File size: 38,132 Bytes
1729de6 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 | """
TenAI PMAI Pro - ๊ธฐ์
์์ฐ๊ด๋ฆฌ AI ํ๋ซํผ
Zero Hardware | Data Monopoly | AI Transformation
Fireworks Vision API + Groq LLM
"""
import gradio as gr
import requests
import os, json, base64, re
from typing import Generator, Optional, List, Dict
import pandas as pd
import numpy as np
try:
import fitz
PDF_AVAILABLE = True
except ImportError:
PDF_AVAILABLE = False
try:
from PIL import Image
IMAGE_AVAILABLE = True
except ImportError:
IMAGE_AVAILABLE = False
try:
import folium
from folium.plugins import HeatMap
FOLIUM_AVAILABLE = True
except ImportError:
FOLIUM_AVAILABLE = False
try:
import plotly.express as px
import plotly.graph_objects as go
PLOTLY_AVAILABLE = True
except ImportError:
PLOTLY_AVAILABLE = False
try:
from datasets import load_dataset
DATASETS_AVAILABLE = True
except ImportError:
DATASETS_AVAILABLE = False
FIREWORKS_API_KEY = os.environ.get("FIREWORKS_API_KEY", "")
GROQ_API_KEY = os.environ.get("GROQ_API_KEY", "")
FIREWORKS_VISION_URL = "https://api.fireworks.ai/inference/v1/chat/completions"
FIREWORKS_VISION_MODEL = "accounts/fireworks/models/qwen3-vl-235b-a22b-thinking"
DEMO_MODE = not FIREWORKS_API_KEY and not GROQ_API_KEY
groq_client = None
if GROQ_API_KEY:
try:
from groq import Groq
groq_client = Groq(api_key=GROQ_API_KEY)
except:
pass
PMAI_SYSTEM_PROMPT = """๋น์ ์ TenAI์ PMAI(Property Management AI)์
๋๋ค. ๊ธฐ์
์์ฐ๊ด๋ฆฌ๋ฅผ ์ํ 24์๊ฐ AI ๋น์์
๋๋ค.
ํต์ฌ ์ญ๋: Vision AI + RAG + ์ถ๋ก ์์ง ๊ธฐ๋ฐ ๋ฌธ์ ๋ถ์, ๋น์ ํ ๋ฌธ์์ ๋๋ฉด ์ดํด
์ ๊ณต ๊ฐ์น: AMC(์์ฐ์ด์ฉ)-์ค์๊ฐ ๊ฐ์นํ๊ฐ, PMC(์๋๊ด๋ฆฌ)-๊ณต์ค๊ด๋ฆฌ ์๋ํ, FMC(์์ค๊ด๋ฆฌ)-์์ค์ ๊ฒ ์๋ํ
ํน์ง: Zero Hardware-์ผ์ ์์ด ์ฆ์ ๋์
, ๋ฌธ์ ๊ธฐ๋ฐ ๋ถ์, ๋น์ฉ ์ ๊ฐ & ์์ต ์ฆ๋
ํ๊ตญ์ด๋ก ์น์ ํ๊ณ ์ ๋ฌธ์ ์ผ๋ก ์๋ตํ์ธ์."""
DOCUMENT_ANALYSIS_PROMPT = """๋น์ ์ TenAI์ ๋ฌธ์ ๋ถ์ AI์
๋๋ค. ์
๋ก๋๋ ๋ฌธ์๋ฅผ ๋ถ์ํ์ฌ:
1. ํต์ฌ ์ ๋ณด ์ถ์ถ 2. ์ ์ฌ์ ๋ฆฌ์คํฌ ์๋ณ 3. ๋น์ฉ ์ต์ ํ ํฌ์ธํธ ๋์ถ 4. ์คํ ๊ฐ๋ฅํ ์ธ์ฌ์ดํธ ์ ๊ณต"""
COST_ANALYSIS_PROMPT = """๋น์ ์ TenAI์ ๋น์ฉ ๋ถ์ AI์
๋๋ค. ์ด์๋น์ฉ ๋ฐ์ดํฐ๋ฅผ ๋ถ์ํ์ฌ:
1. ๋น์ฉ ๊ตฌ์กฐ ๋ถ์ 2. ๋์ ์ง์ ์๋ณ 3. ์ ๊ฐ ๊ฐ๋ฅ ํญ๋ชฉ ๋์ถ 4. ROI ๊ธฐ๋ฐ ์ฐ์ ์์ ์ ์"""
SOMA_AGENTS = {
"coordinator": {"name": "๐ฏ ์ข
ํฉ ์ฝ๋๋ค์ดํฐ", "role": "SOMA ํ์ฅ", "prompt": "๋น์ ์ SOMA ํ์ ์ข
ํฉ ์ฝ๋๋ค์ดํฐ์
๋๋ค. ๊ฐ ์ ๋ฌธ๊ฐ์ ๋ถ์ ๊ฒฐ๊ณผ๋ฅผ ์ข
ํฉํ์ฌ Executive Summary๋ฅผ ์์ฑํฉ๋๋ค. ํ๊ตญ์ด๋ก ์๋ตํ์ธ์."},
"document_analyst": {"name": "๐ ๋ฌธ์ ๋ถ์๊ฐ", "role": "๋ฌธ์/๊ณ์ฝ์ ์ ๋ฌธ", "prompt": "๋น์ ์ SOMA ํ์ ๋ฌธ์ ๋ถ์ ์ ๋ฌธ๊ฐ์
๋๋ค. ์๋์ฐจ๊ณ์ฝ์, ์ ์ง๋ณด์ ๊ณ์ฝ ๋ฑ์ ์ ๋ฐ ๋ถ์ํฉ๋๋ค. ํ๊ตญ์ด๋ก ์๋ตํ์ธ์."},
"financial_expert": {"name": "๐ฐ ์ฌ๋ฌด ์ ๋ฌธ๊ฐ", "role": "๋น์ฉ/์์ต ๋ถ์", "prompt": "๋น์ ์ SOMA ํ์ ์ฌ๋ฌด ๋ถ์ ์ ๋ฌธ๊ฐ์
๋๋ค. ์ฌ๋ฌด์ ์ํฅ์ ๋ถ์ํ๊ณ ๋น์ฉ ์ ๊ฐ ๋ฐฉ์์ ์ ์ํฉ๋๋ค. ํ๊ตญ์ด๋ก ์๋ตํ์ธ์."},
"legal_advisor": {"name": "โ๏ธ ๋ฒ๋ฅ ์๋ฌธ๊ฐ", "role": "๋ฒ์ ๋ฆฌ์คํฌ ๊ฒํ ", "prompt": "๋น์ ์ SOMA ํ์ ๋ฒ๋ฅ ์๋ฌธ ์ ๋ฌธ๊ฐ์
๋๋ค. ๋ฒ์ ๋ฆฌ์คํฌ, ๋ถ๋ฆฌํ ์กฐํญ์ ์๋ณํฉ๋๋ค. ํ๊ตญ์ด๋ก ์๋ตํ์ธ์."},
"facility_manager": {"name": "๐ง ์์ค ๊ด๋ฆฌ์", "role": "์ด์/์์ค ์ ๋ฌธ", "prompt": "๋น์ ์ SOMA ํ์ ์์ค ๊ด๋ฆฌ ์ ๋ฌธ๊ฐ์
๋๋ค. ์์ค ์ด์ ๊ด์ ์์ ๋ถ์ํฉ๋๋ค. ํ๊ตญ์ด๋ก ์๋ตํ์ธ์."},
"market_analyst": {"name": "๐ ์๊ถ ๋ถ์๊ฐ", "role": "์
์ง/์๊ถ ์ ๋ฌธ", "prompt": "๋น์ ์ SOMA ํ์ ์๊ถ ๋ถ์ ์ ๋ฌธ๊ฐ์
๋๋ค. ์
์ง, ์ฃผ๋ณ ์๊ถ, ์ ๋์ธ๊ตฌ๋ฅผ ๋ถ์ํฉ๋๋ค. ํ๊ตญ์ด๋ก ์๋ตํ์ธ์."}
}
SEOUL_DISTRICTS = {
"๊ฐ๋จ๊ตฌ": {"lat": 37.5172, "lng": 127.0473, "ํน์ฑ": "IT/๊ธ์ต ์ค์ฌ, ๊ณ ๊ธ ์คํผ์ค", "ํ๊ท ์๋๋ฃ": 85000},
"์์ด๊ตฌ": {"lat": 37.4837, "lng": 127.0324, "ํน์ฑ": "๋ฒ์กฐํ์ด, ๊ต์ก/๋ฌธํ", "ํ๊ท ์๋๋ฃ": 75000},
"์กํ๊ตฌ": {"lat": 37.5145, "lng": 127.1050, "ํน์ฑ": "์ ์ค ์๊ถ, ์ฃผ๊ฑฐ/์์
๋ณตํฉ", "ํ๊ท ์๋๋ฃ": 65000},
"๋งํฌ๊ตฌ": {"lat": 37.5663, "lng": 126.9014, "ํน์ฑ": "ํ๋/ํฉ์ ์๊ถ, ๋ฌธํ/์์ ", "ํ๊ท ์๋๋ฃ": 55000},
"์๋ฑํฌ๊ตฌ": {"lat": 37.5264, "lng": 126.8963, "ํน์ฑ": "์ฌ์๋ ๊ธ์ต, ํ์์คํ์ด", "ํ๊ท ์๋๋ฃ": 70000},
"์ฉ์ฐ๊ตฌ": {"lat": 37.5324, "lng": 126.9903, "ํน์ฑ": "์ดํ์/ํ๋จ, ์ฌ๊ฐ๋ฐ", "ํ๊ท ์๋๋ฃ": 60000},
"์ฑ๋๊ตฌ": {"lat": 37.5634, "lng": 127.0369, "ํน์ฑ": "์ฑ์๋ ํซํ, ์ง์์ฐ์
", "ํ๊ท ์๋๋ฃ": 55000},
"์ข
๋ก๊ตฌ": {"lat": 37.5735, "lng": 126.9790, "ํน์ฑ": "๋์ฌ CBD, ์ ํต ์๊ถ", "ํ๊ท ์๋๋ฃ": 80000},
"์ค๊ตฌ": {"lat": 37.5641, "lng": 126.9979, "ํน์ฑ": "๋ช
๋/์์ง๋ก, ๊ด๊ด/์์
", "ํ๊ท ์๋๋ฃ": 90000},
"๊ฐ์๊ตฌ": {"lat": 37.5510, "lng": 126.8495, "ํน์ฑ": "๋ง๊ณก์ง๊ตฌ, ์ฐ์
๋จ์ง", "ํ๊ท ์๋๋ฃ": 45000},
"๊ตฌ๋ก๊ตฌ": {"lat": 37.4954, "lng": 126.8874, "ํน์ฑ": "๋์งํธ๋จ์ง, ์ฐ์
", "ํ๊ท ์๋๋ฃ": 40000},
"๊ธ์ฒ๊ตฌ": {"lat": 37.4569, "lng": 126.8956, "ํน์ฑ": "๊ฐ์ฐ๋์งํธ, IT", "ํ๊ท ์๋๋ฃ": 38000},
}
MARKET_REGIONS = {'์์ธ': '์์ธ_202506', '๊ฒฝ๊ธฐ': '๊ฒฝ๊ธฐ_202506', '๋ถ์ฐ': '๋ถ์ฐ_202506', '๋๊ตฌ': '๋๊ตฌ_202506', '์ธ์ฒ': '์ธ์ฒ_202506', '๊ด์ฃผ': '๊ด์ฃผ_202506', '๋์ ': '๋์ _202506', '์ธ์ฐ': '์ธ์ฐ_202506', '์ธ์ข
': '์ธ์ข
_202506', '๊ฒฝ๋จ': '๊ฒฝ๋จ_202506', '๊ฒฝ๋ถ': '๊ฒฝ๋ถ_202506', '์ ๋จ': '์ ๋จ_202506', '์ ๋ถ': '์ ๋ถ_202506', '์ถฉ๋จ': '์ถฉ๋จ_202506', '์ถฉ๋ถ': '์ถฉ๋ถ_202506', '๊ฐ์': '๊ฐ์_202506', '์ ์ฃผ': '์ ์ฃผ_202506'}
class MarketDataLoader:
@staticmethod
def load_region_data(region: str, sample_size: int = 20000) -> pd.DataFrame:
if not DATASETS_AVAILABLE:
return pd.DataFrame()
try:
file_name = f"์์๊ณต์ธ์์ฅ์งํฅ๊ณต๋จ_์๊ฐ(์๊ถ)์ ๋ณด_{MARKET_REGIONS[region]}.csv"
dataset = load_dataset("ginipick/market", data_files=file_name, split="train")
df = dataset.to_pandas()
return df.sample(n=min(sample_size, len(df)), random_state=42)
except:
return pd.DataFrame()
@staticmethod
def load_multiple_regions(regions: List[str], sample_per_region: int = 20000) -> pd.DataFrame:
dfs = [MarketDataLoader.load_region_data(r, sample_per_region) for r in regions]
return pd.concat([d for d in dfs if not d.empty], ignore_index=True) if any(not d.empty for d in dfs) else pd.DataFrame()
class MarketAnalyzer:
def __init__(self, df: pd.DataFrame):
self.df = df
self.prepare_data()
def prepare_data(self):
for col in ['๊ฒฝ๋', '์๋']:
if col in self.df.columns:
self.df[col] = pd.to_numeric(self.df[col], errors='coerce')
self.df = self.df.dropna(subset=['๊ฒฝ๋', '์๋'])
if '์ธต์ ๋ณด' in self.df.columns:
self.df['์ธต์ ๋ณด_์ซ์'] = self.df['์ธต์ ๋ณด'].apply(self._parse_floor)
def _parse_floor(self, floor_str):
if pd.isna(floor_str): return None
s = str(floor_str)
if '์งํ' in s or 'B' in s:
m = re.search(r'\d+', s)
return -int(m.group()) if m else -1
m = re.search(r'\d+', s)
return int(m.group()) if m else None
def get_summary(self) -> Dict:
return {
'์ด์ ํฌ์': len(self.df),
'์
์ข
์': self.df['์๊ถ์
์ข
์ค๋ถ๋ฅ๋ช
'].nunique() if '์๊ถ์
์ข
์ค๋ถ๋ฅ๋ช
' in self.df.columns else 0,
'์์์
์ข
': self.df['์๊ถ์
์ข
์ค๋ถ๋ฅ๋ช
'].value_counts().head(5).to_dict() if '์๊ถ์
์ข
์ค๋ถ๋ฅ๋ช
' in self.df.columns else {},
}
def create_category_chart(self):
if not PLOTLY_AVAILABLE or '์๊ถ์
์ข
์ค๋ถ๋ฅ๋ช
' not in self.df.columns: return None
top = self.df['์๊ถ์
์ข
์ค๋ถ๋ฅ๋ช
'].value_counts().head(15)
fig = px.bar(x=top.values, y=top.index, orientation='h', title='๐ ์์ ์
์ข
TOP 15', color=top.values, color_continuous_scale='blues')
fig.update_layout(showlegend=False, height=450)
return fig
def create_district_chart(self):
if not PLOTLY_AVAILABLE or '์๊ตฐ๊ตฌ๋ช
' not in self.df.columns: return None
counts = self.df['์๊ตฐ๊ตฌ๋ช
'].value_counts().head(15)
fig = px.bar(x=counts.values, y=counts.index, orientation='h', title='๐ ์ง์ญ๋ณ ์ ํฌ ๋ฐ์ง๋', color=counts.values, color_continuous_scale='reds')
fig.update_layout(showlegend=False, height=450)
return fig
def create_heatmap(self, sample_size: int = 2000) -> str:
if not FOLIUM_AVAILABLE: return "<p>folium ์ค์น ํ์</p>"
df_sample = self.df.sample(n=min(sample_size, len(self.df)), random_state=42)
m = folium.Map(location=[df_sample['์๋'].mean(), df_sample['๊ฒฝ๋'].mean()], zoom_start=11, tiles='cartodbpositron')
HeatMap([[r['์๋'], r['๊ฒฝ๋']] for _, r in df_sample.iterrows()], radius=15, blur=25).add_to(m)
return m._repr_html_()
class AppState:
def __init__(self):
self.analyzer = None
app_state = AppState()
def encode_image_to_base64(image_path: str) -> str:
with open(image_path, "rb") as f:
return base64.b64encode(f.read()).decode('utf-8')
def get_image_mime_type(file_path: str) -> str:
ext = file_path.lower().split('.')[-1]
return {'jpg': 'image/jpeg', 'jpeg': 'image/jpeg', 'png': 'image/png', 'gif': 'image/gif', 'webp': 'image/webp'}.get(ext, 'image/jpeg')
def extract_text_from_image_fireworks(image_path: str) -> str:
"""Fireworks AI Vision API๋ฅผ ์ฌ์ฉํ ์ด๋ฏธ์ง OCR"""
if not FIREWORKS_API_KEY:
return """[๋ฐ๋ชจ ๋ชจ๋] ์ด๋ฏธ์ง OCR ๊ฒฐ๊ณผ:
---
์๋์ฐจ ๊ณ์ฝ์
์ 1์กฐ (๋ชฉ์ ) ์๋์ธ์ ์๋ ๋ถ๋์ฐ์ ์์ฐจ์ธ์๊ฒ ์๋ํ๋ค.
์์ฌ์ง: ์์ธ์ ๊ฐ๋จ๊ตฌ ํ
ํค๋๋ก 123, 5์ธต
์๋๋ฉด์ : 330.58ใก (100ํ)
์๋๊ธฐ๊ฐ: 2024.01.01 ~ 2026.12.31 (3๋
)
๋ณด์ฆ๊ธ: ๊ธ ์ผ์ต์์ (โฉ300,000,000)
์์๋๋ฃ: ๊ธ ์ผ์ฒ์ค๋ฐฑ๋ง์์ (โฉ15,000,000)
๊ด๋ฆฌ๋น: ํ๋น 25,000์ (์ 250๋ง์)
---
โ ๏ธ FIREWORKS_API_KEY ์ค์ ์ ์ค์ OCR ๊ฒฐ๊ณผ ์ ๊ณต"""
try:
base64_image = encode_image_to_base64(image_path)
mime_type = get_image_mime_type(image_path)
headers = {
"Accept": "application/json",
"Content-Type": "application/json",
"Authorization": f"Bearer {FIREWORKS_API_KEY}"
}
payload = {
"model": FIREWORKS_VISION_MODEL,
"max_tokens": 4096,
"temperature": 0.2,
"messages": [
{
"role": "user",
"content": [
{
"type": "text",
"text": """์ด ์ด๋ฏธ์ง๋ ๋ถ๋์ฐ ๊ด๋ จ ๋ฌธ์(๊ณ์ฝ์, ๋๋ฉด, ๊ด๋ฆฌ๋ฌธ์ ๋ฑ)์
๋๋ค.
์ด๋ฏธ์ง์ ์๋ ๋ชจ๋ ํ
์คํธ๋ฅผ ์ ํํ๊ฒ ์ถ์ถํด์ฃผ์ธ์.
ํ๋ ๋ํ๊ฐ ์๋ค๋ฉด ๊ตฌ์กฐ๋ฅผ ์ ์งํ์ฌ ํ
์คํธ๋ก ๋ณํํด์ฃผ์ธ์.
ํ๊ตญ์ด์ ์์ด ๋ชจ๋ ์ ํํ๊ฒ ์ธ์ํด์ฃผ์ธ์.
์ถ์ถํ ํ
์คํธ๋ง ์ถ๋ ฅํ๊ณ , ๋ค๋ฅธ ์ค๋ช
์ ํ์ง ๋ง์ธ์."""
},
{
"type": "image_url",
"image_url": {
"url": f"data:{mime_type};base64,{base64_image}"
}
}
]
}
]
}
response = requests.post(FIREWORKS_VISION_URL, headers=headers, json=payload, timeout=60)
if response.status_code == 200:
result = response.json()
content = result.get("choices", [{}])[0].get("message", {}).get("content", "")
if content:
if "<think>" in content and "</think>" in content:
content = re.sub(r'<think>.*?</think>', '', content, flags=re.DOTALL).strip()
return content
return "โ ๏ธ OCR ๊ฒฐ๊ณผ๊ฐ ๋น์ด์์ต๋๋ค."
else:
return f"โ API ์ค๋ฅ ({response.status_code}): {response.text[:200]}"
except requests.exceptions.Timeout:
return "โ API ์๋ต ์๊ฐ ์ด๊ณผ. ๋ค์ ์๋ํด์ฃผ์ธ์."
except Exception as e:
return f"โ OCR ์ฒ๋ฆฌ ์ค๋ฅ: {str(e)}"
def extract_text_from_pdf(file_path: str) -> str:
if not PDF_AVAILABLE:
return "โ PyMuPDF ์ค์น ํ์: pip install pymupdf"
try:
doc = fitz.open(file_path)
texts = [f"--- ํ์ด์ง {i+1} ---\n{page.get_text()}" for i, page in enumerate(doc) if page.get_text().strip()]
doc.close()
return "\n\n".join(texts) if texts else "โ ๏ธ ํ
์คํธ ์ถ์ถ ์คํจ (์ด๋ฏธ์ง PDF์ผ ์ ์์)"
except Exception as e:
return f"โ PDF ์ค๋ฅ: {e}"
def generate_response_fireworks(message: str, history: list, system_prompt: str) -> Generator:
"""Fireworks AI๋ฅผ ์ฌ์ฉํ ํ
์คํธ ์์ฑ (์คํธ๋ฆฌ๋ฐ)"""
if not FIREWORKS_API_KEY:
demo = f"""## ๐ข PMAI ๋ถ์ ๊ฒฐ๊ณผ
**์ง๋ฌธ**: {message}
---
### ๐ ๋ถ์ ๋ด์ฉ
1. **ํํฉ**: ์
๋ ฅ ๋ด์ฉ ๊ธฐ๋ฐ ์์ฐ๊ด๋ฆฌ ๊ด์ ๋ถ์ ์๋ฃ
2. **ํต์ฌ**: Zero Hardware ์ ๊ทผ, ๋ฌธ์ ๊ธฐ๋ฐ ๋น์ฉ ์ ๊ฐ ํฌ์ธํธ ๋์ถ
3. **๊ถ์ฅ**: ์ด์๋น์ฉ ๊ตฌ์กฐ ์ฌ๊ฒํ , ์๋์ง ํจ์จํ, ๊ณต์ค๋ฅ ๊ด๋ฆฌ ์ ๋ต
> โ ๏ธ ๋ฐ๋ชจ ๋ชจ๋ - FIREWORKS_API_KEY ์ค์ ์ ์์ธ ๋ถ์ ์ ๊ณต"""
for i in range(0, len(demo), 20):
yield demo[:i+20]
return
messages = [{"role": "system", "content": system_prompt}]
for h in history:
if isinstance(h, (list, tuple)) and len(h) >= 2:
messages.extend([{"role": "user", "content": str(h[0])}, {"role": "assistant", "content": str(h[1])}])
messages.append({"role": "user", "content": message})
headers = {"Accept": "application/json", "Content-Type": "application/json", "Authorization": f"Bearer {FIREWORKS_API_KEY}"}
payload = {"model": "accounts/fireworks/models/qwen3-235b-a22b-instruct-2507", "max_tokens": 4096, "temperature": 0.7, "stream": True, "messages": messages}
try:
response = requests.post(FIREWORKS_VISION_URL, headers=headers, json=payload, stream=True, timeout=60)
if response.status_code == 200:
full_response = ""
for line in response.iter_lines():
if line:
line_text = line.decode('utf-8')
if line_text.startswith('data: '):
data_str = line_text[6:]
if data_str.strip() == '[DONE]':
break
try:
data = json.loads(data_str)
content = data.get('choices', [{}])[0].get('delta', {}).get('content', '')
if content:
full_response += content
clean = re.sub(r'<think>.*?</think>', '', full_response, flags=re.DOTALL).strip()
yield clean
except:
continue
else:
yield f"โ API ์ค๋ฅ: {response.status_code}"
except Exception as e:
yield f"โ ์ค๋ฅ: {e}"
def generate_response(message: str, history: list, system_prompt: str = PMAI_SYSTEM_PROMPT) -> Generator:
"""Groq ๋๋ Fireworks๋ฅผ ์ฌ์ฉํ ์๋ต ์์ฑ"""
if groq_client:
messages = [{"role": "system", "content": system_prompt}]
for h in history:
if isinstance(h, (list, tuple)) and len(h) >= 2:
messages.extend([{"role": "user", "content": str(h[0])}, {"role": "assistant", "content": str(h[1])}])
messages.append({"role": "user", "content": message})
try:
completion = groq_client.chat.completions.create(model="llama-3.3-70b-versatile", messages=messages, temperature=0.7, max_tokens=4096, stream=True)
response = ""
for chunk in completion:
if chunk.choices[0].delta.content:
response += chunk.choices[0].delta.content
yield response
return
except:
pass
yield from generate_response_fireworks(message, history, system_prompt)
def chat_respond(message: str, history: list):
if not message or not message.strip():
yield history or []
return
history = history or []
history_api = []
for h in history:
if isinstance(h, dict):
r, c = h.get("role", ""), h.get("content", "")
if r == "user": history_api.append([c, ""])
elif r == "assistant" and history_api: history_api[-1][1] = c
new_history = list(history) + [{"role": "user", "content": message}, {"role": "assistant", "content": ""}]
for chunk in generate_response(message, history_api, PMAI_SYSTEM_PROMPT):
new_history[-1] = {"role": "assistant", "content": chunk}
yield new_history
def run_soma_analysis(document_text: str, selected_agents: List[str]) -> Generator:
if not document_text.strip():
yield "๐ ๋ฌธ์ ๋ด์ฉ์ด ํ์ํฉ๋๋ค."
return
if not selected_agents:
selected_agents = ["document_analyst", "financial_expert", "legal_advisor"]
output = "# ๐ค SOMA ๋ฉํฐ ์์ด์ ํธ ํ์
๋ถ์\n\n---\n\n"
yield output
results = {}
for key in selected_agents:
agent = SOMA_AGENTS.get(key)
if not agent: continue
output += f"## {agent['name']}\n**์ญํ **: {agent['role']}\n\nโณ ๋ถ์ ์ค...\n\n"
yield output
prompt = f"{agent['prompt']}\n\n๋ถ์ํ ๋ฌธ์:\n---\n{document_text[:6000]}\n---\n๊ตฌ์ฒด์ ์ธ ์ธ์ฌ์ดํธ๋ฅผ ์ ๊ณตํ์ธ์."
agent_response = ""
for chunk in generate_response(prompt, [], agent['prompt']):
agent_response = chunk
results[key] = agent_response
output = output.replace("โณ ๋ถ์ ์ค...\n\n", f"{agent_response}\n\n---\n\n")
yield output
if len(results) > 1 and "coordinator" not in selected_agents:
output += f"## {SOMA_AGENTS['coordinator']['name']}\nโณ ์ข
ํฉ ๋ถ์ ์ค...\n\n"
yield output
summary_prompt = f"๊ฐ ์ ๋ฌธ๊ฐ ๋ถ์์ ์ข
ํฉํ์ฌ Executive Summary๋ฅผ ์์ฑํ์ธ์:\n" + "\n".join([f"### {SOMA_AGENTS[k]['name']}:\n{v[:1000]}" for k, v in results.items()])
coord_response = ""
for chunk in generate_response(summary_prompt, [], SOMA_AGENTS['coordinator']['prompt']):
coord_response = chunk
output = output.replace("โณ ์ข
ํฉ ๋ถ์ ์ค...\n\n", f"{coord_response}\n\n")
yield output
yield output + "\nโ
**SOMA ๋ถ์ ์๋ฃ**"
def analyze_document(document_text: str, document_type: str, file_upload: Optional[str] = None) -> Generator:
if file_upload:
ext = file_upload.lower().split('.')[-1]
if ext == 'pdf':
yield "๐ PDF ํ
์คํธ ์ถ์ถ ์ค..."
extracted = extract_text_from_pdf(file_upload)
if extracted.startswith("โ") or extracted.startswith("โ ๏ธ"):
yield extracted
return
document_text = extracted
yield f"๐ PDF ์ถ์ถ ์๋ฃ ({len(extracted):,}์)\n\n๋ถ์ ์ค..."
elif ext in ['jpg', 'jpeg', 'png', 'gif', 'webp', 'bmp']:
yield f"๐ผ๏ธ ์ด๋ฏธ์ง OCR ์ฒ๋ฆฌ ์ค... (Fireworks Vision AI)"
extracted = extract_text_from_image_fireworks(file_upload)
if extracted.startswith("โ"):
yield extracted
return
document_text = extracted
yield f"๐ผ๏ธ OCR ์๋ฃ ({len(extracted):,}์)\n\n**์ถ์ถ๋ ํ
์คํธ:**\n```\n{extracted[:2000]}{'...' if len(extracted) > 2000 else ''}\n```\n\n๋ถ์ ์ค..."
else:
yield f"โ ์ง์ํ์ง ์๋ ํ์: .{ext}"
return
if not document_text or not document_text.strip():
yield "๐ ๋ฌธ์ ๋ด์ฉ์ ์
๋ ฅํ๊ฑฐ๋ ํ์ผ์ ์
๋ก๋ํด์ฃผ์ธ์."
return
prompt = f"""๋ค์ {document_type} ๋ฌธ์๋ฅผ ๋ถ์ํด์ฃผ์ธ์:
---
{document_text[:8000]}
---
๋ค์ ํ์์ผ๋ก ๋ถ์:
## ๐ ๋ฌธ์ ์์ฝ (3์ค)
## ๐ ํต์ฌ ์ ๋ณด
## โ ๏ธ ๋ฆฌ์คํฌ ํฌ์ธํธ
## ๐ก ์ต์ ํ ์ ์
## ๐ ์ก์
์์ดํ
"""
full_output = ""
for chunk in generate_response(prompt, [], DOCUMENT_ANALYSIS_PROMPT):
full_output = chunk
yield full_output
full_output += "\n\n---\n\n## ๐ค SOMA ์ ๋ฌธ๊ฐ ์ถ๊ฐ ์ธ์ฌ์ดํธ\n\n"
yield full_output
mini_agents = [("legal_advisor", "โ๏ธ ๋ฒ๋ฅ ์๋ฌธ๊ฐ"), ("financial_expert", "๐ฐ ์ฌ๋ฌด ์ ๋ฌธ๊ฐ")]
for agent_key, agent_label in mini_agents:
agent = SOMA_AGENTS.get(agent_key)
if not agent: continue
full_output += f"### {agent_label}\n"
yield full_output + "๋ถ์ ์ค...\n"
mini_prompt = f"{agent['prompt']}\n\n๋ฌธ์ ๋ด์ฉ:\n{document_text[:3000]}\n\nํต์ฌ ํฌ์ธํธ 3๊ฐ์ง๋ง ๊ฐ๊ฒฐํ๊ฒ ๋ถ์ํด์ฃผ์ธ์. ๊ฐ ํฌ์ธํธ๋ 1-2๋ฌธ์ฅ์ผ๋ก."
agent_response = ""
for chunk in generate_response(mini_prompt, [], agent['prompt']):
agent_response = chunk
full_output += f"{agent_response}\n\n"
yield full_output
def analyze_cost(building_name, monthly_rent, maintenance, utility, personnel, repair, other, vacancy_rate, additional_info) -> Generator:
total = maintenance + utility + personnel + repair + other
noi = monthly_rent * (1 - vacancy_rate/100) - total
cost_data = f"""๊ฑด๋ฌผ๋ช
: {building_name}
์ ์๋์์
: {monthly_rent:,.0f}์ | ๊ณต์ค๋ฅ : {vacancy_rate}%
์ด์๋น์ฉ ๋ด์ญ:
- ๊ด๋ฆฌ๋น: {maintenance:,.0f}์ ({maintenance/total*100:.1f}%)
- ์ ํธ๋ฆฌํฐ: {utility:,.0f}์ ({utility/total*100:.1f}%)
- ์ธ๊ฑด๋น: {personnel:,.0f}์ ({personnel/total*100:.1f}%)
- ์์ ์ ์ง๋น: {repair:,.0f}์ ({repair/total*100:.1f}%)
- ๊ธฐํ: {other:,.0f}์ ({other/total*100:.1f}%)
- ์ด ์ด์๋น์ฉ: {total:,.0f}์
- ์์ด์์์ต(NOI): {noi:,.0f}์
์ถ๊ฐ์ ๋ณด: {additional_info or '์์'}"""
prompt = f"""๊ฑด๋ฌผ ์ด์๋น์ฉ์ ๋ถ์ํด์ฃผ์ธ์:
{cost_data}
๋ถ์ ํ์:
## ๐ ๋น์ฉ ๊ตฌ์กฐ ๋ถ์
## ๐ด ๋์ ํฌ์ธํธ ์๋ณ
## ๐ฐ ์ ๊ฐ ๊ฐ๋ฅ ํญ๋ชฉ (๊ตฌ์ฒด์ ๊ธ์ก ์ ์)
## ๐ ROI ๊ธฐ๋ฐ ์ฐ์ ์์
## ๐ต ์์ ์ ๊ฐ ํจ๊ณผ (์๊ฐ/์ฐ๊ฐ)"""
full_output = ""
for chunk in generate_response(prompt, [], COST_ANALYSIS_PROMPT):
full_output = chunk
yield full_output
full_output += "\n\n---\n\n## ๐ค SOMA ์ ๋ฌธ๊ฐ ์ถ๊ฐ ์ธ์ฌ์ดํธ\n\n"
yield full_output
mini_agents = [("financial_expert", "๐ฐ ์ฌ๋ฌด ์ ๋ฌธ๊ฐ"), ("facility_manager", "๐ง ์์ค ๊ด๋ฆฌ์")]
for agent_key, agent_label in mini_agents:
agent = SOMA_AGENTS.get(agent_key)
if not agent: continue
full_output += f"### {agent_label}\n"
yield full_output + "๋ถ์ ์ค...\n"
mini_prompt = f"{agent['prompt']}\n\n๋น์ฉ ๋ฐ์ดํฐ:\n{cost_data}\n\n๋น์ ์ ์ ๋ฌธ ๋ถ์ผ ๊ด์ ์์ ํต์ฌ ์ธ์ฌ์ดํธ 3๊ฐ์ง๋ง ๊ฐ๊ฒฐํ๊ฒ ์ ์ํด์ฃผ์ธ์."
agent_response = ""
for chunk in generate_response(mini_prompt, [], agent['prompt']):
agent_response = chunk
full_output += f"{agent_response}\n\n"
yield full_output
def create_seoul_map(selected: str = None) -> str:
if not FOLIUM_AVAILABLE:
return "<div style='padding:40px;text-align:center;'><p>folium ์ค์น ํ์</p></div>"
m = folium.Map(location=[37.5665, 126.9780], zoom_start=11, tiles='cartodbpositron')
for name, info in SEOUL_DISTRICTS.items():
color = 'red' if name == selected else 'blue'
popup = f"<b>{name}</b><br>{info['ํน์ฑ']}<br>์๋๋ฃ: {info['ํ๊ท ์๋๋ฃ']:,}์/ํ"
folium.Marker([info['lat'], info['lng']], popup=popup, tooltip=name, icon=folium.Icon(color=color, icon='building', prefix='fa')).add_to(m)
if name == selected:
folium.Circle([info['lat'], info['lng']], radius=1500, color='red', fill=True, fillOpacity=0.2).add_to(m)
return m._repr_html_()
def analyze_location(district: str) -> Generator:
if district not in SEOUL_DISTRICTS:
yield "์ง์ญ ์ ๋ณด ์์"
return
info = SEOUL_DISTRICTS[district]
location_data = f"""์ง์ญ: ์์ธ์ {district}
ํน์ฑ: {info['ํน์ฑ']}
ํ๊ท ์๋๋ฃ: {info['ํ๊ท ์๋๋ฃ']:,}์/ํ
์์น: ์๋ {info['lat']}, ๊ฒฝ๋ {info['lng']}"""
prompt = f"""์์ธ์ {district}์ ์๊ถ ๋ฐ ๋ถ๋์ฐ ์
์ง๋ฅผ ๋ถ์ํด์ฃผ์ธ์.
{location_data}
๋ค์ ๊ด์ ์์ ์์ธํ ๋ถ์:
## ๐ ์๊ถ ํน์ฑ ๋ถ์
(์ฃผ์ ์
์ข
, ์ ๋์ธ๊ตฌ ํจํด, ์๋น ํน์ฑ)
## ๐ข ์คํผ์ค/์๊ฐ ์์ฅ ํํฉ
(์๋๋ฃ ์์ค, ๊ณต์ค๋ฅ ์ถ์ด, ์์-๊ณต๊ธ)
## ๐ ํฌ์ ๋งค๋ ฅ๋ ํ๊ฐ
(์์ฐ๊ฐ์น ์์น ๊ฐ๋ฅ์ฑ, ๊ฐ๋ฐ ํธ์ฌ)
## โ ๏ธ ๋ฆฌ์คํฌ ์์ธ
## ๐ฏ ์ถ์ฒ ์ ๋ต"""
full_output = ""
for chunk in generate_response(prompt, [], SOMA_AGENTS['market_analyst']['prompt']):
full_output = chunk
yield full_output
full_output += "\n\n---\n\n## ๐ค SOMA ์ ๋ฌธ๊ฐ ์ถ๊ฐ ์ธ์ฌ์ดํธ\n\n"
yield full_output
mini_agents = [("financial_expert", "๐ฐ ์ฌ๋ฌด ์ ๋ฌธ๊ฐ"), ("facility_manager", "๐ง ์์ค ๊ด๋ฆฌ์")]
for agent_key, agent_label in mini_agents:
agent = SOMA_AGENTS.get(agent_key)
if not agent: continue
full_output += f"### {agent_label}\n"
yield full_output + "๋ถ์ ์ค...\n"
mini_prompt = f"{agent['prompt']}\n\n์
์ง ์ ๋ณด:\n{location_data}\n\n์ด ์ง์ญ์์ ์์ฐ๊ด๋ฆฌ ์ ๋น์ ์ ์ ๋ฌธ ๋ถ์ผ ๊ด์ ์์ ํต์ฌ ์กฐ์ธ 3๊ฐ์ง๋ง ๊ฐ๊ฒฐํ๊ฒ ์ ์ํด์ฃผ์ธ์."
agent_response = ""
for chunk in generate_response(mini_prompt, [], agent['prompt']):
agent_response = chunk
full_output += f"{agent_response}\n\n"
yield full_output
def create_cost_chart(m, u, p, r, o):
if not PLOTLY_AVAILABLE: return None
fig = go.Figure(data=[go.Pie(labels=['๊ด๋ฆฌ๋น','์ ํธ๋ฆฌํฐ','์ธ๊ฑด๋น','์์ ๋น','๊ธฐํ'], values=[m,u,p,r,o], hole=0.4, marker_colors=['#3B82F6','#10B981','#F59E0B','#EF4444','#8B5CF6'])])
fig.update_layout(title='์๊ฐ ์ด์๋น์ฉ', height=350, paper_bgcolor='#ffffff', font=dict(color='#1e293b'))
return fig
CSS = """
@import url('https://fonts.googleapis.com/css2?family=Noto+Sans+KR:wght@400;500;700&display=swap');
.gradio-container { background: linear-gradient(135deg, #f8fafc 0%, #e2e8f0 50%, #f1f5f9 100%) !important; font-family: 'Noto Sans KR', sans-serif !important; min-height: 100vh; }
.gr-button-primary { background: linear-gradient(135deg, #2563eb, #3b82f6) !important; border: none !important; color: white !important; font-weight: 600 !important; box-shadow: 0 4px 14px rgba(37,99,235,0.35) !important; }
.gr-button-primary:hover { background: linear-gradient(135deg, #1d4ed8, #2563eb) !important; transform: translateY(-1px) !important; box-shadow: 0 6px 20px rgba(37,99,235,0.4) !important; }
.gr-panel, .block { background: #ffffff !important; border: 1px solid #e2e8f0 !important; border-radius: 16px !important; box-shadow: 0 4px 20px rgba(0,0,0,0.08) !important; }
textarea, input[type="text"], input[type="number"] { background: #ffffff !important; border: 2px solid #e2e8f0 !important; color: #1e293b !important; border-radius: 10px !important; }
textarea:focus, input:focus { border-color: #3b82f6 !important; box-shadow: 0 0 0 3px rgba(59,130,246,0.15) !important; }
label, .gr-input-label { color: #334155 !important; font-weight: 500 !important; }
.gr-markdown { color: #334155 !important; }
.gr-markdown h1, .gr-markdown h2, .gr-markdown h3 { color: #1e40af !important; font-weight: 700 !important; }
.gr-markdown h1 { font-size: 1.8em !important; }
.gr-markdown h2 { font-size: 1.4em !important; }
.gr-markdown h3 { font-size: 1.2em !important; }
.gr-chatbot { background: #ffffff !important; border: 1px solid #e2e8f0 !important; border-radius: 16px !important; }
.gr-tab-nav { background: #f1f5f9 !important; border-radius: 12px !important; padding: 4px !important; }
.gr-tab-nav button { color: #64748b !important; font-weight: 500 !important; border-radius: 8px !important; }
.gr-tab-nav button.selected { background: #ffffff !important; color: #2563eb !important; box-shadow: 0 2px 8px rgba(0,0,0,0.1) !important; }
.gr-accordion { background: #f8fafc !important; border: 1px solid #e2e8f0 !important; border-radius: 12px !important; }
.gr-dropdown { background: #ffffff !important; border: 2px solid #e2e8f0 !important; border-radius: 10px !important; }
.gr-checkbox-group { background: #f8fafc !important; border-radius: 10px !important; padding: 12px !important; }
::-webkit-scrollbar { width: 8px; height: 8px; }
::-webkit-scrollbar-track { background: #f1f5f9; border-radius: 4px; }
::-webkit-scrollbar-thumb { background: #94a3b8; border-radius: 4px; }
::-webkit-scrollbar-thumb:hover { background: #64748b; }
footer { display: none !important; }
"""
def create_demo():
with gr.Blocks(title="TenAI PMAI Pro", css=CSS) as demo:
gr.HTML("""<div style="text-align:center;padding:35px 20px;background:linear-gradient(135deg,#ffffff,#f8fafc);border-radius:20px;border:1px solid #e2e8f0;margin-bottom:25px;box-shadow:0 8px 30px rgba(0,0,0,0.08);">
<h1 style="color:#1e40af;font-size:2.6em;margin:0;font-weight:800;">๐ข TenAI PMAI Pro</h1>
<p style="color:#475569;margin:12px 0 5px 0;font-size:1.15em;font-weight:500;">๊ธฐ์
์์ฐ๊ด๋ฆฌ AI ํ๋ซํผ</p>
<p style="color:#64748b;margin:5px 0 20px 0;font-size:0.95em;">"ํ๋์จ์ด ์๋ ๊ฑด๋ฌผ ์ด์์ฒด์ (OS), TenAI"</p>
<div style="display:flex;justify-content:center;gap:12px;flex-wrap:wrap;">
<span style="background:linear-gradient(135deg,#2563eb,#3b82f6);padding:10px 22px;border-radius:25px;color:white;font-size:0.9em;font-weight:600;box-shadow:0 4px 12px rgba(37,99,235,0.3);">โก Zero Hardware</span>
<span style="background:linear-gradient(135deg,#059669,#10b981);padding:10px 22px;border-radius:25px;color:white;font-size:0.9em;font-weight:600;box-shadow:0 4px 12px rgba(16,185,129,0.3);">๐ผ๏ธ Fireworks Vision AI</span>
<span style="background:linear-gradient(135deg,#7c3aed,#8b5cf6);padding:10px 22px;border-radius:25px;color:white;font-size:0.9em;font-weight:600;box-shadow:0 4px 12px rgba(139,92,246,0.3);">๐ค SOMA Multi-Agent</span>
</div></div>""")
with gr.Tabs():
with gr.Tab("๐ฌ PMAI ์๋ด"):
gr.Markdown("### ๐ค 24์๊ฐ AI ์์ฐ๊ด๋ฆฌ ๋น์")
chatbot = gr.Chatbot(height=400)
with gr.Row():
msg = gr.Textbox(placeholder="์ง๋ฌธ์ ์
๋ ฅํ์ธ์...", show_label=False, scale=9)
btn = gr.Button("์ ์ก", variant="primary", scale=1)
gr.Examples(["๊ณต์ค๋ฅ ์ ๋ฎ์ถ๋ ๋ฐฉ๋ฒ์?", "๊ฑด๋ฌผ ์ ์ง๋ณด์ ๋น์ฉ ์ ๊ฐ ์ ๋ต", "์๋์ฐจ ๊ณ์ฝ ๊ฐฑ์ ์ ์ฃผ์์ "], inputs=msg)
msg.submit(chat_respond, [msg, chatbot], [chatbot]).then(lambda: "", None, [msg])
btn.click(chat_respond, [msg, chatbot], [chatbot]).then(lambda: "", None, [msg])
with gr.Tab("๐ ๋ฌธ์ ๋ถ์"):
gr.Markdown("### ๐ Vision AI ๋ฌธ์ ๋ถ์\n**PDF** ๋๋ **์ด๋ฏธ์ง**(๊ณ์ฝ์ ์ฌ์ง, ์ค์บ๋ณธ) ์
๋ก๋ ์ ์๋ OCR ์ฒ๋ฆฌ")
with gr.Row():
with gr.Column(scale=1):
doc_type = gr.Dropdown(["์๋์ฐจ๊ณ์ฝ์", "์ ์ง๋ณด์ ๋ฌธ์", "์์ค์ ๊ฒ ๋ณด๊ณ ์", "๊ด๋ฆฌ๋น ๋ด์ญ์", "๊ฑด๋ฌผ ๋๋ฉด", "๊ธฐํ"], value="์๋์ฐจ๊ณ์ฝ์", label="๋ฌธ์ ์ ํ")
file_upload = gr.File(label="๐ ํ์ผ ์
๋ก๋ (PDF/์ด๋ฏธ์ง)", file_types=[".pdf",".jpg",".jpeg",".png",".gif",".webp"], type="filepath")
gr.Markdown("<small>โ
์ง์: PDF, JPG, PNG, GIF, WEBP | ๐ผ๏ธ ์ด๋ฏธ์ง๋ Fireworks Vision AI๋ก OCR</small>")
doc_input = gr.Textbox(lines=6, placeholder="๋๋ ํ
์คํธ ์ง์ ์
๋ ฅ...", label="๋ฌธ์ ๋ด์ฉ")
analyze_btn = gr.Button("๐ ๋ถ์ ์์", variant="primary", size="lg")
with gr.Column(scale=1):
analysis_output = gr.Markdown("### ๐ ๋ถ์ ๊ฒฐ๊ณผ\nํ์ผ ์
๋ก๋ ๋๋ ํ
์คํธ ์
๋ ฅ ํ ๋ถ์ ์์")
analyze_btn.click(analyze_document, [doc_input, doc_type, file_upload], analysis_output)
with gr.Tab("๐ฐ ๋น์ฉ ๋ถ์"):
gr.Markdown("### ๐ ์ด์๋น์ฉ ์ต์ ํ ๋ถ์")
with gr.Row():
with gr.Column():
building = gr.Textbox(label="๊ฑด๋ฌผ๋ช
", value="๊ฐ๋จํ
ํฌํ์")
rent = gr.Number(label="์ ์๋์์
(์)", value=50000000)
vacancy = gr.Slider(0, 100, 10, label="๊ณต์ค๋ฅ (%)")
gr.Markdown("#### ์๊ฐ ์ด์๋น์ฉ")
m_cost = gr.Number(label="๊ด๋ฆฌ๋น", value=5000000)
u_cost = gr.Number(label="์ ํธ๋ฆฌํฐ", value=8000000)
p_cost = gr.Number(label="์ธ๊ฑด๋น", value=12000000)
r_cost = gr.Number(label="์์ ์ ์ง๋น", value=3000000)
o_cost = gr.Number(label="๊ธฐํ", value=2000000)
add_info = gr.Textbox(label="์ถ๊ฐ ์ ๋ณด", lines=2)
cost_btn = gr.Button("๐ก ๋น์ฉ ๋ถ์", variant="primary")
with gr.Column():
cost_chart = gr.Plot()
cost_output = gr.Markdown("### ๐ ๋ถ์ ๊ฒฐ๊ณผ")
cost_btn.click(lambda m,u,p,r,o: create_cost_chart(m,u,p,r,o), [m_cost,u_cost,p_cost,r_cost,o_cost], cost_chart)
cost_btn.click(analyze_cost, [building,rent,m_cost,u_cost,p_cost,r_cost,o_cost,vacancy,add_info], cost_output)
with gr.Tab("๐บ๏ธ ์
์ง ๋ถ์"):
gr.Markdown("### ๐ ์์ธ์ ์๊ถ ๋ถ์")
with gr.Row():
with gr.Column():
district = gr.Dropdown(list(SEOUL_DISTRICTS.keys()), value="๊ฐ๋จ๊ตฌ", label="์ง์ญ ์ ํ")
loc_btn = gr.Button("๐ ์
์ง ๋ถ์", variant="primary")
with gr.Column():
map_html = gr.HTML(value=create_seoul_map())
loc_output = gr.Markdown("### ๋ถ์ ๊ฒฐ๊ณผ")
district.change(create_seoul_map, [district], [map_html])
loc_btn.click(analyze_location, [district], loc_output)
with gr.Tab("๐ค SOMA ํ์
"):
gr.Markdown("### ๐ค ๋ฉํฐ ์์ด์ ํธ ํ์
๋ถ์\n6๋ช
์ AI ์ ๋ฌธ๊ฐ๊ฐ ๋ฌธ์๋ฅผ ๋ค๊ฐ๋๋ก ๋ถ์")
with gr.Row():
with gr.Column():
soma_agents = gr.CheckboxGroup([("๐ ๋ฌธ์๋ถ์๊ฐ","document_analyst"),("๐ฐ ์ฌ๋ฌด์ ๋ฌธ๊ฐ","financial_expert"),("โ๏ธ ๋ฒ๋ฅ ์๋ฌธ๊ฐ","legal_advisor"),("๐ง ์์ค๊ด๋ฆฌ์","facility_manager"),("๐ ์๊ถ๋ถ์๊ฐ","market_analyst")], value=["document_analyst","financial_expert","legal_advisor"], label="๋ถ์ ํ")
soma_file = gr.File(label="ํ์ผ ์
๋ก๋", file_types=[".pdf",".jpg",".jpeg",".png"], type="filepath")
soma_text = gr.Textbox(lines=5, placeholder="๋๋ ํ
์คํธ ์
๋ ฅ...", label="๋ฌธ์ ๋ด์ฉ")
soma_btn = gr.Button("๐ SOMA ๋ถ์", variant="primary")
with gr.Column():
soma_output = gr.Markdown("### SOMA ๋ถ์ ๊ฒฐ๊ณผ")
def soma_with_file(text, agents, file):
doc = text or ""
if file:
ext = file.lower().split('.')[-1]
if ext == 'pdf':
doc = extract_text_from_pdf(file)
elif ext in ['jpg','jpeg','png','gif','webp']:
doc = extract_text_from_image_fireworks(file)
if doc.startswith("โ"):
yield doc
return
if not doc.strip():
yield "๋ฌธ์๋ฅผ ์
๋ ฅํด์ฃผ์ธ์."
return
yield from run_soma_analysis(doc, agents)
soma_btn.click(soma_with_file, [soma_text, soma_agents, soma_file], soma_output)
with gr.Tab("โน๏ธ About"):
gr.Markdown("""
## ๐ข TenAI PMAI Pro
### ๋น์ : "ํ๋์จ์ด ์๋ ๊ฑด๋ฌผ ์ด์์ฒด์ (OS)"
### ํต์ฌ ๊ธฐ๋ฅ
| ๊ธฐ๋ฅ | ์ค๋ช
|
|-----|-----|
| ๐ผ๏ธ **Fireworks Vision AI** | ์ด๋ฏธ์ง OCR (Qwen3-VL-235B) |
| ๐ **๋ฌธ์ ๋ถ์** | PDF/์ด๋ฏธ์ง ์๋ ํ
์คํธ ์ถ์ถ ๋ฐ ๋ถ์ |
| ๐ค **SOMA ๋ฉํฐ์์ด์ ํธ** | 6๋ช
AI ์ ๋ฌธ๊ฐ ํ์
|
| ๐บ๏ธ **์๊ถ ๋ถ์** | ์์ธ์ 12๊ฐ ๊ตฌ ์
์ง ๋ถ์ |
### API ์ค์
```
FIREWORKS_API_KEY=your_key # Vision AI + LLM
GROQ_API_KEY=your_key # ๋น ๋ฅธ LLM (์ ํ)
```
### Contact
๐ง ten@tenspace.co.kr | ๐ฑ 010-2710-6246
""")
gr.HTML("<div style='text-align:center;padding:20px;margin-top:10px;border-top:1px solid #e2e8f0;background:#f8fafc;border-radius:0 0 16px 16px;'><p style='color:#64748b;font-size:0.9em;margin:0;'>๐ Powered by <strong style='color:#2563eb;'>Ten-AX Engine</strong> | Fireworks Vision AI | SOMA Multi-Agent</p></div>")
return demo
if __name__ == "__main__":
demo = create_demo()
demo.launch(server_name="0.0.0.0", server_port=7860) |