Spaces:
Running
Running
File size: 43,685 Bytes
8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 87ef853 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 87ef853 8ed5fed 8c0cc35 8ed5fed 8c0cc35 8ed5fed 8c0cc35 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 | # ============================================================
# 한지(HANJI) · HWP AI Agent 서비스 — App Router
# core.so (또는 core.py)에서 엔진을 import
# ============================================================
import os, re, json, time, tempfile, threading
import gradio as gr
# ── core 엔진 import (.so 또는 .py) ──
import core
from core import *
def build_ui():
with gr.Blocks(title="한지(HANJI) · HWP AI Agent 서비스") as app:
gr.HTML(f"<style>{SOMA_CUSTOM_CSS}</style>")
# ── Top Bar ──
gr.HTML("""
<div class="soma-topbar">
<span class="soma-logo">한지<em>(HANJI)</em></span>
<span class="soma-sep"></span>
<span class="soma-desc">HWP AI Agent 서비스</span>
<a class="soma-url" href="https://hanji.ginigen.ai" target="_blank">🔗 hanji.ginigen.ai</a>
<span class="soma-right">
<a class="soma-contact" href="mailto:ginigenaihp@gmail.com">📧 문의 · 온프레미스 · 제휴</a>
</span>
</div>""")
# ── States ──
ref_text_state = gr.State("")
ref_hwpx_path_state = gr.State("")
state = gr.State({"final_doc": "", "search_count": 0})
dummy_state = gr.State("")
doc_text_state = gr.State("")
_transform_result_path = "" # 문서 변환 결과 경로 (run_soma에서 설정)
# ══════════════════════════════════════════════════
# MAIN LAYOUT: Left 1/3 Controls | Right 2/3 Viewer
# ══════════════════════════════════════════════════
with gr.Row(equal_height=False):
# ── LEFT PANEL (1/3) ──────────────────────────
with gr.Column(scale=1, min_width=320):
# Prompt
prompt_input = gr.Textbox(
label="📌 프롬프트",
placeholder="예: 2026년 AI 보안 유망기업 육성 지원사업 공모 안내문을 작성해주세요.",
lines=3)
# File upload
ref_file_upload = gr.File(
label="📎 레퍼런스 문서",
file_types=[".hwp",".hwpx",".hml",".pdf",".docx",".txt",".md",
".csv",".json",".xml",".xlsx",".xls",".py",".html",".log"])
ref_upload_status = gr.Textbox(label="파일 상태", interactive=False, lines=2,
placeholder="레퍼런스 파일을 업로드하면 여기에 상태가 표시됩니다.")
# Generation Mode
mode_radio = gr.Radio(
choices=[
"새로 생성 — AI가 주제에 맞는 문서를 처음부터 작성",
"서식 유지 · 내용 변경 — 원본 레이아웃 100% 보존, 텍스트만 교체",
"구조 참고 · 새로 생성 — 원본 구조를 참고하여 새 내용으로 작성",
],
value="새로 생성 — AI가 주제에 맞는 문서를 처음부터 작성",
label="⚙️ 생성 모드",
interactive=True)
mode_state = gr.State(1) # 1=새로, 2=서식유지, 3=구조참고
# Settings (compact)
with gr.Row():
max_search_slider = gr.Slider(minimum=5, maximum=100, value=20, step=5,
label="🔍 검색", scale=1)
temperature_slider = gr.Slider(minimum=0.1, maximum=1.0, value=0.6, step=0.05,
label="🌡 Temp", scale=1)
# Action buttons
with gr.Row():
run_btn = gr.Button("🚀 문서 생성", variant="primary", scale=2)
stop_btn = gr.Button("⛔", variant="secondary", scale=0)
# Status indicator
search_counter = gr.Markdown("대기 중")
# HWPX Download
with gr.Row():
gen_hml_btn = gr.Button("📥 HWPX 변환", variant="primary", scale=2)
copy_text_btn = gr.Button("📋", variant="secondary", scale=0)
hml_status = gr.Textbox(label="", interactive=False, value="",
placeholder="HWPX 변환 상태", lines=1)
hml_file = gr.File(label="다운로드", file_types=[".hwpx"], visible=True)
# Generated text (collapsed)
with gr.Accordion("📝 생성된 텍스트", open=False):
final_doc_box = gr.Textbox(label="", value="", interactive=True, lines=12,
placeholder="SOMA 파이프라인 실행 후 최종 문서 텍스트")
# Pipeline internals (collapsed)
with gr.Accordion("🧬 파이프라인 로그", open=False):
agent_stream = gr.Textbox(label="Agent Stream", value="", interactive=False, lines=6)
search_log_box = gr.Textbox(label="Search Log", value="", interactive=False, lines=4)
agent_log_box = gr.Textbox(label="Pipeline Log", value="", interactive=False, lines=6)
# Doc Chat (collapsed)
with gr.Accordion("📎 문서 분석 챗", open=False):
doc_upload = gr.File(label="📄 문서 업로드",
file_types=[".hwp",".hwpx",".hml",".pdf",".docx",".txt",".md",
".csv",".json",".xml",".xlsx",".xls",".py",".html",".log"])
doc_upload_status = gr.Textbox(label="", interactive=False, lines=1)
doc_chatbot = gr.Chatbot(label="💬 Chat", height=200)
with gr.Row():
doc_msg = gr.Textbox(label="", placeholder="질문하세요...", lines=1, scale=4)
doc_send_btn = gr.Button("🚀", variant="primary", scale=0)
doc_clear_btn = gr.Button("🗑️ Clear", size="sm")
# ── 문서 변환 (XML 직접 치환) ──
with gr.Accordion("🔄 문서 변환 (서식 100% 보존)", open=False):
gr.HTML('<div style="font-size:11px;color:#475569;padding:4px 0;border-bottom:1px solid #e2e8f0">'
'원본 HWPX의 XML 구조를 보존하면서 LLM이 맥락을 이해하여 텍스트만 교체합니다.'
'</div>')
transform_file = gr.File(label="📂 원본 HWPX 업로드", file_types=[".hwpx"])
transform_instruction = gr.Textbox(
label="📝 변환 지시",
placeholder="예: 경기도→서울, 노인말벗서비스→청년창업지원벗서비스로 변경하되 맥락에 맞게 조정",
lines=3,
)
transform_temp = gr.Slider(0.0, 1.0, 0.3, step=0.1, label="Temperature (낮을수록 정확)")
transform_btn = gr.Button("🔄 변환 실행", variant="primary", size="lg")
transform_status = gr.Textbox(label="상태", interactive=False)
transform_diff = gr.HTML(label="변경 사항")
transform_output = gr.File(label="📥 변환된 HWPX 다운로드")
# ── RIGHT PANEL (2/3) — DOCUMENT VIEWER ──────
with gr.Column(scale=2, min_width=500, elem_classes=["viewer-panel"]):
viewer_main = gr.HTML(value=_SAMPLE_PREVIEW)
# ── Hidden component for ohaeng (pipeline needs it) ──
ohaeng_display = gr.HTML(value="", visible=False)
# ── Event Handlers
def handle_ref_upload(file):
if file is None:
return "", "", "", _viewer_empty("파일을 선택하면 여기에 미리보기가 표시됩니다.")
fpath = file.name if hasattr(file, 'name') else str(file)
fname = os.path.basename(fpath)
ext = Path(fpath).suffix.lower()
# ── 바이너리 HWP 감지 → 변환 안내 ──
if ext == '.hwp' and _is_binary_hwp(fpath):
text, err = process_uploaded_file(fpath)
preview = hwpx_to_html_preview(fpath)
if text:
status = f"📄 {fname} ({len(text):,}자 추출)\n\n{_HWP_CONVERT_GUIDE}"
return text, status, "", preview
return "", f"❌ {fname}: {err}", "", preview
# ── HWPX 파일 → 스타일 복원 모드 ──
hwpx_path = ""
if ext == '.hwpx':
try:
with zipfile.ZipFile(fpath, 'r') as zf:
if 'Contents/header.xml' in zf.namelist():
hwpx_path = fpath
styles = analyze_hwpx_styles(fpath)
print(f"📋 레퍼런스 HWPX 분석 완료: charPr {styles['char_count']}개, "
f"paraPr {styles['para_count']}개, "
f"borderFill {styles['bf_count']}개")
except:
pass
# ── 그 외 파일 ──
text, err = process_uploaded_file(fpath)
# 뷰어: HWP/HWPX만 렌더링
if ext in ('.hwp', '.hwpx'):
preview = hwpx_to_html_preview(fpath)
else:
preview = _viewer_empty(f"{fname} — HWP/HWPX 파일만 미리보기 지원됩니다.")
if text:
status = f"✅ {fname} ({len(text):,}자)"
if hwpx_path:
status += "\n🔄 '서식 유지 · 내용 변경' 및 '구조 참고 · 새로 생성' 모드 사용 가능"
return text, status, hwpx_path, preview
return "", f"❌ {fname}: {err}", "", preview
ref_file_upload.change(fn=handle_ref_upload, inputs=[ref_file_upload],
outputs=[ref_text_state, ref_upload_status, ref_hwpx_path_state, viewer_main])
def _radio_to_mode(radio_val):
"""라디오 레이블 → 모드 번호 변환"""
if not radio_val:
return 1
if "서식 유지" in radio_val:
return 2
if "구조 참고" in radio_val:
return 3
return 1
def run_soma(prompt, max_search, temperature, ref_text, ref_hwpx_path="", mode_val=1):
mode = mode_val if isinstance(mode_val, int) else _radio_to_mode(str(mode_val))
if not prompt.strip():
yield (ohaeng_cards_html("水"), "⚠️ 프롬프트를 입력하세요.", "", "", "", "대기 중", "")
return
# ════════════════════════════════════════════════════════
# MODE 2: 서식 유지 · 내용 변경 (XML 직접 치환)
# ════════════════════════════════════════════════════════
if mode == 2 and ref_hwpx_path and os.path.exists(ref_hwpx_path):
# ── XML 직접 치환 모드: SOMA 전체 바이패스 ──
yield (ohaeng_cards_html("水"),
"🔄 **서식 유지 · 내용 변경** 모드 — XML 직접 치환 (서식 100% 보존)\n\n"
"📖 원본 텍스트 노드 추출 중...\n",
"", "🔄 Mode 2: XML 키워드 치환\n", "", "🔄 변환 중", "")
try:
text_list, raw_xml, orig_flags = extract_text_nodes(ref_hwpx_path)
yield (ohaeng_cards_html("木"),
f"📖 텍스트 노드 {len(text_list)}개 추출 완료\n\n"
f"🤖 LLM 키워드 매핑 생성 중...\n",
"", f"📖 {len(text_list)}개 노드 추출\n", "", "🔄 LLM 분석 중", "")
mapping = generate_keyword_mapping(raw_xml, prompt, temperature)
if not mapping:
yield (ohaeng_cards_html("金"),
"⚠️ 변경할 키워드가 없습니다. 지시를 더 구체적으로 입력하세요.",
"", "❌ 매핑 0건\n", "", "⚠️ 변경 없음", "")
return
yield (ohaeng_cards_html("火"),
f"🤖 키워드 매핑 {len(mapping)}쌍 생성\n\n"
f"🔧 XML 적용 중...\n",
"", f"🤖 {len(mapping)}쌍 매핑\n", "", "🔧 적용 중", "")
new_xml, details = apply_keyword_mapping(raw_xml, mapping)
output_path = repack_transform_hwpx(ref_hwpx_path, new_xml, orig_flags)
orig_name = os.path.splitext(os.path.basename(ref_hwpx_path))[0]
final_name = f"{orig_name}_변환.hwpx"
final_path = os.path.join(os.path.dirname(output_path), final_name)
os.rename(output_path, final_path)
total_count = sum(d.get("count",0) for d in details)
summary_lines = []
for d in details:
summary_lines.append(f"• '{d['original']}' → '{d['replacement']}' ({d.get('count',0)}회)")
summary = "\n".join(summary_lines)
final_doc = (
f"## 🔄 문서 변환 완료 (서식 100% 보존)\n\n"
f"**{len(details)}개 키워드 · {total_count}회 치환**\n\n"
f"{summary}\n\n"
f"---\n"
f"✅ header.xml: 원본 그대로\n"
f"✅ 이미지/스크립트: 원본 그대로\n"
f"✅ charPr/paraPr: 원본 그대로\n"
f"✅ 문단 구조: 원본 그대로\n"
f"✅ section0.xml: 키워드만 {total_count}회 치환\n"
)
preview = hwpx_to_html_preview(final_path) if 'hwpx_to_html_preview' in dir() else ""
yield (ohaeng_cards_html("金"),
f"🎉 **문서 변환 완료!**\n\n"
f"서식 100% 보존 · {len(details)}개 키워드 · {total_count}회 치환\n\n"
f"아래 'HWPX 생성' 버튼으로 다운로드하세요.\n",
"", f"✅ 변환 완료: {total_count}회\n", final_doc,
f"✅ 변환 완료", final_doc)
nonlocal _transform_result_path
_transform_result_path = final_path
return
except Exception as e:
import traceback
traceback.print_exc()
yield (ohaeng_cards_html("金"),
f"❌ 변환 오류: {e}\n\n모드를 '새로 생성'으로 변경하여 다시 시도하세요.",
"", f"❌ {e}\n", "", "❌ 오류", "")
return
# ════════════════════════════════════════════════════════
# MODE 1 & 3: SOMA 파이프라인 (문서 신규 생성)
# ════════════════════════════════════════════════════════
full_prompt = prompt
if mode == 3 and ref_text and ref_text.strip():
# MODE 3: 참조 문서의 구조 골격을 압축 추출하여 주입
structure = extract_structure_summary(ref_text)
full_prompt = f"{prompt}\n\n{structure}"
elif ref_text and ref_text.strip():
# MODE 1: 레퍼런스 텍스트가 있으면 참고자료로만 활용
ref_content = ref_text.strip()[:8000]
full_prompt = f"{prompt}\n\n[참고자료]\n{ref_content}"
stream_acc, log_acc, search_log, final_doc = "", "", "", ""
active_agent, sc = "水", 0
for chunk in soma_pipeline(full_prompt, int(max_search), temperature):
if chunk.get("done"):
final_doc = chunk.get("final_doc", "")
sc = chunk.get("search_count", sc)
log_acc = chunk.get("log", "")
search_log = chunk.get("search_log", "")
break
active_agent = chunk.get("active", active_agent)
tok = chunk.get("stream", "")
if tok:
stream_acc += tok
if len(stream_acc) > 8000:
stream_acc = "...(이전 생략)...\n" + stream_acc[-6000:]
if chunk.get("log"): log_acc = chunk["log"]
if chunk.get("search_log"): search_log = chunk["search_log"]
if chunk.get("search_count") is not None: sc = chunk["search_count"]
if chunk.get("final_doc"): final_doc = chunk["final_doc"]
yield (ohaeng_cards_html(active_agent), stream_acc, search_log, log_acc,
final_doc if final_doc else "", f"🔍 {sc} / {int(max_search)}", "")
yield (ohaeng_cards_html("金"), stream_acc + "\n\n🎉 완료!", search_log, log_acc,
final_doc, f"✅ 완료: {sc}회 검색", final_doc)
run_btn.click(
fn=run_soma,
inputs=[prompt_input, max_search_slider, temperature_slider, ref_text_state, ref_hwpx_path_state, mode_radio],
outputs=[ohaeng_display, agent_stream, search_log_box, agent_log_box, final_doc_box, search_counter, dummy_state])
def make_hml(doc_text, ref_hwpx_path, mode_val=1):
mode = mode_val if isinstance(mode_val, int) else _radio_to_mode(str(mode_val))
nonlocal _transform_result_path
# ── MODE 2: 문서 변환 결과가 있으면 바로 반환 ──
if mode == 2 and _transform_result_path and os.path.exists(_transform_result_path):
path = _transform_result_path
_transform_result_path = "" # 1회 사용 후 리셋
preview = hwpx_to_html_preview(path)
return path, "✅ 문서 변환 완료 (서식 100% 보존) — XML 직접 치환", preview
if not doc_text or not doc_text.strip():
return None, "❌ 문서를 먼저 생성하세요.", _viewer_empty("HWPX 생성 후 여기에 표시됩니다.")
try:
# MODE 3: 레퍼런스 HWPX 구조 참고 → SectionCloner
if mode == 3 and ref_hwpx_path and os.path.exists(ref_hwpx_path):
path = generate_hwpx(doc_text.strip(), ref_hwpx_path=ref_hwpx_path)
gen_mode = "🧩 구조 참고 · SectionCloner"
elif ref_hwpx_path and os.path.exists(ref_hwpx_path):
path = generate_hwpx(doc_text.strip(), ref_hwpx_path=ref_hwpx_path)
gen_mode = "🎯 레퍼런스 스타일 복원"
else:
path = generate_hwpx(doc_text.strip())
gen_mode = "📄 report 템플릿"
title = normalize_text_for_title(doc_text.strip())
safe_title = re.sub(r'[\\/:*?"<>|]', '', title)[:40].strip() or "문서"
new_path = os.path.join(os.path.dirname(path), f"{safe_title}.hwpx")
os.rename(path, new_path)
# page_guard 결과 표시
status = f"✅ 생성 완료 ({gen_mode})"
if ref_hwpx_path and os.path.exists(ref_hwpx_path):
guard = page_guard_check(ref_hwpx_path, new_path)
if guard["status"] == "PASS":
status += f" | 📏 page_guard PASS (ref={guard['ref_chars']}자 → out={guard['out_chars']}자)"
else:
status += f" | ⚠️ page_guard {len(guard['errors'])}건 경고"
# 생성된 HWPX 뷰어 렌더링
preview = hwpx_to_html_preview(new_path)
return new_path, status, preview
except Exception as e:
return None, f"❌ 오류: {e}", _viewer_empty(f"생성 오류: {e}")
gen_hml_btn.click(fn=make_hml, inputs=[final_doc_box, ref_hwpx_path_state, mode_radio],
outputs=[hml_file, hml_status, viewer_main])
# Doc Chat handlers
def handle_doc_upload(file):
if file is None:
return "", "파일을 선택해주세요."
fpath = file.name if hasattr(file, 'name') else str(file)
fname = os.path.basename(fpath)
ext = Path(fpath).suffix.lower()
# 바이너리 HWP 감지
is_bin_hwp = (ext == '.hwp' and _is_binary_hwp(fpath))
text, err = process_uploaded_file(fpath)
if text:
status = f"✅ {fname} ({len(text):,}자)"
if is_bin_hwp:
status = f"📄 {fname} ({len(text):,}자 추출) — 바이너리 HWP (텍스트만 추출됨)"
return text, status
return "", f"❌ {fname}: {err}"
doc_upload.change(fn=handle_doc_upload, inputs=[doc_upload], outputs=[doc_text_state, doc_upload_status])
doc_send_btn.click(fn=doc_chat_respond, inputs=[doc_msg, doc_chatbot, doc_text_state], outputs=[doc_chatbot])
doc_msg.submit(fn=doc_chat_respond, inputs=[doc_msg, doc_chatbot, doc_text_state], outputs=[doc_chatbot])
doc_clear_btn.click(fn=lambda: ([], ""), outputs=[doc_chatbot, doc_text_state])
# ── 문서 변환 이벤트 핸들러 ──
def handle_transform(hwpx_file, instruction, temperature):
if hwpx_file is None:
return None, "❌ HWPX 파일을 업로드하세요.", ""
if not instruction or not instruction.strip():
return None, "❌ 변환 지시를 입력하세요.", ""
fpath = hwpx_file.name if hasattr(hwpx_file, 'name') else str(hwpx_file)
try:
output_path, replacements, diff_html = transform_hwpx(
fpath, instruction.strip(), temperature)
orig_name = os.path.splitext(os.path.basename(fpath))[0]
new_name = f"{orig_name}_변환.hwpx"
final_path = os.path.join(os.path.dirname(output_path), new_name)
os.rename(output_path, final_path)
return final_path, f"✅ 변환 완료: {len(replacements)}건 변경 | 서식 100% 보존", diff_html
except Exception as e:
return None, f"❌ 오류: {e}", f"<p style='color:red'>{e}</p>"
transform_btn.click(
fn=handle_transform,
inputs=[transform_file, transform_instruction, transform_temp],
outputs=[transform_output, transform_status, transform_diff])
return app
# ============================================================
# ⑧ Entry Point — FastAPI 메인 + Gradio 서브마운트
# ============================================================
from fastapi import FastAPI, Request as _FAReq
from fastapi.responses import FileResponse, JSONResponse, HTMLResponse, StreamingResponse
import uvicorn
# ── FastAPI 메인 앱 ──
app = FastAPI()
_APP_DIR = os.path.dirname(os.path.abspath(__file__))
_index_path = os.path.join(_APP_DIR, "index.html")
# ── ohah/hwpjs 백그라운드 설치 ──
threading.Thread(target=_install_hwpjs, daemon=True).start()
# ── "/" → index.html 서빙 ──
@app.get("/")
async def _serve_index():
if os.path.exists(_index_path):
return FileResponse(_index_path, media_type="text/html")
return HTMLResponse("<h1>index.html not found</h1>", status_code=404)
@app.get("/ui")
async def _serve_ui():
return await _serve_index()
# ── SOMA API ──
import asyncio as _asyncio
import queue as _queue
_file_registry = {}
_doc_text_store = {} # sid → text
_doc_hwpx_store = {} # sid → hwpx file path (변환 모드용)
_transform_store = {} # sid → 변환 결과 hwpx path
_last_transform = {"path": "", "ts": 0} # 마지막 변환 결과 (index.html용)
@app.post("/soma/run")
async def _soma_run(req: _FAReq):
try:
body = await req.json()
prompt = body.get("prompt", "").strip()
max_search = int(body.get("max_search", 20))
temperature = float(body.get("temperature", 0.6))
ref_text = body.get("ref_text", "") # 직접 전달
ref_sid = body.get("ref_sid", "") # doc-upload에서 받은 sid
if not ref_text and ref_sid:
ref_text = _doc_text_store.get(ref_sid, "")
if not prompt:
return JSONResponse({"error": "prompt 없음"}, status_code=400)
# ════════════════════════════════════════════════════════
# 모드 분기: 1=새로 생성, 2=서식 유지·내용 변경, 3=구조 참고·새로 생성
# ════════════════════════════════════════════════════════
mode = int(body.get("mode", 1))
ref_hwpx_path = _doc_hwpx_store.get(ref_sid, "")
# ── 디버그 로그 ──
print(f"[MODE] mode={mode} ref_sid='{ref_sid}' hwpx='{ref_hwpx_path}' exists={os.path.exists(ref_hwpx_path) if ref_hwpx_path else False}")
print(f"[MODE] prompt[:100]='{prompt[:100]}'")
# ════════════════════════════════════════════════════════
# MODE 2: 서식 유지 · 내용 변경 (XML 직접 치환)
# ════════════════════════════════════════════════════════
if mode == 2 and ref_hwpx_path and os.path.exists(ref_hwpx_path):
# ── XML 직접 치환 모드: SOMA 전체 바이패스 ──
def _transform_in_thread():
try:
q.put(json.dumps({"active": "水",
"stream": "🔄 **문서 변환 모드** — 키워드 매핑 치환 (서식 100% 보존)\n\n📖 텍스트 추출 중...\n"}, ensure_ascii=False))
text_list, raw_xml, orig_flags = extract_text_nodes(ref_hwpx_path)
q.put(json.dumps({"active": "木",
"stream": f"📖 텍스트 노드 {len(text_list)}개 추출\n🤖 LLM 키워드 매핑 생성 중...\n"}, ensure_ascii=False))
mapping = generate_keyword_mapping(raw_xml, prompt, temperature)
if not mapping:
q.put(json.dumps({"active": "金", "done": True,
"final_doc": "⚠️ 변경할 키워드가 없습니다.",
"stream": "⚠️ 매핑 0건\n"}, ensure_ascii=False))
return
q.put(json.dumps({"active": "火",
"stream": f"🤖 {len(mapping)}쌍 매핑 생성\n🔧 XML 적용 중...\n"}, ensure_ascii=False))
new_xml, details = apply_keyword_mapping(raw_xml, mapping)
output_path = repack_transform_hwpx(ref_hwpx_path, new_xml, orig_flags)
# 파일 등록
orig_name = os.path.splitext(os.path.basename(ref_hwpx_path))[0]
final_name = f"{orig_name}_변환.hwpx"
final_path = os.path.join(os.path.dirname(output_path), final_name)
os.rename(output_path, final_path)
_file_registry[final_name] = final_path
if ref_sid:
_transform_store[ref_sid] = final_path
_last_transform["path"] = final_path
_last_transform["ts"] = time.time()
# 변경 사항 요약
total_count = sum(d.get("count",0) for d in details)
summary = "\n".join(f"• '{d['original']}' → '{d['replacement']}' ({d.get('count',0)}회)" for d in details)
final_doc = (
f"## 🔄 문서 변환 완료 (서식 100% 보존)\n\n"
f"**{len(details)}개 키워드 · {total_count}회 치환**\n\n{summary}\n\n"
f"---\n✅ header.xml/이미지/스크립트/charPr/paraPr: 원본 100% 보존\n"
f"✅ section0.xml: 키워드만 {total_count}회 치환\n"
)
q.put(json.dumps({"active": "金", "done": True,
"final_doc": final_doc,
"transform_file": f"/file/{final_name}",
"transform_filename": final_name,
"transform_path": final_path,
"stream": f"🎉 변환 완료! {len(details)}개 키워드 · {total_count}회 · 서식 100% 보존\n",
"search_count": 0}, ensure_ascii=False))
except Exception as e:
import traceback; traceback.print_exc()
q.put(json.dumps({"error": str(e), "done": True}, ensure_ascii=False))
finally:
q.put(None) # SSE 종료 신호
q = _queue.Queue()
threading.Thread(target=_transform_in_thread, daemon=True).start()
async def _async_gen():
while True:
try:
item = await _asyncio.get_event_loop().run_in_executor(
None, lambda: q.get(timeout=300))
except: break
if item is None:
yield "data: [DONE]\n\n"; break
yield f"data: {item}\n\n"
if '"done": true' in item or '"done":true' in item:
yield "data: [DONE]\n\n"; break
return StreamingResponse(_async_gen(),
media_type="text/event-stream",
headers={"Cache-Control": "no-cache",
"X-Accel-Buffering": "no"})
# ════════════════════════════════════════════════════════
# MODE 1 & 3: SOMA 파이프라인 (문서 신규 생성)
# MODE 3은 구조 참고 힌트 추가 + HWPX 생성 시 SectionCloner 사용
# ════════════════════════════════════════════════════════
full_prompt = prompt
if mode == 3 and ref_text and ref_text.strip():
# MODE 3: 참조 문서의 구조 골격을 압축 추출하여 주입
structure = extract_structure_summary(ref_text)
full_prompt = f"{prompt}\n\n{structure}"
elif mode == 1 and ref_text and ref_text.strip():
# MODE 1: 레퍼런스 텍스트가 있어도 참고자료로만 활용
ref_snippet = ref_text.strip()[:8000]
full_prompt = f"{prompt}\n\n[참고자료]\n{ref_snippet}"
# 동기 제너레이터를 별도 스레드에서 실행 → 이벤트 루프 블로킹 방지
q = _queue.Queue()
def _run_in_thread():
try:
for chunk in soma_pipeline(full_prompt, max_search, temperature):
q.put(json.dumps(chunk, ensure_ascii=False))
except Exception as e:
q.put(json.dumps({"error": str(e), "done": True}))
finally:
q.put(None) # 종료 시그널
threading.Thread(target=_run_in_thread, daemon=True).start()
async def _async_generate():
while True:
# 큐에서 비동기로 가져오기 (이벤트 루프 블로킹 없음)
try:
item = await _asyncio.get_event_loop().run_in_executor(
None, lambda: q.get(timeout=300))
except:
break
if item is None:
yield "data: [DONE]\n\n"
break
yield f"data: {item}\n\n"
return StreamingResponse(_async_generate(),
media_type="text/event-stream",
headers={"Cache-Control": "no-cache",
"X-Accel-Buffering": "no"})
except Exception as e:
return JSONResponse({"error": str(e)}, status_code=500)
@app.post("/soma/hml")
async def _soma_hml(req: _FAReq):
try:
body = await req.json()
content = body.get("content", "").strip()
ref_sid = body.get("ref_sid", "")
mode = int(body.get("mode", 1))
# ── MODE 2: 문서 변환 결과가 있으면 바로 반환 ──
if mode == 2:
# 1) ref_sid로 찾기
if ref_sid and ref_sid in _transform_store:
path = _transform_store.pop(ref_sid)
if os.path.exists(path):
fname = os.path.basename(path)
_file_registry[fname] = path
return JSONResponse({"file_url": f"/file/{fname}",
"filename": fname,
"file_path": path,
"mode": "transform"})
# 2) 글로벌 최근 변환 결과
if _last_transform["path"] and os.path.exists(_last_transform["path"]):
if time.time() - _last_transform["ts"] < 300:
path = _last_transform["path"]
_last_transform["path"] = ""
fname = os.path.basename(path)
_file_registry[fname] = path
return JSONResponse({"file_url": f"/file/{fname}",
"filename": fname,
"file_path": path,
"mode": "transform"})
# ── MODE 3: 레퍼런스 HWPX 구조 참고 → SectionCloner ──
ref_hwpx_path = ""
if mode == 3 and ref_sid:
ref_hwpx_path = _doc_hwpx_store.get(ref_sid, "")
if not content:
return JSONResponse({"error": "content 없음"}, status_code=400)
def _blocking():
if ref_hwpx_path and os.path.exists(ref_hwpx_path):
path = generate_hwpx(content, ref_hwpx_path=ref_hwpx_path)
else:
path = generate_hwpx(content)
title = normalize_text_for_title(content)
safe = re.sub(r'[\\/:*?"<>|]', '', title)[:40].strip() or "문서"
new_path = os.path.join(os.path.dirname(path), f"{safe}.hwpx")
os.rename(path, new_path)
return new_path
new_path = await _asyncio.get_event_loop().run_in_executor(None, _blocking)
fname = os.path.basename(new_path)
_file_registry[fname] = new_path
return JSONResponse({"file_url": f"/file/{fname}",
"filename": fname,
"file_path": new_path})
except Exception as e:
return JSONResponse({"error": str(e)}, status_code=500)
@app.get("/file/{fname}")
async def _serve_file(fname: str):
fpath = _file_registry.get(fname)
if fpath and os.path.exists(fpath):
return FileResponse(fpath, filename=fname,
media_type="application/octet-stream")
return JSONResponse({"error": "파일 없음"}, status_code=404)
@app.post("/soma/preview")
async def _soma_preview(req: _FAReq):
try:
body = await req.json()
if "b64" in body:
import base64 as _b64
ext = body.get("ext", ".hwpx").lower()
tmp = tempfile.NamedTemporaryFile(suffix=ext, delete=False)
tmp.write(_b64.b64decode(body["b64"]))
tmp.close()
fpath = tmp.name
else:
fpath = body.get("file_path", "")
if not fpath or not os.path.exists(fpath):
return HTMLResponse(_viewer_empty("파일을 찾을 수 없습니다."))
preview = await _asyncio.get_event_loop().run_in_executor(
None, hwpx_to_html_preview, fpath)
return HTMLResponse(preview)
except Exception as e:
return HTMLResponse(_viewer_empty(f"미리보기 오류: {e}"))
@app.get("/soma/status")
async def _soma_status():
return JSONResponse({
"status": "ok",
"hwpjs_ready": core._HWPJS_READY,
"engine": "ohah/hwpjs WASM" if core._HWPJS_READY else "Python lxml"
})
# HF Spaces 호환 — 헬스체크
@app.get("/api/health")
async def _health():
return JSONResponse({"status": "ok"})
# ── 문서 업로드 (텍스트 추출) ──
@app.post("/soma/doc-upload")
async def _soma_doc_upload(req: _FAReq):
"""업로드된 문서에서 텍스트 추출 (b64 또는 file_path)"""
try:
body = await req.json()
fpath = body.get("file_path", "")
if not (fpath and os.path.exists(fpath)):
import base64 as _b64
b64 = body.get("b64", "")
fname = body.get("filename", "document")
ext = body.get("ext", ".txt").lower()
tmp = tempfile.NamedTemporaryFile(suffix=ext, delete=False)
tmp.write(_b64.b64decode(b64))
tmp.close()
fpath = tmp.name
text, err = await _asyncio.get_event_loop().run_in_executor(
None, process_uploaded_file, fpath)
if text:
sid = str(id(text))[-8:]
_doc_text_store[sid] = text
# HWPX 파일이면 경로도 저장 (변환 모드용)
is_hwpx = fpath.lower().endswith('.hwpx')
if is_hwpx:
_doc_hwpx_store[sid] = fpath
print(f"[DOC-UPLOAD] sid={sid} fpath={fpath} is_hwpx={is_hwpx} hwpx_store_keys={list(_doc_hwpx_store.keys())}")
return JSONResponse({"ok": True, "sid": sid,
"chars": len(text),
"is_hwpx": is_hwpx,
"preview": text[:200]})
return JSONResponse({"ok": False, "error": err or "텍스트 추출 실패"})
except Exception as e:
return JSONResponse({"ok": False, "error": str(e)})
# ── 문서 QnA 챗 (SSE 스트리밍) ──
@app.post("/soma/chat")
async def _soma_chat(req: _FAReq):
"""문서 기반 QnA 챗 — SSE 스트리밍"""
try:
body = await req.json()
message = body.get("message", "").strip()
sid = body.get("sid", "")
history = body.get("history", [])
if not message:
return JSONResponse({"error": "message 없음"}, status_code=400)
if not FIREWORKS_API_KEY:
return JSONResponse({"error": "FIREWORKS_API_KEY 미설정"}, status_code=500)
doc_text = _doc_text_store.get(sid, "")
# 메시지 구성
if doc_text:
user_content = f"## 📄 업로드된 문서 내용\n---\n{doc_text[:12000]}\n---\n\n## 💬 질문\n{message}\n\n위 문서 내용을 바탕으로 답변해주세요."
else:
user_content = message
api_messages = [{"role": "system", "content": DOC_CHAT_SYSTEM}]
for h in (history or [])[-6:]:
if isinstance(h, (list, tuple, dict)):
if isinstance(h, dict):
api_messages.append({"role": h.get("role","user"), "content": h.get("content","")})
elif len(h) == 2:
api_messages.append({"role": "user", "content": h[0] or ""})
api_messages.append({"role": "assistant", "content": h[1] or ""})
api_messages.append({"role": "user", "content": user_content})
q2 = _queue.Queue()
def _chat_thread():
try:
headers = {"Accept":"application/json","Content-Type":"application/json",
"Authorization": f"Bearer {FIREWORKS_API_KEY}"}
payload = {"model": FIREWORKS_MODEL, "max_tokens": 16000,
"temperature": 0.6, "stream": True, "messages": api_messages}
resp = requests.post(FIREWORKS_URL, headers=headers, json=payload,
stream=True, timeout=180)
resp.raise_for_status()
for raw_line in resp.iter_lines():
if not raw_line: continue
line = raw_line.decode("utf-8") if isinstance(raw_line, bytes) else raw_line
if not line.startswith("data: "): continue
data = line[6:]
if data.strip() == "[DONE]":
break
try:
chunk = json.loads(data)
delta = chunk["choices"][0]["delta"].get("content", "")
if delta:
q2.put(json.dumps({"delta": delta}, ensure_ascii=False))
except:
pass
except Exception as e:
q2.put(json.dumps({"error": str(e)}))
finally:
q2.put(None)
threading.Thread(target=_chat_thread, daemon=True).start()
async def _async_chat():
while True:
try:
item = await _asyncio.get_event_loop().run_in_executor(
None, lambda: q2.get(timeout=300))
except:
break
if item is None:
yield "data: [DONE]\n\n"
break
yield f"data: {item}\n\n"
return StreamingResponse(_async_chat(),
media_type="text/event-stream",
headers={"Cache-Control": "no-cache",
"X-Accel-Buffering": "no"})
except Exception as e:
return JSONResponse({"error": str(e)}, status_code=500)
# ── Gradio를 /gradio 서브경로에 마운트 ──
demo = build_ui()
app = gr.mount_gradio_app(app, demo, path="/gradio")
print("✅ FastAPI 메인 서버")
print(" / → index.html")
print(" /gradio → Gradio UI")
print(" /soma/* → API")
uvicorn.run(app, host="0.0.0.0", port=7860) |