors
env_cols = [col for col in df.columns if any(term in col.lower() for term in
['temp', 'rainfall', 'soil', 'fertilizer', 'nutrient', 'irrigation',
'기온', 'κ°•μˆ˜λŸ‰', 'ν† μ–‘', 'λΉ„λ£Œ', 'κ΄€κ°œ'])]
if env_cols:
analysis_result += f"- ν™˜κ²½ κ΄€λ ¨ λ³€μˆ˜: {', '.join(env_cols)}\n\n"
# Sample data
analysis_result += "### 데이터 μƒ˜ν”Œ:\n\n"
analysis_result += df.head(5).to_markdown() + "\n\n"
# Statistical summary of key variables
key_vars = yield_cols + env_cols
numeric_vars = df[key_vars].select_dtypes(include=['number']).columns
if len(numeric_vars) > 0:
analysis_result += "### μ£Όμš” λ³€μˆ˜ 톡계:\n\n"
stats_df = df[numeric_vars].describe().round(2)
analysis_result += stats_df.to_markdown() + "\n\n"
# Time series analysis if possible
year_cols = [col for col in df.columns if 'year' in col.lower() or 'date' in col.lower() or '연도' in col.lower()]
if year_cols:
analysis_result += "### μ‹œκ³„μ—΄ 생산성 νŒ¨ν„΄:\n\n"
analysis_result += "이 데이터셋은 μ‹œκ°„μ— λ”°λ₯Έ λŒ€λ‘ μƒμ‚°μ„±μ˜ λ³€ν™”λ₯Ό 좔적할 수 μžˆμŠ΅λ‹ˆλ‹€.\n\n"
# Check for correlations between environmental factors and yield
if len(env_cols) > 0 and len(yield_cols) > 0:
numeric_env = df[env_cols].select_dtypes(include=['number']).columns
numeric_yield = df[yield_cols].select_dtypes(include=['number']).columns
if len(numeric_env) > 0 and len(numeric_yield) > 0:
analysis_result += "### ν™˜κ²½ μš”μ†Œμ™€ λŒ€λ‘ 생산성 κ°„μ˜ 상관관계:\n\n"
try:
corr_vars = list(numeric_env)[:3] + list(numeric_yield)[:2] # Limit variables
corr_df = df[corr_vars].corr().round(3)
analysis_result += corr_df.to_markdown() + "\n\n"
except:
analysis_result += "상관관계 계산 쀑 였λ₯˜κ°€ λ°œμƒν–ˆμŠ΅λ‹ˆλ‹€.\n\n"
except Exception as e:
logging.error(f"Error analyzing soybean file {file_info['name']}: {e}")
analysis_result += f"뢄석 였λ₯˜: {str(e)}\n\n"
analysis_result += "## λŒ€λ‘ 농업 μΈμ‚¬μ΄νŠΈ\n\n"
analysis_result += "κ³ κΈ‰ λŒ€λ‘ 농업 데이터셋 뢄석 κ²°κ³Ό, λ‹€μŒκ³Ό 같은 μΈμ‚¬μ΄νŠΈλ₯Ό μ œκ³΅ν•©λ‹ˆλ‹€:\n\n"
analysis_result += "1. 졜적의 λŒ€λ‘ 생산을 μœ„ν•œ ν™˜κ²½ 쑰건\n"
analysis_result += "2. 지역별 λŒ€λ‘ 생산성 λ³€ν™” νŒ¨ν„΄\n"
analysis_result += "3. 생산성 ν–₯상을 μœ„ν•œ 농업 기술 및 접근법\n"
analysis_result += "4. μ‹œμž₯ μˆ˜μš”μ— λ§žλŠ” λŒ€λ‘ ν’ˆμ’… 선택 κ°€μ΄λ“œ\n\n"
return analysis_result
except Exception as e:
logging.error(f"Soybean dataset analysis error: {e}")
return "λŒ€λ‘ 농업 데이터셋 뢄석 쀑 였λ₯˜κ°€ λ°œμƒν–ˆμŠ΅λ‹ˆλ‹€."
# ──────────────────────────────── System Prompt ─────────────────────────
def get_system_prompt(mode="price_forecast", style="professional", include_search_results=True, include_uploaded_files=False) -> str:
"""
Generate a system prompt for the 'Agricultural Price & Demand Forecast AI Assistant' interface based on:
- The selected analysis mode and style
- Guidelines for using agricultural datasets, web search results and uploaded files
"""
base_prompt = """
당신은 농업 데이터 μ „λ¬Έκ°€λ‘œμ„œ 농산물 가격 예츑과 μˆ˜μš” 뢄석을 μˆ˜ν–‰ν•˜λŠ” AI μ–΄μ‹œμŠ€ν„΄νŠΈμž…λ‹ˆλ‹€.
μ£Όμš” μž„λ¬΄:
1. UN κΈ€λ‘œλ²Œ μ‹λŸ‰ 및 농업 톡계 데이터셋을 기반으둜 농산물 μ‹œμž₯ 뢄석
2. 농산물 가격 μΆ”μ„Έ 예츑 및 μˆ˜μš” νŒ¨ν„΄ 뢄석
3. 데이터λ₯Ό λ°”νƒ•μœΌλ‘œ λͺ…ν™•ν•˜κ³  κ·Όκ±° μžˆλŠ” 뢄석 제곡
4. κ΄€λ ¨ 정보와 μΈμ‚¬μ΄νŠΈλ₯Ό μ²΄κ³„μ μœΌλ‘œ κ΅¬μ„±ν•˜μ—¬ μ œμ‹œ
5. μ‹œκ°μ  이해λ₯Ό 돕기 μœ„ν•΄ 차트, κ·Έλž˜ν”„ 등을 적절히 ν™œμš©
6. ν† μ–‘ 및 ν™˜κ²½ λ³€μˆ˜ 기반 μž‘λ¬Ό μΆ”μ²œ λ°μ΄ν„°μ…‹μ—μ„œ μΆ”μΆœν•œ μΈμ‚¬μ΄νŠΈ 적용
7. κΈ°ν›„ λ³€ν™”κ°€ 농업에 λ―ΈμΉ˜λŠ” 영ν–₯ 데이터셋을 ν†΅ν•œ ν™˜κ²½ λ³€ν™” μ‹œλ‚˜λ¦¬μ˜€ 뢄석
μ€‘μš” κ°€μ΄λ“œλΌμΈ:
- 데이터에 κΈ°λ°˜ν•œ 객관적 뢄석을 μ œκ³΅ν•˜μ„Έμš”
- 뢄석 κ³Όμ •κ³Ό 방법둠을 λͺ…ν™•νžˆ μ„€λͺ…ν•˜μ„Έμš”
- 톡계적 μ‹ λ’°μ„±κ³Ό ν•œκ³„μ μ„ 투λͺ…ν•˜κ²Œ μ œμ‹œν•˜μ„Έμš”
- μ΄ν•΄ν•˜κΈ° μ‰¬μš΄ μ‹œκ°μ  μš”μ†Œλ‘œ 뢄석 κ²°κ³Όλ₯Ό λ³΄μ™„ν•˜μ„Έμš”
- λ§ˆν¬λ‹€μš΄μ„ ν™œμš©ν•΄ 응닡을 μ²΄κ³„μ μœΌλ‘œ κ΅¬μ„±ν•˜μ„Έμš”
"""
mode_prompts = {
"price_forecast": """
농산물 가격 예츑 및 μ‹œμž₯ 뢄석에 μ§‘μ€‘ν•©λ‹ˆλ‹€:
- κ³Όκ±° 가격 데이터 νŒ¨ν„΄μ— κΈ°λ°˜ν•œ 예츑 제곡
- 가격 변동성 μš”μΈ 뢄석(κ³„μ ˆμ„±, 날씨, μ •μ±… λ“±)
- 단기 및 쀑μž₯κΈ° 가격 전망 μ œμ‹œ
- 가격에 영ν–₯을 λ―ΈμΉ˜λŠ” κ΅­λ‚΄μ™Έ μš”μΈ 식별
- μ‹œμž₯ λΆˆν™•μ‹€μ„±κ³Ό 리슀크 μš”μ†Œ κ°•μ‘°
""",
"market_trend": """
μ‹œμž₯ 동ν–₯ 및 μˆ˜μš” νŒ¨ν„΄ 뢄석에 μ§‘μ€‘ν•©λ‹ˆλ‹€:
- μ£Όμš” 농산물 μˆ˜μš” λ³€ν™” νŒ¨ν„΄ 식별
- μ†ŒλΉ„μž μ„ ν˜Έλ„ 및 ꡬ맀 행동 뢄석
- μ‹œμž₯ μ„Έκ·Έλ¨ΌνŠΈ 및 ν‹ˆμƒˆμ‹œμž₯ 기회 탐색
- μ‹œμž₯ ν™•λŒ€/μΆ•μ†Œ νŠΈλ Œλ“œ 평가
- μˆ˜μš” 탄λ ₯μ„± 및 가격 민감도 뢄석
""",
"production_analysis": """
μƒμ‚°λŸ‰ 뢄석 및 μ‹λŸ‰ μ•ˆλ³΄ 전망에 μ§‘μ€‘ν•©λ‹ˆλ‹€:
- μž‘λ¬Ό μƒμ‚°λŸ‰ μΆ”μ„Έ 및 변동 μš”μΈ 뢄석
- μ‹λŸ‰ 생산과 인ꡬ μ„±μž₯ κ°„μ˜ 관계 평가
- κ΅­κ°€/지역별 생산 μ—­λŸ‰ 비ꡐ
- μ‹λŸ‰ μ•ˆλ³΄ μœ„ν˜‘ μš”μ†Œ 및 취약점 식별
- 생산성 ν–₯상 μ „λž΅ 및 기회 μ œμ•ˆ
""",
"agricultural_policy": """
농업 μ •μ±… 및 규제 영ν–₯ 뢄석에 μ§‘μ€‘ν•©λ‹ˆλ‹€:
- μ •λΆ€ μ •μ±…κ³Ό, 보쑰금, 규제의 μ‹œμž₯ 영ν–₯ 뢄석
- ꡭ제 무역 μ •μ±…κ³Ό κ΄€μ„Έμ˜ 농산물 가격 영ν–₯ 평가
- 농업 지원 ν”„λ‘œκ·Έλž¨μ˜ νš¨κ³Όμ„± κ²€ν† 
- 규제 ν™˜κ²½ 변화에 λ”°λ₯Έ μ‹œμž₯ μ‘°μ • 예츑
- 정책적 κ°œμž…μ˜ μ˜λ„λœ/μ˜λ„μΉ˜ μ•Šμ€ κ²°κ³Ό 뢄석
""",
"climate_impact": """
κΈ°ν›„ λ³€ν™”κ°€ 농업에 λ―ΈμΉ˜λŠ” 영ν–₯ 뢄석에 μ§‘μ€‘ν•©λ‹ˆλ‹€:
- κΈ°ν›„ 변화와 농산물 μƒμ‚°λŸ‰/ν’ˆμ§ˆ κ°„μ˜ 상관관계 뢄석
- 기상 이변이 가격 변동성에 λ―ΈμΉ˜λŠ” 영ν–₯ 평가
- μž₯기적 κΈ°ν›„ 좔세에 λ”°λ₯Έ 농업 νŒ¨ν„΄ λ³€ν™” 예츑
- κΈ°ν›„ 회볡λ ₯ μžˆλŠ” 농업 μ‹œμŠ€ν…œ μ „λž΅ μ œμ•ˆ
- 지역별 κΈ°ν›„ μœ„ν—˜ λ…ΈμΆœλ„ 및 μ·¨μ•½μ„± λ§€ν•‘
"""
}
style_guides = {
"professional": "전문적이고 ν•™μˆ μ μΈ μ–΄μ‘°λ₯Ό μ‚¬μš©ν•˜μ„Έμš”. 기술적 μš©μ–΄λ₯Ό 적절히 μ‚¬μš©ν•˜κ³  체계적인 데이터 뢄석을 μ œκ³΅ν•˜μ„Έμš”.",
"simple": "쉽고 κ°„κ²°ν•œ μ–Έμ–΄λ‘œ μ„€λͺ…ν•˜μ„Έμš”. μ „λ¬Έ μš©μ–΄λŠ” μ΅œμ†Œν™”ν•˜κ³  핡심 κ°œλ…μ„ 일상적인 ν‘œν˜„μœΌλ‘œ μ „λ‹¬ν•˜μ„Έμš”.",
"detailed": "μƒμ„Έν•˜κ³  포괄적인 뢄석을 μ œκ³΅ν•˜μ„Έμš”. λ‹€μ–‘ν•œ 데이터 포인트, 톡계적 λ‰˜μ•™μŠ€, 그리고 μ—¬λŸ¬ μ‹œλ‚˜λ¦¬μ˜€λ₯Ό κ³ λ €ν•œ 심측 뢄석을 μ œμ‹œν•˜μ„Έμš”.",
"action_oriented": "μ‹€ν–‰ κ°€λŠ₯ν•œ μΈμ‚¬μ΄νŠΈμ™€ ꡬ체적인 ꢌμž₯사항에 μ΄ˆμ μ„ λ§žμΆ”μ„Έμš”. 'λ‹€μŒ 단계' 및 'μ‹€μ§ˆμ  μ‘°μ–Έ' μ„Ήμ…˜μ„ ν¬ν•¨ν•˜μ„Έμš”."
}
dataset_guide = """
농업 데이터셋 ν™œμš© μ§€μΉ¨:
- UN κΈ€λ‘œλ²Œ μ‹λŸ‰ 및 농업 톡계 데이터셋을 κΈ°λ³Έ λΆ„μ„μ˜ 근거둜 μ‚¬μš©ν•˜μ„Έμš”
- ν† μ–‘ 및 ν™˜κ²½ λ³€μˆ˜ 기반 μž‘λ¬Ό μΆ”μ²œ λ°μ΄ν„°μ…‹μ˜ μΈμ‚¬μ΄νŠΈλ₯Ό μž‘λ¬Ό 선택 및 재배 쑰건 뢄석에 ν†΅ν•©ν•˜μ„Έμš”
- κΈ°ν›„ λ³€ν™”κ°€ 농업에 λ―ΈμΉ˜λŠ” 영ν–₯ λ°μ΄ν„°μ…‹μ˜ 정보λ₯Ό 지속 κ°€λŠ₯μ„± 및 미래 전망 뢄석에 ν™œμš©ν•˜μ„Έμš”
- λ°μ΄ν„°μ˜ μΆœμ²˜μ™€ 연도λ₯Ό λͺ…ν™•νžˆ μΈμš©ν•˜μ„Έμš”
- 데이터셋 λ‚΄ μ£Όμš” λ³€μˆ˜ κ°„μ˜ 관계λ₯Ό λΆ„μ„ν•˜μ—¬ μΈμ‚¬μ΄νŠΈλ₯Ό λ„μΆœν•˜μ„Έμš”
- λ°μ΄ν„°μ˜ ν•œκ³„μ™€ λΆˆν™•μ‹€μ„±μ„ 투λͺ…ν•˜κ²Œ μ–ΈκΈ‰ν•˜μ„Έμš”
- ν•„μš”μ‹œ 데이터 격차λ₯Ό μ‹λ³„ν•˜κ³  μΆ”κ°€ 연ꡬ가 ν•„μš”ν•œ μ˜μ—­μ„ μ œμ•ˆν•˜μ„Έμš”
"""
soybean_guide = """
κ³ κΈ‰ λŒ€λ‘ 농업 데이터셋 ν™œμš© μ§€μΉ¨:
- λŒ€λ‘ 생산 쑰건 및 μˆ˜ν™•λŸ‰ νŒ¨ν„΄μ„ λ‹€λ₯Έ μž‘λ¬Όκ³Ό λΉ„κ΅ν•˜μ—¬ λΆ„μ„ν•˜μ„Έμš”
- λŒ€λ‘ λ†μ—…μ˜ 경제적 κ°€μΉ˜μ™€ μ‹œμž₯ κΈ°νšŒμ— λŒ€ν•œ μΈμ‚¬μ΄νŠΈλ₯Ό μ œκ³΅ν•˜μ„Έμš”
- λŒ€λ‘ 생산성에 영ν–₯을 λ―ΈμΉ˜λŠ” μ£Όμš” ν™˜κ²½ μš”μΈμ„ κ°•μ‘°ν•˜μ„Έμš”
- λŒ€λ‘ 재배 기술 ν˜μ‹ κ³Ό μˆ˜μ΅μ„± ν–₯상 λ°©μ•ˆμ„ μ œμ•ˆν•˜μ„Έμš”
- 지속 κ°€λŠ₯ν•œ λŒ€λ‘ 농업을 μœ„ν•œ μ‹€μ§ˆμ μΈ 접근법을 κ³΅μœ ν•˜μ„Έμš”
"""
crop_recommendation_guide = """
ν† μ–‘ 및 ν™˜κ²½ λ³€μˆ˜ 기반 μž‘λ¬Ό μΆ”μ²œ ν™œμš© μ§€μΉ¨:
- μ§€μ—­ νŠΉμ„±μ— λ§žλŠ” 졜적의 μž‘λ¬Ό 선택 기쀀을 μ œμ‹œν•˜μ„Έμš”
- ν† μ–‘ 쑰건과 μž‘λ¬Ό 적합성 κ°„μ˜ 상관관계λ₯Ό λΆ„μ„ν•˜μ„Έμš”
- ν™˜κ²½ λ³€μˆ˜μ— λ”°λ₯Έ μž‘λ¬Ό 생산성 예츑 λͺ¨λΈμ„ ν™œμš©ν•˜μ„Έμš”
- 농업 생산성과 μˆ˜μ΅μ„± ν–₯상을 μœ„ν•œ μž‘λ¬Ό 선택 μ „λž΅μ„ μ œμ•ˆν•˜μ„Έμš”
- 지속 κ°€λŠ₯ν•œ 농업을 μœ„ν•œ μž‘λ¬Ό λ‹€μ–‘ν™” 접근법을 ꢌμž₯ν•˜μ„Έμš”
"""
climate_impact_guide = """
κΈ°ν›„ λ³€ν™”κ°€ 농업에 λ―ΈμΉ˜λŠ” 영ν–₯ 데이터셋 ν™œμš© μ§€μΉ¨:
- κΈ°ν›„ λ³€ν™” μ‹œλ‚˜λ¦¬μ˜€μ— λ”°λ₯Έ μž‘λ¬Ό 생산성 λ³€ν™”λ₯Ό μ˜ˆμΈ‘ν•˜μ„Έμš”
- κΈ°ν›„ μ μ‘ν˜• 농업 기술 및 μ „λž΅μ„ μ œμ•ˆν•˜μ„Έμš”
- 지역별 κΈ°ν›„ μœ„ν—˜ μš”μ†Œμ™€ λŒ€μ‘ λ°©μ•ˆμ„ λΆ„μ„ν•˜μ„Έμš”
- κΈ°ν›„ 변화에 λŒ€μ‘ν•˜κΈ° μœ„ν•œ μž‘λ¬Ό 선택 및 재배 μ‹œκΈ° μ‘°μ • λ°©μ•ˆμ„ μ œμ‹œν•˜μ„Έμš”
- κΈ°ν›„ λ³€ν™”κ°€ 농산물 가격 및 μ‹œμž₯ 동ν–₯에 λ―ΈμΉ˜λŠ” 영ν–₯을 ν‰κ°€ν•˜μ„Έμš”
"""
search_guide = """
μ›Ή 검색 κ²°κ³Ό ν™œμš© μ§€μΉ¨:
- 데이터셋 뢄석을 λ³΄μ™„ν•˜λŠ” μ΅œμ‹  μ‹œμž₯ μ •λ³΄λ‘œ 검색 κ²°κ³Όλ₯Ό ν™œμš©ν•˜μ„Έμš”
- 각 μ •λ³΄μ˜ 좜처λ₯Ό λ§ˆν¬λ‹€μš΄ 링크둜 ν¬ν•¨ν•˜μ„Έμš”: [좜처λͺ…](URL)
- μ£Όμš” μ£Όμž₯μ΄λ‚˜ 데이터 ν¬μΈνŠΈλ§ˆλ‹€ 좜처λ₯Ό ν‘œμ‹œν•˜μ„Έμš”
- μΆœμ²˜κ°€ 상좩할 경우, λ‹€μ–‘ν•œ 관점과 신뒰도λ₯Ό μ„€λͺ…ν•˜μ„Έμš”
- κ΄€λ ¨ λ™μ˜μƒ λ§ν¬λŠ” [λΉ„λ””μ˜€: 제λͺ©](video_url) ν˜•μ‹μœΌλ‘œ ν¬ν•¨ν•˜μ„Έμš”
- 검색 정보λ₯Ό μΌκ΄€λ˜κ³  체계적인 μ‘λ‹΅μœΌλ‘œ ν†΅ν•©ν•˜μ„Έμš”
- λͺ¨λ“  μ£Όμš” 좜처λ₯Ό λ‚˜μ—΄ν•œ "μ°Έκ³  자료" μ„Ήμ…˜μ„ λ§ˆμ§€λ§‰μ— ν¬ν•¨ν•˜μ„Έμš”
"""
upload_guide = """
μ—…λ‘œλ“œλœ 파일 ν™œμš© μ§€μΉ¨:
- μ—…λ‘œλ“œλœ νŒŒμΌμ„ μ‘λ‹΅μ˜ μ£Όμš” μ •λ³΄μ›μœΌλ‘œ ν™œμš©ν•˜μ„Έμš”
- 쿼리와 직접 κ΄€λ ¨λœ 파일 정보λ₯Ό μΆ”μΆœν•˜κ³  κ°•μ‘°ν•˜μ„Έμš”
- κ΄€λ ¨ κ΅¬μ ˆμ„ μΈμš©ν•˜κ³  νŠΉμ • νŒŒμΌμ„ 좜처둜 μΈμš©ν•˜μ„Έμš”
- CSV 파일의 수치 λ°μ΄ν„°λŠ” μš”μ•½ λ¬Έμž₯으둜 λ³€ν™˜ν•˜μ„Έμš”
- PDF μ½˜ν…μΈ λŠ” νŠΉμ • μ„Ήμ…˜μ΄λ‚˜ νŽ˜μ΄μ§€λ₯Ό μ°Έμ‘°ν•˜μ„Έμš”
- 파일 정보λ₯Ό μ›Ή 검색 결과와 μ›ν™œν•˜κ²Œ ν†΅ν•©ν•˜μ„Έμš”
- 정보가 상좩할 경우, 일반적인 μ›Ή 결과보닀 파일 μ½˜ν…μΈ λ₯Ό μš°μ„ μ‹œν•˜μ„Έμš”
"""
# Base prompt
final_prompt = base_prompt
# Add mode-specific guidance
if mode in mode_prompts:
final_prompt += "\n" + mode_prompts[mode]
# Style
if style in style_guides:
final_prompt += f"\n\n뢄석 μŠ€νƒ€μΌ: {style_guides[style]}"
# Always include dataset guides
final_prompt += f"\n\n{dataset_guide}"
final_prompt += f"\n\n{crop_recommendation_guide}"
final_prompt += f"\n\n{climate_impact_guide}"
# Conditionally add soybean dataset guide if selected in UI
if st.session_state.get('use_soybean_dataset', False):
final_prompt += f"\n\n{soybean_guide}"
if include_search_results:
final_prompt += f"\n\n{search_guide}"
if include_uploaded_files:
final_prompt += f"\n\n{upload_guide}"
final_prompt += """
\n\n응닡 ν˜•μ‹ μš”κ΅¬μ‚¬ν•­:
- λ§ˆν¬λ‹€μš΄ 제λͺ©(## 및 ###)을 μ‚¬μš©ν•˜μ—¬ 응닡을 μ²΄κ³„μ μœΌλ‘œ κ΅¬μ„±ν•˜μ„Έμš”
- μ€‘μš”ν•œ 점은 ꡡ은 ν…μŠ€νŠΈ(**ν…μŠ€νŠΈ**)둜 κ°•μ‘°ν•˜μ„Έμš”
- 3-5개의 후속 μ§ˆλ¬Έμ„ ν¬ν•¨ν•œ "κ΄€λ ¨ 질문" μ„Ήμ…˜μ„ λ§ˆμ§€λ§‰μ— μΆ”κ°€ν•˜μ„Έμš”
- μ μ ˆν•œ 간격과 단락 κ΅¬λΆ„μœΌλ‘œ 응닡을 μ„œμ‹ν™”ν•˜μ„Έμš”
- λͺ¨λ“  λ§ν¬λŠ” λ§ˆν¬λ‹€μš΄ ν˜•μ‹μœΌλ‘œ 클릭 κ°€λŠ₯ν•˜κ²Œ λ§Œλ“œμ„Έμš”: [ν…μŠ€νŠΈ](url)
- κ°€λŠ₯ν•œ 경우 데이터λ₯Ό μ‹œκ°μ μœΌλ‘œ ν‘œν˜„(ν‘œ, κ·Έλž˜ν”„ λ“±μ˜ μ„€λͺ…)ν•˜μ„Έμš”
"""
return final_prompt
# ──────────────────────────────── Brave Search API ────────────────────────
@st.cache_data(ttl=3600)
def brave_search(query: str, count: int = 10):
if not BRAVE_KEY:
raise RuntimeError("⚠️ SERPHOUSE_API_KEY (Brave API Key) environment variable is empty.")
headers = {"Accept": "application/json", "Accept-Encoding": "gzip", "X-Subscription-Token": BRAVE_KEY}
params = {"q": query + " 농산물 가격 동ν–₯ 농업 데이터", "count": str(count)}
for attempt in range(3):
try:
r = requests.get(BRAVE_ENDPOINT, headers=headers, params=params, timeout=15)
r.raise_for_status()
data = r.json()
raw = data.get("web", {}).get("results") or data.get("results", [])
if not raw:
logging.warning(f"No Brave search results found. Response: {data}")
raise ValueError("No search results found.")
arts = []
for i, res in enumerate(raw[:count], 1):
url = res.get("url", res.get("link", ""))
host = re.sub(r"https?://(www\.)?", "", url).split("/")[0]
arts.append({
"index": i,
"title": res.get("title", "No title"),
"link": url,
"snippet": res.get("description", res.get("text", "No snippet")),
"displayed_link": host
})
return arts
except Exception as e:
logging.error(f"Brave search failure (attempt {attempt+1}/3): {e}")
if attempt < 2:
time.sleep(5)
return []
@st.cache_data(ttl=3600)
def brave_video_search(query: str, count: int = 3):
if not BRAVE_KEY:
raise RuntimeError("⚠️ SERPHOUSE_API_KEY (Brave API Key) environment variable is empty.")
headers = {"Accept": "application/json","Accept-Encoding": "gzip","X-Subscription-Token": BRAVE_KEY}
params = {"q": query + " 농산물 가격 농업 μ‹œμž₯", "count": str(count)}
for attempt in range(3):
try:
r = requests.get(BRAVE_VIDEO_ENDPOINT, headers=headers, params=params, timeout=15)
r.raise_for_status()
data = r.json()
results = []
for i, vid in enumerate(data.get("results", [])[:count], 1):
results.append({
"index": i,
"title": vid.get("title", "Video"),
"video_url": vid.get("url", ""),
"thumbnail_url": vid.get("thumbnail", {}).get("src", ""),
"source": vid.get("provider", {}).get("name", "Unknown source")
})
return results
except Exception as e:
logging.error(f"Brave video search failure (attempt {attempt+1}/3): {e}")
if attempt < 2:
time.sleep(5)
return []
@st.cache_data(ttl=3600)
def brave_news_search(query: str, count: int = 3):
if not BRAVE_KEY:
raise RuntimeError("⚠️ SERPHOUSE_API_KEY (Brave API Key) environment variable is empty.")
headers = {"Accept": "application/json","Accept-Encoding": "gzip","X-Subscription-Token": BRAVE_KEY}
params = {"q": query + " 농산물 가격 동ν–₯ 농업", "count": str(count)}
for attempt in range(3):
try:
r = requests.get(BRAVE_NEWS_ENDPOINT, headers=headers, params=params, timeout=15)
r.raise_for_status()
data = r.json()
results = []
for i, news in enumerate(data.get("results", [])[:count], 1):
results.append({
"index": i,
"title": news.get("title", "News article"),
"url": news.get("url", ""),
"description": news.get("description", ""),
"source": news.get("source", "Unknown source"),
"date": news.get("age", "Unknown date")
})
return results
except Exception as e:
logging.error(f"Brave news search failure (attempt {attempt+1}/3): {e}")
if attempt < 2:
time.sleep(5)
return []
def mock_results(query: str) -> str:
ts = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
return (f"# λŒ€μ²΄ 검색 μ½˜ν…μΈ  (생성 μ‹œκ°„: {ts})\n\n"
f"'{query}'에 λŒ€ν•œ 검색 API μš”μ²­μ΄ μ‹€νŒ¨ν–ˆκ±°λ‚˜ κ²°κ³Όκ°€ μ—†μŠ΅λ‹ˆλ‹€. "
f"κΈ°μ‘΄ 지식을 기반으둜 응닡을 μƒμ„±ν•΄μ£Όμ„Έμš”.\n\n"
f"λ‹€μŒ 사항을 κ³ λ €ν•˜μ„Έμš”:\n\n"
f"- {query}에 κ΄€ν•œ κΈ°λ³Έ κ°œλ…κ³Ό μ€‘μš”μ„±\n"
f"- 일반적으둜 μ•Œλ €μ§„ κ΄€λ ¨ ν†΅κ³„λ‚˜ μΆ”μ„Έ\n"
f"- 이 μ£Όμ œμ— λŒ€ν•œ μ „λ¬Έκ°€ 의견\n"
f"- λ…μžκ°€ κ°€μ§ˆ 수 μžˆλŠ” 질문\n\n"
f"μ°Έκ³ : μ΄λŠ” μ‹€μ‹œκ°„ 데이터가 μ•„λ‹Œ λŒ€μ²΄ μ§€μΉ¨μž…λ‹ˆλ‹€.\n\n")
def do_web_search(query: str) -> str:
try:
arts = brave_search(query, 10)
if not arts:
logging.warning("No search results, using fallback content")
return mock_results(query)
videos = brave_video_search(query, 2)
news = brave_news_search(query, 3)
result = "# μ›Ή 검색 κ²°κ³Ό\nλ‹€μŒ κ²°κ³Όλ₯Ό ν™œμš©ν•˜μ—¬ 데이터셋 뢄석을 λ³΄μ™„ν•˜λŠ” 포괄적인 닡변을 μ œκ³΅ν•˜μ„Έμš”.\n\n"
result += "## μ›Ή κ²°κ³Ό\n\n"
for a in arts[:5]:
result += f"### κ²°κ³Ό {a['index']}: {a['title']}\n\n{a['snippet']}\n\n"
result += f"**좜처**: [{a['displayed_link']}]({a['link']})\n\n---\n"
if news:
result += "## λ‰΄μŠ€ κ²°κ³Ό\n\n"
for n in news:
result += f"### {n['title']}\n\n{n['description']}\n\n"
result += f"**좜처**: [{n['source']}]({n['url']}) - {n['date']}\n\n---\n"
if videos:
result += "## λΉ„λ””μ˜€ κ²°κ³Ό\n\n"
for vid in videos:
result += f"### {vid['title']}\n\n"
if vid.get('thumbnail_url'):
result += f"![썸넀일]({vid['thumbnail_url']})\n\n"
result += f"**μ‹œμ²­**: [{vid['source']}]({vid['video_url']})\n\n"
return result
except Exception as e:
logging.error(f"Web search process failed: {str(e)}")
return mock_results(query)
# ──────────────────────────────── File Upload Handling ─────────────────────
def process_text_file(file):
try:
content = file.read()
file.seek(0)
text = content.decode('utf-8', errors='ignore')
if len(text) > 10000:
text = text[:9700] + "...(truncated)..."
result = f"## ν…μŠ€νŠΈ 파일: {file.name}\n\n" + text
return result
except Exception as e:
logging.error(f"Error processing text file: {str(e)}")
return f"ν…μŠ€νŠΈ 파일 처리 였λ₯˜: {str(e)}"
def process_csv_file(file):
try:
content = file.read()
file.seek(0)
df = pd.read_csv(io.BytesIO(content))
result = f"## CSV 파일: {file.name}\n\n"
result += f"- ν–‰: {len(df)}\n"
result += f"- μ—΄: {len(df.columns)}\n"
result += f"- μ—΄ 이름: {', '.join(df.columns.tolist())}\n\n"
result += "### 데이터 미리보기\n\n"
preview_df = df.head(10)
try:
markdown_table = preview_df.to_markdown(index=False)
if markdown_table:
result += markdown_table + "\n\n"
else:
result += "CSV 데이터λ₯Ό ν‘œμ‹œν•  수 μ—†μŠ΅λ‹ˆλ‹€.\n\n"
except Exception as e:
logging.error(f"Markdown table conversion error: {e}")
result += "ν…μŠ€νŠΈλ‘œ 데이터 ν‘œμ‹œ:\n\n" + str(preview_df) + "\n\n"
num_cols = df.select_dtypes(include=['number']).columns
if len(num_cols) > 0:
result += "### κΈ°λ³Έ 톡계 정보\n\n"
try:
stats_df = df[num_cols].describe().round(2)
stats_markdown = stats_df.to_markdown()
if stats_markdown:
result += stats_markdown + "\n\n"
else:
result += "톡계 정보λ₯Ό ν‘œμ‹œν•  수 μ—†μŠ΅λ‹ˆλ‹€.\n\n"
except Exception as e:
logging.error(f"Statistical info conversion error: {e}")
result += "톡계 정보λ₯Ό 생성할 수 μ—†μŠ΅λ‹ˆλ‹€.\n\n"
return result
except Exception as e:
logging.error(f"CSV file processing error: {str(e)}")
return f"CSV 파일 처리 였λ₯˜: {str(e)}"
def process_pdf_file(file):
try:
file_bytes = file.read()
file.seek(0)
pdf_file = io.BytesIO(file_bytes)
reader = PyPDF2.PdfReader(pdf_file, strict=False)
result = f"## PDF 파일: {file.name}\n\n- 총 νŽ˜μ΄μ§€: {len(reader.pages)}\n\n"
max_pages = min(5, len(reader.pages))
all_text = ""
for i in range(max_pages):
try:
page = reader.pages[i]
page_text = page.extract_text()
current_page_text = f"### νŽ˜μ΄μ§€ {i+1}\n\n"
if page_text and len(page_text.strip()) > 0:
if len(page_text) > 1500:
current_page_text += page_text[:1500] + "...(좕약됨)...\n\n"
else:
current_page_text += page_text + "\n\n"
else:
current_page_text += "(ν…μŠ€νŠΈλ₯Ό μΆ”μΆœν•  수 μ—†μŒ)\n\n"
all_text += current_page_text
if len(all_text) > 8000:
all_text += "...(λ‚˜λ¨Έμ§€ νŽ˜μ΄μ§€ 좕약됨)...\n\n"
break
except Exception as page_err:
logging.error(f"Error processing PDF page {i+1}: {str(page_err)}")
all_text += f"### νŽ˜μ΄μ§€ {i+1}\n\n(λ‚΄μš© μΆ”μΆœ 였λ₯˜: {str(page_err)})\n\n"
if len(reader.pages) > max_pages:
all_text += f"\nμ°Έκ³ : 처음 {max_pages} νŽ˜μ΄μ§€λ§Œ ν‘œμ‹œλ©λ‹ˆλ‹€.\n\n"
result += "### PDF λ‚΄μš©\n\n" + all_text
return result
except Exception as e:
logging.error(f"PDF file processing error: {str(e)}")
return f"## PDF 파일: {file.name}\n\n였λ₯˜: {str(e)}\n\nμ²˜λ¦¬ν•  수 μ—†μŠ΅λ‹ˆλ‹€."
def process_uploaded_files(files):
if not files:
return None
result = "# μ—…λ‘œλ“œλœ 파일 λ‚΄μš©\n\nμ‚¬μš©μžκ°€ μ œκ³΅ν•œ 파일의 λ‚΄μš©μž…λ‹ˆλ‹€.\n\n"
for file in files:
try:
ext = file.name.split('.')[-1].lower()
if ext == 'txt':
result += process_text_file(file) + "\n\n---\n\n"
elif ext == 'csv':
result += process_csv_file(file) + "\n\n---\n\n"
elif ext == 'pdf':
result += process_pdf_file(file) + "\n\n---\n\n"
else:
result += f"### μ§€μ›λ˜μ§€ μ•ŠλŠ” 파일: {file.name}\n\n---\n\n"
except Exception as e:
logging.error(f"File processing error {file.name}: {e}")
result += f"### 파일 처리 였λ₯˜: {file.name}\n\n였λ₯˜: {e}\n\n---\n\n"
return result
# ──────────────────────────────── Image & Utility ─────────────────────────
def generate_image(prompt, w=768, h=768, g=3.5, steps=30, seed=3):
if not prompt:
return None, "Insufficient prompt"
try:
res = Client(IMAGE_API_URL).predict(
prompt=prompt, width=w, height=h, guidance=g,
inference_steps=steps, seed=seed,
do_img2img=False, init_image=None,
image2image_strength=0.8, resize_img=True,
api_name="/generate_image"
)
return res[0], f"Seed: {res[1]}"
except Exception as e:
logging.error(e)
return None, str(e)
def extract_image_prompt(response_text: str, topic: str):
client = get_openai_client()
try:
response = client.chat.completions.create(
model="gpt-4.1-mini",
messages=[
{"role": "system", "content": "농업 및 농산물에 κ΄€ν•œ 이미지 ν”„λ‘¬ν”„νŠΈλ₯Ό μƒμ„±ν•©λ‹ˆλ‹€. ν•œ μ€„μ˜ μ˜μ–΄λ‘œ 된 ν”„λ‘¬ν”„νŠΈλ§Œ λ°˜ν™˜ν•˜μ„Έμš”, λ‹€λ₯Έ ν…μŠ€νŠΈλŠ” ν¬ν•¨ν•˜μ§€ λ§ˆμ„Έμš”."},
{"role": "user", "content": f"주제: {topic}\n\n---\n{response_text}\n\n---"}
],
temperature=1,
max_tokens=80,
top_p=1
)
return response.choices[0].message.content.strip()
except Exception as e:
logging.error(f"OpenAI image prompt generation error: {e}")
return f"A professional photograph of agricultural produce and farm fields, data visualization of crop prices and trends, high quality"
def md_to_html(md: str, title="농산물 μˆ˜μš” 예츑 뢄석 κ²°κ³Ό"):
return f"<!DOCTYPE html><html><head><title>{title}</title><meta charset='utf-8'></head><body>{markdown.markdown(md)}</body></html>"
def keywords(text: str, top=5):
cleaned = re.sub(r"[^κ°€-힣a-zA-Z0-9\s]", "", text)
return " ".join(cleaned.split()[:top])
# ──────────────────────────────── Streamlit UI ────────────────────────────
def agricultural_price_forecast_app():
st.title("농산물 μˆ˜μš” 및 가격 예츑 AI μ–΄μ‹œμŠ€ν„΄νŠΈ")
st.markdown("UN κΈ€λ‘œλ²Œ μ‹λŸ‰ 및 농업 톡계 데이터셋 뢄석 기반의 농산물 μ‹œμž₯ 예츑")
if "ai_model" not in st.session_state:
st.session_state.ai_model = "gpt-4.1-mini"
if "messages" not in st.session_state:
st.session_state.messages = []
if "auto_save" not in st.session_state:
st.session_state.auto_save = True
if "generate_image" not in st.session_state:
st.session_state.generate_image = False
if "web_search_enabled" not in st.session_state:
st.session_state.web_search_enabled = True
if "analysis_mode" not in st.session_state:
st.session_state.analysis_mode = "price_forecast"
if "response_style" not in st.session_state:
st.session_state.response_style = "professional"
if "use_soybean_dataset" not in st.session_state:
st.session_state.use_soybean_dataset = False
sb = st.sidebar
sb.title("뢄석 μ„€μ •")
# Kaggle dataset info display
if sb.checkbox("데이터셋 정보 ν‘œμ‹œ", value=False):
st.info("UN κΈ€λ‘œλ²Œ μ‹λŸ‰ 및 농업 톡계 데이터셋을 λΆˆλŸ¬μ˜€λŠ” 쀑...")
dataset_info = load_agriculture_dataset()
if dataset_info:
st.success(f"데이터셋 λ‘œλ“œ μ™„λ£Œ: {len(dataset_info['files'])}개 파일")
with st.expander("데이터셋 미리보기", expanded=False):
for file_info in dataset_info['files'][:5]:
st.write(f"**{file_info['name']}** ({file_info['size_mb']} MB)")
else:
st.error("데이터셋을 λΆˆλŸ¬μ˜€λŠ”λ° μ‹€νŒ¨ν–ˆμŠ΅λ‹ˆλ‹€. Kaggle API 섀정을 ν™•μΈν•˜μ„Έμš”.")
sb.subheader("뢄석 ꡬ성")
sb.selectbox(
"뢄석 λͺ¨λ“œ",
options=list(ANALYSIS_MODES.keys()),
format_func=lambda x: ANALYSIS_MODES[x],
key="analysis_mode"
)
sb.selectbox(
"응닡 μŠ€νƒ€μΌ",
options=list(RESPONSE_STYLES.keys()),
format_func=lambda x: RESPONSE_STYLES[x],
key="response_style"
)
# Dataset selection
sb.subheader("데이터셋 선택")
sb.checkbox(
"κ³ κΈ‰ λŒ€λ‘ 농업 데이터셋 μ‚¬μš©",
key="use_soybean_dataset",
help="λŒ€λ‘(콩) κ΄€λ ¨ μ§ˆλ¬Έμ— 더 μ •ν™•ν•œ 정보λ₯Ό μ œκ³΅ν•©λ‹ˆλ‹€."
)
# Always enabled datasets info
sb.info("κΈ°λ³Έ ν™œμ„±ν™”λœ 데이터셋:\n- UN κΈ€λ‘œλ²Œ μ‹λŸ‰ 및 농업 톡계\n- ν† μ–‘ 및 ν™˜κ²½ λ³€μˆ˜ 기반 μž‘λ¬Ό μΆ”μ²œ\n- κΈ°ν›„ λ³€ν™”κ°€ 농업에 λ―ΈμΉ˜λŠ” 영ν–₯")
# Example queries
sb.subheader("μ˜ˆμ‹œ 질문")
c1, c2, c3 = sb.columns(3)
if c1.button("μŒ€ 가격 전망", key="ex1"):
process_example(EXAMPLE_QUERIES["example1"])
if c2.button("κΈ°ν›„ 영ν–₯", key="ex2"):
process_example(EXAMPLE_QUERIES["example2"])
if c3.button("증평ꡰ μž‘λ¬Ό", key="ex3"):
process_example(EXAMPLE_QUERIES["example3"])
sb.subheader("기타 μ„€μ •")
sb.toggle("μžλ™ μ €μž₯", key="auto_save")
sb.toggle("이미지 μžλ™ 생성", key="generate_image")
web_search_enabled = sb.toggle("μ›Ή 검색 μ‚¬μš©", value=st.session_state.web_search_enabled)
st.session_state.web_search_enabled = web_search_enabled
if web_search_enabled:
st.sidebar.info("βœ… μ›Ή 검색 κ²°κ³Όκ°€ 응닡에 ν†΅ν•©λ©λ‹ˆλ‹€.")
# Download the latest response
latest_response = next(
(m["content"] for m in reversed(st.session_state.messages)
if m["role"] == "assistant" and m["content"].strip()),
None
)
if latest_response:
title_match = re.search(r"# (.*?)(\n|$)", latest_response)
if title_match:
title = title_match.group(1).strip()
else:
first_line = latest_response.split('\n', 1)[0].strip()
title = first_line[:40] + "..." if len(first_line) > 40 else first_line
sb.subheader("μ΅œμ‹  응닡 λ‹€μš΄λ‘œλ“œ")
d1, d2 = sb.columns(2)
d1.download_button("λ§ˆν¬λ‹€μš΄μœΌλ‘œ λ‹€μš΄λ‘œλ“œ", latest_response,
file_name=f"{title}.md", mime="text/markdown")
d2.download_button("HTML둜 λ‹€μš΄λ‘œλ“œ", md_to_html(latest_response, title),
file_name=f"{title}.html", mime="text/html")
# JSON conversation record upload
up = sb.file_uploader("λŒ€ν™” 기둝 뢈러였기 (.json)", type=["json"], key="json_uploader")
if up:
try:
st.session_state.messages = json.load(up)
sb.success("λŒ€ν™” 기둝을 μ„±κ³΅μ μœΌλ‘œ λΆˆλŸ¬μ™”μŠ΅λ‹ˆλ‹€")
except Exception as e:
sb.error(f"뢈러였기 μ‹€νŒ¨: {e}")
# JSON conversation record download
if sb.button("λŒ€ν™” 기둝을 JSON으둜 λ‹€μš΄λ‘œλ“œ"):
sb.download_button(
"μ €μž₯",
data=json.dumps(st.session_state.messages, ensure_ascii=False, indent=2),
file_name="conversation_history.json",
mime="application/json"
)
# File Upload
st.subheader("파일 μ—…λ‘œλ“œ")
uploaded_files = st.file_uploader(
"μ°Έκ³  자료둜 μ‚¬μš©ν•  파일 μ—…λ‘œλ“œ (txt, csv, pdf)",
type=["txt", "csv", "pdf"],
accept_multiple_files=True,
key="file_uploader"
)
if uploaded_files:
file_count = len(uploaded_files)
st.success(f"{file_count}개 파일이 μ—…λ‘œλ“œλ˜μ—ˆμŠ΅λ‹ˆλ‹€. μ§ˆμ˜μ— λŒ€ν•œ μ†ŒμŠ€λ‘œ μ‚¬μš©λ©λ‹ˆλ‹€.")
with st.expander("μ—…λ‘œλ“œλœ 파일 미리보기", expanded=False):
for idx, file in enumerate(uploaded_files):
st.write(f"**파일λͺ…:** {file.name}")
ext = file.name.split('.')[-1].lower()
if ext == 'txt':
preview = file.read(1000).decode('utf-8', errors='ignore')
file.seek(0)
st.text_area(
f"{file.name} 미리보기",
preview + ("..." if len(preview) >= 1000 else ""),
height=150
)
elif ext == 'csv':
try:
df = pd.read_csv(file)
file.seek(0)
st.write("CSV 미리보기 (μ΅œλŒ€ 5ν–‰)")
st.dataframe(df.head(5))
except Exception as e:
st.error(f"CSV 미리보기 μ‹€νŒ¨: {e}")
elif ext == 'pdf':
try:
file_bytes = file.read()
file.seek(0)
pdf_file = io.BytesIO(file_bytes)
reader = PyPDF2.PdfReader(pdf_file, strict=False)
pc = len(reader.pages)
st.write(f"PDF 파일: {pc}νŽ˜μ΄μ§€")
if pc > 0:
try:
page_text = reader.pages[0].extract_text()
preview = page_text[:500] if page_text else "(ν…μŠ€νŠΈ μΆ”μΆœ λΆˆκ°€)"
st.text_area("첫 νŽ˜μ΄μ§€ 미리보기", preview + "...", height=150)
except:
st.warning("첫 νŽ˜μ΄μ§€ ν…μŠ€νŠΈ μΆ”μΆœ μ‹€νŒ¨")
except Exception as e:
st.error(f"PDF 미리보기 μ‹€νŒ¨: {e}")
if idx < file_count - 1:
st.divider()
# Display existing messages
for m in st.session_state.messages:
with st.chat_message(m["role"]):
st.markdown(m["content"], unsafe_allow_html=True)
# Videos
if "videos" in m and m["videos"]:
st.subheader("κ΄€λ ¨ λΉ„λ””μ˜€")
for video in m["videos"]:
video_title = video.get('title', 'κ΄€λ ¨ λΉ„λ””μ˜€')
video_url = video.get('url', '')
thumbnail = video.get('thumbnail', '')
if thumbnail:
col1, col2 = st.columns([1, 3])
with col1:
st.write("🎬")
with col2:
st.markdown(f"**[{video_title}]({video_url})**")
st.write(f"좜처: {video.get('source', 'μ•Œ 수 μ—†μŒ')}")
else:
st.markdown(f"🎬 **[{video_title}]({video_url})**")
st.write(f"좜처: {video.get('source', 'μ•Œ 수 μ—†μŒ')}")
# User input
query = st.chat_input("농산물 가격, μˆ˜μš” λ˜λŠ” μ‹œμž₯ 동ν–₯ κ΄€λ ¨ μ§ˆλ¬Έμ„ μž…λ ₯ν•˜μ„Έμš”.")
if query:
process_input(query, uploaded_files)
sb.markdown("---")
sb.markdown("Created by Vidraft | [Community](https://discord.gg/openfreeai)")
def process_example(topic):
process_input(topic, [])
def process_input(query: str, uploaded_files):
if not any(m["role"] == "user" and m["content"] == query for m in st.session_state.messages):
st.session_state.messages.append({"role": "user", "content": query})
with st.chat_message("user"):
st.markdown(query)
with st.chat_message("assistant"):
placeholder = st.empty()
message_placeholder = st.empty()
full_response = ""
use_web_search = st.session_state.web_search_enabled
has_uploaded_files = bool(uploaded_files) and len(uploaded_files) > 0
try:
status = st.status("μ§ˆλ¬Έμ— λ‹΅λ³€ μ€€λΉ„ 쀑...")
status.update(label="ν΄λΌμ΄μ–ΈνŠΈ μ΄ˆκΈ°ν™” 쀑...")
client = get_openai_client()
search_content = None
video_results = []
news_results = []
# 농업 데이터셋 뢄석 κ²°κ³Ό κ°€μ Έμ˜€κΈ°
status.update(label="농업 데이터셋 뢄석 쀑...")
with st.spinner("데이터셋 뢄석 쀑..."):
dataset_analysis = analyze_dataset_for_query(query)
# 항상 ν¬ν•¨λ˜λŠ” μΆ”κ°€ 데이터셋 뢄석
crop_recommendation_analysis = analyze_crop_recommendation_dataset(query)
climate_impact_analysis = analyze_climate_impact_dataset(query)
#
# 쑰건뢀 데이터셋 뢄석
soybean_analysis = None
if st.session_state.use_soybean_dataset:
status.update(label="λŒ€λ‘ 농업 데이터셋 뢄석 쀑...")
with st.spinner("λŒ€λ‘ 데이터셋 뢄석 쀑..."):
soybean_analysis = analyze_soybean_dataset(query)
if use_web_search:
# μ›Ή 검색 과정은 λ…ΈμΆœν•˜μ§€ μ•Šκ³  쑰용히 μ§„ν–‰
with st.spinner("정보 μˆ˜μ§‘ 쀑..."):
search_content = do_web_search(keywords(query, top=5))
video_results = brave_video_search(query, 2)
news_results = brave_news_search(query, 3)
file_content = None
if has_uploaded_files:
status.update(label="μ—…λ‘œλ“œλœ 파일 처리 쀑...")
with st.spinner("파일 뢄석 쀑..."):
file_content = process_uploaded_files(uploaded_files)
valid_videos = []
for vid in video_results:
url = vid.get('video_url')
if url and url.startswith('http'):
valid_videos.append({
'url': url,
'title': vid.get('title', 'λΉ„λ””μ˜€'),
'thumbnail': vid.get('thumbnail_url', ''),
'source': vid.get('source', 'λΉ„λ””μ˜€ 좜처')
})
status.update(label="μ’…ν•© 뢄석 μ€€λΉ„ 쀑...")
sys_prompt = get_system_prompt(
mode=st.session_state.analysis_mode,
style=st.session_state.response_style,
include_search_results=use_web_search,
include_uploaded_files=has_uploaded_files
)
api_messages = [
{"role": "system", "content": sys_prompt}
]
user_content = query
# 항상 κΈ°λ³Έ 데이터셋 뢄석 κ²°κ³Ό 포함
user_content += "\n\n" + dataset_analysis
user_content += "\n\n" + crop_recommendation_analysis
user_content += "\n\n" + climate_impact_analysis
# 쑰건뢀 데이터셋 κ²°κ³Ό 포함
if soybean_analysis:
user_content += "\n\n" + soybean_analysis
if search_content:
user_content += "\n\n" + search_content
if file_content:
user_content += "\n\n" + file_content
if valid_videos:
user_content += "\n\n# κ΄€λ ¨ λ™μ˜μƒ\n"
for i, vid in enumerate(valid_videos):
user_content += f"\n{i+1}. **{vid['title']}** - [{vid['source']}]({vid['url']})\n"
api_messages.append({"role": "user", "content": user_content})
try:
stream = client.chat.completions.create(
model="gpt-4.1-mini",
messages=api_messages,
temperature=1,
max_tokens=MAX_TOKENS,
top_p=1,
stream=True
)
for chunk in stream:
if chunk.choices and len(chunk.choices) > 0 and chunk.choices[0].delta.content is not None:
content_delta = chunk.choices[0].delta.content
full_response += content_delta
message_placeholder.markdown(full_response + "β–Œ", unsafe_allow_html=True)
message_placeholder.markdown(full_response, unsafe_allow_html=True)
if valid_videos:
st.subheader("κ΄€λ ¨ λΉ„λ””μ˜€")
for video in valid_videos:
video_title = video.get('title', 'κ΄€λ ¨ λΉ„λ””μ˜€')
video_url = video.get('url', '')
st.markdown(f"🎬 **[{video_title}]({video_url})**")
st.write(f"좜처: {video.get('source', 'μ•Œ 수 μ—†μŒ')}")
status.update(label="응닡 μ™„λ£Œ!", state="complete")
st.session_state.messages.append({
"role": "assistant",
"content": full_response,
"videos": valid_videos
})
except Exception as api_error:
error_message = str(api_error)
logging.error(f"API 였λ₯˜: {error_message}")
status.update(label=f"였λ₯˜: {error_message}", state="error")
raise Exception(f"응닡 생성 였λ₯˜: {error_message}")
if st.session_state.generate_image and full_response:
with st.spinner("λ§žμΆ€ν˜• 이미지 생성 쀑..."):
try:
ip = extract_image_prompt(full_response, query)
img, cap = generate_image(ip)
if img:
st.subheader("AI 생성 이미지")
st.image(img, caption=cap, use_container_width=True)
except Exception as img_error:
logging.error(f"이미지 생성 였λ₯˜: {str(img_error)}")
st.warning("λ§žμΆ€ν˜• 이미지 생성에 μ‹€νŒ¨ν–ˆμŠ΅λ‹ˆλ‹€.")
if full_response:
st.subheader("이 응닡 λ‹€μš΄λ‘œλ“œ")
c1, c2 = st.columns(2)
c1.download_button(
"λ§ˆν¬λ‹€μš΄",
data=full_response,
file_name=f"{query[:30]}.md",
mime="text/markdown"
)
c2.download_button(
"HTML",
data=md_to_html(full_response, query[:30]),
file_name=f"{query[:30]}.html",
mime="text/html"
)
if st.session_state.auto_save and st.session_state.messages:
try:
fn = f"conversation_history_auto_{datetime.now():%Y%m%d_%H%M%S}.json"
with open(fn, "w", encoding="utf-8") as fp:
json.dump(st.session_state.messages, fp, ensure_ascii=False, indent=2)
except Exception as e:
logging.error(f"μžλ™ μ €μž₯ μ‹€νŒ¨: {e}")
except Exception as e:
error_message = str(e)
placeholder.error(f"였λ₯˜ λ°œμƒ: {error_message}")
logging.error(f"μž…λ ₯ 처리 였λ₯˜: {error_message}")
ans = f"μš”μ²­ 처리 쀑 였λ₯˜κ°€ λ°œμƒν–ˆμŠ΅λ‹ˆλ‹€: {error_message}"
st.session_state.messages.append({"role": "assistant", "content": ans})
# ──────────────────────────────── main ────────────────────────────────────
def main():
st.write("==== μ• ν”Œλ¦¬μΌ€μ΄μ…˜ μ‹œμž‘ μ‹œκ°„:", datetime.now().strftime("%Y-%m-%d %H:%M:%S"), "=====")
agricultural_price_forecast_app()
if __name__ == "__main__":
main()