File size: 46,609 Bytes
a23394a | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 | import ast
from json import JSONDecodeError, JSONDecoder
import json
import re
from typing import Any, Dict, Union, List
from pathlib import Path
from dataflow_agent.logger import get_logger
log = get_logger(__name__)
import ast
from json import JSONDecodeError, JSONDecoder
import json
import re
from typing import Any, Dict, Union, List
from pathlib import Path
import asyncio
import logging
from typing import List, Dict, Any, Optional
from math import ceil
import time
import os
import math
from pptx.enum.text import PP_ALIGN
from pptx.dml.color import RGBColor
from pptx.enum.shapes import MSO_SHAPE
from pptx.util import Inches, Pt
import fitz # PyMuPDF
from PIL import Image
import pdfplumber
import uuid
from dataflow_agent.toolkits.imtool.mineru_tool import run_aio_batch_two_step_extract, run_aio_two_step_extract
def get_project_root() -> Path:
return Path(__file__).resolve().parent.parent
def robust_parse_json(
text: str,
*,
merge_dicts: bool = False,
strip_double_braces: bool = False
) -> Union[Dict[str, Any], List[Any]]:
"""
尽量从 LLM / 日志 / jsonl / Markdown 片段中提取合法 JSON。
参数
----
text : str
输入原始文本
merge_dicts : bool, default False
提取到多个对象且全部是 dict 时，是否用 dict.update 合并返回
strip_double_braces : bool, default False
把 '{{' / '}}' 替换成 '{' / '}'（某些模板语言会加双层花括号）
返回
----
Dict / List / List[Dict | List]
"""
s = text.strip()
# ---------- 预处理：剥去外层包裹 ----------
s = _remove_markdown_fence(s) # ```json ... ```
s = _remove_outer_triple_quotes(s) # ''' ... ''' / """ ... """
s = _remove_leading_json_word(s) # 开头一个 json/JSON 标记
if strip_double_braces:
s = s.replace("{{", "{").replace("}}", "}")
# ---------- 清理注释 & 尾逗号 ----------
s = _strip_json_comments(s)
# ---------- 新增：清理非法控制字符 ----------
# 移除所有 JSON 规范不允许的 ASCII 控制字符。
# 合法的 \n, \r, \t, 和 \f, \b, \" 都不会被移除，但这里只针对不可打印的控制码。
s = re.sub(r'[\x00-\x08\x0b\x0c\x0e-\x1f\x7f]', '', s)
# ---------- 新增：转义未转义的反斜杠（修复 LaTeX 公式等问题）----------
# 这会将所有单个反斜杠转换为双反斜杠，但保留已经正确转义的序列
# 先保护已经转义的序列（如 \\n, \\t, \\", \\\\）
s = s.replace('\\\\', '\x00DOUBLE_BACKSLASH\x00') # 临时标记
s = s.replace('\\n', '\x00NEWLINE\x00')
s = s.replace('\\r', '\x00RETURN\x00')
s = s.replace('\\t', '\x00TAB\x00')
s = s.replace('\\"', '\x00QUOTE\x00')
s = s.replace('\\/', '\x00SLASH\x00')
s = s.replace('\\b', '\x00BACKSPACE\x00')
s = s.replace('\\f', '\x00FORMFEED\x00')
# 现在转义所有剩余的单个反斜杠
s = s.replace('\\', '\\\\')
# 恢复之前保护的序列
s = s.replace('\x00DOUBLE_BACKSLASH\x00', '\\\\')
s = s.replace('\x00NEWLINE\x00', '\\n')
s = s.replace('\x00RETURN\x00', '\\r')
s = s.replace('\x00TAB\x00', '\\t')
s = s.replace('\x00QUOTE\x00', '\\"')
s = s.replace('\x00SLASH\x00', '\\/')
s = s.replace('\x00BACKSPACE\x00', '\\b')
s = s.replace('\x00FORMFEED\x00', '\\f')
log.debug(f'清洗完之后内容是： {s}')
# ---------- Step-1：整体解析 ----------
# Step-1
try:
result = json.loads(s)
log.info(f"整体解析成功，类型: {type(result)}")
return result
except JSONDecodeError as e:
log.warning(f"整体解析失败: {e}")
# ---------- Step-2：尝试 JSON Lines ----------
objs = _parse_json_lines(s)
if objs is not None:
return _maybe_merge(objs, merge_dicts)
# ---------- Step-3：流式提取多个对象 ----------
objs = _extract_json_objects(s)
log.warning(f"提取到 {len(objs)} 个对象")
if not objs:
raise ValueError("Unable to locate any valid JSON fragment.")
return _maybe_merge(objs, merge_dicts)
# ======================================================================
# 工具函数
# ======================================================================
_fence_pat = re.compile(r'```[\w-]*\s*([\s\S]*?)```', re.I)
# 只匹配外层包裹的代码块（整个内容被 ``` 包裹）
_outer_fence_pat = re.compile(r'^\s*```[\w-]*\s*([\s\S]*?)```\s*$', re.I)
def _remove_markdown_fence(src: str) -> str:
"""只提取外层包裹的 ``` … ``` 内文本；若内容不是被代码块包裹则原样返回"""
# 只处理整个内容被代码块包裹的情况，避免提取 JSON 内嵌的代码块
match = _outer_fence_pat.match(src)
if match:
return match.group(1).strip()
return src
def _remove_outer_triple_quotes(src: str) -> str:
if (src.startswith("'''") and src.endswith("'''")) or (
src.startswith('"""') and src.endswith('"""')
):
return src[3:-3].strip()
return src
def _remove_leading_json_word(src: str) -> str:
return src[4:].lstrip() if src.lower().startswith("json") else src
def _strip_json_comments(src: str) -> str:
# /* ... */ 块注释
src = re.sub(r'/\*[\s\S]*?\*/', '', src)
# // ... 行注释，排除 URL 中的 :// 和字符串内的 //）
src = re.sub(r'(?<![:\"\'])//.*', '', src)
# 尾逗号 ,}
src = re.sub(r',\s*([}\]])', r'\1', src)
return src.strip()
# ---------------- JSON Lines ----------------
def _parse_json_lines(src: str) -> Union[List[Any], None]:
lines = [ln.strip() for ln in src.splitlines() if ln.strip()]
if len(lines) <= 1: # 只有 0/1 行就不是 jsonl
return None
objs: List[Any] = []
for ln in lines:
try:
objs.append(json.loads(ln))
except JSONDecodeError:
return None # 某行不是合法 JSON，放弃 jsonl 方案
return objs
# ------------ 多对象提取（改进版） ------------
def _extract_json_objects(src: str) -> List[Any]:
dec = JSONDecoder()
idx, n = 0, len(src)
objs: List[Any] = []
while idx < n:
m = re.search(r'[{\[]', src[idx:])
if not m:
break
idx += m.start()
try:
obj, end = dec.raw_decode(src, idx)
# ========== 严格性检查 ==========
tail = src[end:].lstrip()
# 允许结束、逗号、换行、右括号、右中括号
if tail and tail[0] not in ',]}>\n\r':
idx += 1 # 可能是误判，如 {"a":1 <-- 缺 }
continue
objs.append(obj)
idx = end
except JSONDecodeError:
idx += 1
return objs
def _maybe_merge(objs: List[Any], merge_dicts: bool) -> Union[Any, List[Any]]:
if len(objs) == 1:
return objs[0]
if merge_dicts and all(isinstance(o, dict) for o in objs):
merged: Dict[str, Any] = {}
for o in objs:
merged.update(o)
return merged
return objs
# ========================================================================
# For Paper2Figure
# ========================================================================
async def run_mineru(image_path: Path, output_dir: Path) -> bool:
"""调用 mineru 并返回是否成功"""
cmd = [
"mineru",
"-p", str(image_path),
"--backend", "vlm-transformers",
"--source", "local",
"-o", str(output_dir)
]
proc = await asyncio.create_subprocess_exec(
*cmd,
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
)
stdout, stderr = await proc.communicate()
if proc.returncode != 0:
log.warning(f"[mineru] 运行失败: {stderr.decode(errors='ignore')}")
return False
log.info("[mineru] 执行成功")
return True
async def replace_item_with_sub_items(
items: List[Dict[str, Any]],
sub_items: List[Dict[str, Any]],
sub_img_path: str
) -> List[Dict[str, Any]]:
"""
处理每个 sub_item，执行坐标变换并替换原有的 item。
"""
# 加载图像并获取宽高
with Image.open(sub_img_path) as sub_img:
sub_img_width, sub_img_height = sub_img.size
expanded_items = [] # 用来存储展开后的 sub_items
items_to_remove = [] # 用来记录需要删除的 items 索引
log.info(f"[replace_item_with_sub_items] 开始替换图像: {sub_img_path}")
for i, item in enumerate(items):
log.info(f"item[type]: {item['type']}")
if item["type"] in ["image", "table"]:
log.info(f"item['img_path']: {item['img_path']}")
log.info(f"sub_img_path: {sub_img_path}")
if item["img_path"] == sub_img_path: # 用 sub_img_path 匹配原图的 img_path
# 获取目标图像的 bbox
target_bbox = item["bbox"]
xmin, ymin, xmax, ymax = target_bbox
# 计算比例
width_ratio = (xmax - xmin) / sub_img_width
height_ratio = (ymax - ymin) / sub_img_height
log.info(f"[replace_item_with_sub_items] 找到匹配的 item，替换 bbox: {target_bbox} -> 展开为 {len(sub_items)} 个 sub_items")
# 替换原始项中的 image 信息，展开 sub_items
for sub_item in sub_items:
# 获取 sub_item 的原始 bbox 坐标
sub_item_bbox = sub_item["bbox"]
sub_xmin, sub_ymin, sub_xmax, sub_ymax = sub_item_bbox
# 进行坐标线性变换
transformed_bbox = [
int(sub_xmin * width_ratio + xmin), # x 坐标变换
int(sub_ymin * height_ratio + ymin), # y 坐标变换
int(sub_xmax * width_ratio + xmin), # x 坐标变换
int(sub_ymax * height_ratio + ymin) # y 坐标变换
]
# 更新 sub_item 的 bbox
sub_item["bbox"] = transformed_bbox
# 将变换后的 sub_item 加入 expanded_items 列表
expanded_items.append(sub_item)
# 记录需要删除的原 item 索引
items_to_remove.append(i)
break
# 替换原项中的 item，删除原有的 item
log.info(f"[replace_item_with_sub_items] 将原 item 删除, 替换为 {len(expanded_items)} 个 sub_item")
for index in sorted(items_to_remove, reverse=True): # 逆序删除，避免索引错乱
del items[index]
# 将展开的 sub_items 插入到 items 中
items.extend(expanded_items)
log.info(f"[replace_item_with_sub_items] 替换完成后，当前 items 长度: {len(items)}")
return items
async def recursive_run_mineru(
img_path: Path,
out_dir: Path,
max_depth: int = 2,
current_depth: int = 0
) -> List[Dict[str, Any]]:
"""递归运行 mineru，处理子图并更新fig_mask"""
if current_depth > max_depth:
return [] # 达到最大递归深度时停止递归
log.info(f"[recursive_run_mineru] 当前深度 {current_depth}，处理图像: {img_path}")
# 调用 mineru 处理当前图像
ok = await run_mineru(img_path, out_dir)
if not ok:
return [] # 如果失败，则返回空结果
# 寻找并读取中间结果的 JSON 文件
content_json = locate_content_json(out_dir / img_path.stem)
if content_json is None:
return [] # 没有找到内容文件，返回空结果
items = load_and_fix_items(content_json, out_dir)
log.info(f"[recursive_run_mineru] 当前 items 长度: {len(items)}")
# 查找子图目录并递归处理每个子图
vlm_images_dir = out_dir / img_path.stem / 'vlm' / 'images'
sub_images = list(vlm_images_dir.glob("*.jpg")) # 假设子图为 .jpg 格式
if(current_depth != max_depth):
for sub_img_path in sub_images: # sub_img_path 是图像路径
log.info(f"[recursive_run_mineru] 处理子图: {sub_img_path}")
# 获取当前 sub_img_path 对应的 sub_items
sub_items = await recursive_run_mineru(sub_img_path, out_dir, max_depth, current_depth + 1)
# 处理替换和坐标变换
items = await replace_item_with_sub_items(items, sub_items, str(sub_img_path))
log.info(f"[recursive_run_mineru] 替换完成后，当前 items 长度: {len(items)}")
return items
def locate_content_json(output_dir: Path) -> Path | None:
"""寻找 *_middle.json 文件"""
files = list(output_dir.rglob("*_middle.json"))
if not files:
log.warning(f"[mineru] 未找到 *_middle.json in {output_dir}")
return None
return files[0]
def load_and_fix_items(content_json: Path, output_dir: Path) -> List[Dict[str, Any]]:
"""
读取 JSON 并提取所有文本和图片元素，修复图片路径为绝对路径
"""
try:
data = json.loads(content_json.read_text(encoding="utf-8"))
except Exception as e:
log.warning(f"[mineru] JSON 读取失败: {e}")
return []
# 提取基础名称
stem = content_json.stem # e.g. "paper1_middle"
base_name = stem.replace("_middle", "") or stem
results = []
# 遍历pdf_info中的para_blocks
if "pdf_info" in data and isinstance(data["pdf_info"], list):
for pdf_info in data["pdf_info"]:
if "para_blocks" in pdf_info and isinstance(pdf_info["para_blocks"], list):
for block in pdf_info["para_blocks"]:
block_type = block.get("type", "")
# 处理标题和普通文本
if block_type in ["title", "text", "paragraph"]:
# 提取文本内容
text_content = extract_text_from_block(block)
if text_content:
results.append({
"type": "text",
"text": text_content,
"bbox": block.get("bbox", []),
"text_level": 1 if block_type == "title" else None,
"page_idx": 0
})
# 处理图片块
elif block_type in ["list", "image", "table"]:
# 提取图片和相关的caption
image_elements = extract_image_elements(block, base_name, output_dir)
results.extend(image_elements)
return results
def extract_text_from_block(block: Dict) -> str:
"""从块中提取文本内容"""
text_parts = []
# 如果有lines字段，遍历提取
if "lines" in block and isinstance(block["lines"], list):
for line in block["lines"]:
if "spans" in line and isinstance(line["spans"], list):
for span in line["spans"]:
if span.get("type") == "text" and "content" in span:
text_parts.append(span["content"])
# 如果没有lines字段，尝试直接获取content
elif "content" in block:
text_parts.append(block["content"])
return " ".join(text_parts) if text_parts else ""
def extract_image_elements(block: Dict, base_name: str, output_dir: Path) -> List[Dict]:
"""从图片块中提取图片和相关的文本元素"""
elements = []
if "blocks" in block and isinstance(block["blocks"], list):
for sub_block in block["blocks"]:
sub_type = sub_block.get("type", "")
# 处理图片caption
if sub_type in ["title", "text", "paragraph","image_caption", "table_caption"]:
caption_text = extract_text_from_block(sub_block)
if caption_text:
elements.append({
"type": "text",
"text": caption_text,
"bbox": sub_block.get("bbox", []),
"text_level": None, # caption作为普通文本
"page_idx": 0
})
# 处理图片主体
elif sub_type in ["image_body", "table_body"]:
image_path = extract_image_path(sub_block, base_name, output_dir)
if image_path:
elements.append({
"type": "image",
"img_path": str(image_path),
"bbox": sub_block.get("bbox", []),
"image_caption": [],
"image_footnote": [],
"page_idx": 0
})
return elements
def extract_image_path(block: Dict, base_name: str, output_dir: Path) -> Optional[Path]:
"""从图片块中提取图片路径并转换为绝对路径"""
if "lines" in block and isinstance(block["lines"], list):
for line in block["lines"]:
if "spans" in line and isinstance(line["spans"], list):
for span in line["spans"]:
if span.get("type") in ["image", "table"] and "image_path" in span:
rel_path = span["image_path"]
if rel_path:
# 构建绝对路径
abs_path = output_dir / base_name / "vlm/images" / rel_path
return abs_path
return None
def build_output_directory(image_path: Path) -> Path:
"""构造 <image_no_ext>_mineru 输出目录"""
base = image_path.with_suffix("")
out_dir = Path(f"{base}_mineru")
out_dir.mkdir(parents=True, exist_ok=True)
return out_dir
# -------------------- NEW UTILS FUNCTIONS ---------------------------------
import asyncio
from pathlib import Path
from PIL import Image
from mineru_vl_utils import MinerUClient
# -----------------------------
# 工具：相对 bbox → 像素 bbox
# -----------------------------
def rel_bbox_to_pixel(bbox, width, height):
x1, y1, x2, y2 = bbox
return [
int(x1 * width),
int(y1 * height),
int(x2 * width),
int(y2 * height),
]
# -----------------------------
# 工具：截图保存图片
# -----------------------------
def crop_and_save(img: Image.Image, bbox_pixel, save_path: Path, margin=3):
W, H = img.size
x1, y1, x2, y2 = bbox_pixel
log.info(f"[crop_and_save] Image size={img.size}, bbox={bbox_pixel}")
# --- 1. 越界判断 ---
if x1 < 0 or y1 < 0 or x2 > W or y2 > H or x2 <= x1 or y2 <= y1:
log.info("[crop_and_save] BBOX out of range → return original path")
return None # 或 return str(original_image_path)
# --- 2. 判断四条边是否都“紧贴边缘” ---
touch_left = x1 <= margin
touch_top = y1 <= margin
touch_right = x2 >= W - margin
touch_bottom = y2 >= H - margin
if touch_left and touch_top and touch_right and touch_bottom:
log.info("[crop_and_save] BBOX covers almost entire image → skip crop, return original")
return None # 或 return str(original_image_path)
# --- 3. 正常裁剪 ---
try:
crop_img = img.crop(bbox_pixel)
crop_img.save(save_path)
log.info(f"[crop_and_save] Cropped size={crop_img.size}, saved={save_path}")
return str(save_path)
except Exception as e:
log.info(f"[crop_and_save] ERROR during crop: {e}")
return None
def transform_sub_bbox(sub_bbox, parent_bbox):
"""把子图内 bbox 映射回原图坐标系"""
px1, py1, px2, py2 = parent_bbox
sx1, sy1, sx2, sy2 = sub_bbox
return [
px1 + sx1,
py1 + sy1,
px1 + sx2,
py1 + sy2
]
# -----------------------------
# 主函数：HTTP异步递归 MinerU
# -----------------------------
async def recursive_run_mineru_http(
image_path: Path,
out_dir: Path,
port:int = 8010,
max_depth: int = 2,
current_depth: int = 0,
):
"""
使用 aio_two_step_extract 执行异步 mineru 提取。
不依赖中间 JSON 文件。
自动截图 image/table/list 等元素作为下一轮输入。
"""
# ---- 深度控制 ----
if current_depth > max_depth:
return []
out_dir.mkdir(parents=True, exist_ok=True)
log.info(f"[recursive_http] ─ Depth={current_depth}, Image={image_path}")
# ---- 加载图片 ----
img = Image.open(image_path)
W, H = img.size
log.info(f"[recursive_http] Image size: W={W}, H={H}")
# ---- 调用 MinerU 异步接口 ----
try:
blocks = await run_aio_two_step_extract(image_path, port)
log.info(f"[recursive_http] MinerU returned {len(blocks)} blocks")
except Exception as e:
log.info(f"[recursive_http] MinerU error: {e}")
return []
results = []
sub_images_paths = [] # 下一层递归的图片路径
# -----------------------------
# 解析 block 并处理不同类型
# -----------------------------
for idx, blk in enumerate(blocks):
btype = blk.get("type")
bbox_rel = blk.get("bbox", [0, 0, 1, 1])
content = blk.get("content")
log.info(f" Block[{idx}] type={btype}, bbox_rel={bbox_rel}, "
f"content={str(content)[:30] if content else None}")
bbox_pixel = rel_bbox_to_pixel(bbox_rel, W, H)
log.info(f" → bbox_pixel={bbox_pixel}")
# ---- 文本类 block：直接保存 ----
if btype in ["title", "text", "paragraph", "caption", "image_caption", "table_caption"]:
results.append({
"type": "text",
"text": content or "",
"bbox": bbox_pixel,
})
log.info(f" Added TEXT block, content preview: {str(content)[:30]}")
# ---- 图片类 block：截图作为下一轮输入 ----
elif btype in ["image", "table", "list"]:
sub_img_name = f"sub_{current_depth}_{uuid.uuid4()}.png"
sub_img_path = out_dir / sub_img_name
log.info(f"Try to crop img: {image_path}")
cropped_path = crop_and_save(img, bbox_pixel, sub_img_path)
sub_img_path = cropped_path if cropped_path else image_path
log.info(f" Cropped IMAGE block → {sub_img_path}")
results.append({
"type": "image",
"img_path": str(sub_img_path),
"bbox": bbox_pixel,
})
sub_images_paths.append(sub_img_path)
log.info(f"[recursive_http] Depth={current_depth} → Parsed {len(results)} items, {len(sub_images_paths)} sub-images")
# ----------------------------------------------
# 递归处理子图，并将对应的 image 元素替换为子结果
# ----------------------------------------------
if current_depth < max_depth and len(sub_images_paths) > 0:
tasks = [
recursive_run_mineru_http(
sub_img_path,
out_dir,
port=port,
max_depth=max_depth,
current_depth=current_depth + 1
)
for sub_img_path in sub_images_paths
]
sub_results_list = await asyncio.gather(*tasks)
new_results = []
sub_map = {
str(path): sub_results_list[i]
for i, path in enumerate(sub_images_paths)
}
for item in results:
img_path = item.get("img_path")
# 只对当前层生成的子图做展开替换
if item["type"] in ["image", 'table', 'list'] and img_path in sub_map:
parent_bbox = item["bbox"]
sub_items = sub_map[img_path]
log.info(f" Replacing sub-image {img_path} with {len(sub_items)} items")
for si in sub_items:
new_item = si.copy()
new_item["bbox"] = transform_sub_bbox(si["bbox"], parent_bbox)
new_results.append(new_item)
else:
new_results.append(item)
results = new_results
return results
# -------------------------------------------------------------------------------------
def get_font_size_for_text(bbox, text, max_font_size=48, min_font_size=10):
"""
根据文本框的 bbox 和文本长度推算字体大小
bbox: [xmin, ymin, xmax, ymax]
text: 要插入的文本
"""
box_height = bbox[3] - bbox[1] # 计算文本框高度
max_chars_per_line = 30 # 最大每行字符数
lines = ceil(len(text) / max_chars_per_line) # 计算需要的行数
font_size = min(box_height // lines, max_font_size)
return max(font_size, min_font_size)
def generate_ppt_filename(output_path):
"""
生成一个唯一的 PPT 文件名，基于当前时间戳
"""
timestamp = time.strftime("%Y%m%d_%H%M%S")
return f"{output_path}/presentation_{timestamp}.pptx"
def pixels_to_inches(pixels: int, dpi: int = 96) -> float:
"""将像素转换为英寸"""
return pixels / dpi
def calculate_font_size(text: str, bbox: List[int], text_level: int = None) -> int:
"""
根据文本框大小、文字内容和文本级别计算合适的字体大小
"""
# 计算文本框的宽度和高度（像素）
width = bbox[2] - bbox[0]
height = bbox[3] - bbox[1]
# 根据文本级别设置基础字体大小
if text_level == 1: # 主标题
base_size = min(height * 0.8, 44)
elif text_level == 2: # 子标题
base_size = min(height * 0.7, 32)
else: # 正文
base_size = min(height * 0.6, 24)
# 根据文本长度调整
char_count = len(text)
if char_count > 0:
chars_per_line = max(1, width / (base_size * 0.6))
lines_needed = math.ceil(char_count / chars_per_line)
max_lines = max(1, height / (base_size * 1.1))
if lines_needed > max_lines:
base_size = base_size * (max_lines / lines_needed)
# 限制字体大小范围
font_size = max(8, min(base_size, 72))
return int(font_size)
def setup_presentation_size(prs, slide_width_px: int = 1024, slide_height_px: int = 1024):
"""设置PPT尺寸"""
prs.slide_width = Inches(pixels_to_inches(slide_width_px))
prs.slide_height = Inches(pixels_to_inches(slide_height_px))
return slide_width_px, slide_height_px
def add_text_element(slide, element: Dict):
"""添加文本元素到幻灯片"""
bbox = element.get('bbox', [0, 0, 100, 50])
text = element.get('text', '')
text_level = element.get('text_level')
# 计算位置和大小
left = pixels_to_inches(bbox[0])
top = pixels_to_inches(bbox[1])
width = pixels_to_inches(bbox[2] - bbox[0])
height = pixels_to_inches(bbox[3] - bbox[1])
# 计算字体大小
font_size = calculate_font_size(text, bbox, text_level)
log.info(f"添加文本框:")
log.info(f" 位置: [{bbox[0]}, {bbox[1]}, {bbox[2]}, {bbox[3]}] 像素")
log.info(f" 英寸坐标: left={left:.2f}, top={top:.2f}, width={width:.2f}, height={height:.2f}")
log.info(f" 文本内容: {text[:30]}{'...' if len(text) > 30 else ''}")
log.info(f" 文本级别: {text_level}, 字体大小: {font_size}pt")
# 添加文本框
textbox = slide.shapes.add_textbox(
Inches(left), Inches(top), Inches(width) * 1.2, Inches(height)
)
text_frame = textbox.text_frame
text_frame.word_wrap = True
# 设置文本内容
paragraph = text_frame.paragraphs[0]
paragraph.text = text
# 设置字体样式
paragraph.font.size = Pt(font_size)
paragraph.font.name = "Comic Sans MS"
# 根据文本级别设置样式
if text_level == 1:
paragraph.font.bold = True
paragraph.alignment = PP_ALIGN.CENTER
log.info(" 样式: 标题(加粗、居中)")
elif text_level == 2:
paragraph.font.bold = True
log.info(" 样式: 子标题(加粗)")
else:
log.info(" 样式: 正文")
return textbox
def add_image_element(slide, element: Dict):
"""添加图片元素到幻灯片"""
bbox = element.get('bbox', [0, 0, 100, 100])
img_path = element.get('img_path', '')
# 计算位置和大小
left = pixels_to_inches(bbox[0])
top = pixels_to_inches(bbox[1])
width = pixels_to_inches(bbox[2] - bbox[0])
height = pixels_to_inches(bbox[3] - bbox[1])
log.info(f"添加图片:")
log.info(f" 位置: [{bbox[0]}, {bbox[1]}, {bbox[2]}, {bbox[3]}] 像素")
log.info(f" 英寸坐标: left={left:.2f}, top={top:.2f}, width={width:.2f}, height={height:.2f}")
log.info(f" 图片路径: {img_path}")
log.info(f" 图片尺寸: {bbox[2]-bbox[0]}x{bbox[3]-bbox[1]} 像素")
# 检查图片文件是否存在
if os.path.exists(img_path):
try:
log.info(" 图片文件存在，正在添加...")
result = slide.shapes.add_picture(
img_path,
Inches(left), Inches(top), Inches(width), Inches(height)
)
log.info(" 图片添加成功")
return result
except Exception as e:
log.error(f" 添加图片时出错: {e}")
return add_image_placeholder(slide, bbox, f"Error: {str(e)}")
else:
log.warning(" 图片文件不存在，使用占位符")
return add_image_placeholder(slide, bbox, "Image not found")
def add_image_placeholder(slide, bbox: List[int], message: str):
"""添加图片占位符"""
left = pixels_to_inches(bbox[0])
top = pixels_to_inches(bbox[1])
width = pixels_to_inches(bbox[2] - bbox[0])
height = pixels_to_inches(bbox[3] - bbox[1])
# 添加矩形作为占位符
shape = slide.shapes.add_shape(
MSO_SHAPE.RECTANGLE,
Inches(left), Inches(top), Inches(width), Inches(height)
)
shape.fill.solid()
shape.fill.fore_color.rgb = RGBColor(240, 240, 240)
shape.line.color.rgb = RGBColor(200, 200, 200)
# 添加说明文字
textbox = slide.shapes.add_textbox(
Inches(left), Inches(top), Inches(width), Inches(height)
)
text_frame = textbox.text_frame
text_frame.text = message
text_frame.paragraphs[0].alignment = PP_ALIGN.CENTER
text_frame.paragraphs[0].font.size = Pt(10)
text_frame.paragraphs[0].font.name = "Comic Sans MS"
text_frame.paragraphs[0].font.color.rgb = RGBColor(128, 128, 128)
return shape
# def robust_parse_json(
# s: str,
# merge_dicts: bool = False, # 想合并时显式打开
# strip_double_braces: bool = False # 可选：把 {{ }} → { }
# ) -> Union[Dict[str, Any], List[Any]]:
# """
# 既能解析普通 JSON，也能从混杂文本中提取多个对象。
# - 支持 // 和 /* */ 注释、尾逗号
# - 自动去除 Markdown 代码块 ```json … ```、三引号 ''' … '''
# - 默认返回 dict / list 的原始结构
# - 若提取到多个独立对象，可用 merge_dicts=True 合并
# """
# # ---------- 预处理 ----------
# # 1) 去掉 ```xxx 代码围栏
# s = re.sub(r'```[\w]*\s*', '', s) # 开始围栏
# s = re.sub(r'```', '', s) # 结束围栏
# # 2) 去掉成对的三引号 ''' 或 """
# s = re.sub(r"^'''|'''$|^\"\"\"|\"\"\"$", '', s.strip())
# # 3) 可选：{{ }} → { }
# if strip_double_braces:
# s = s.replace('{{', '{').replace('}}', '}')
# # 4) 去注释 + 尾逗号
# s = _strip_json_comments(s)
# # ---------- 步骤 1：整体解析 ----------
# try:
# return json.loads(s)
# except JSONDecodeError:
# pass
# # ---------- 步骤 2：提取多个对象 ----------
# objs: List[Any] = _extract_json_objects(s)
# if not objs:
# raise ValueError("No valid JSON found.")
# # 单个对象
# if len(objs) == 1:
# return objs[0]
# # 多对象：根据参数决定
# if merge_dicts and all(isinstance(o, dict) for o in objs):
# merged: Dict[str, Any] = {}
# for o in objs:
# merged.update(o)
# return merged
# return objs
# def _strip_json_comments(s: str) -> str:
# # s = re.sub(r'/\*.*?\*/', '', s, flags=re.S) # 块注释
# # s = re.sub(r'//.*?$', '', s, flags=re.M) # 行注释
# s = re.sub(r',\s*([}\]])', r'\1', s) # 尾逗号
# return s
# def _extract_json_objects(s: str) -> List[Any]:
# dec = JSONDecoder()
# idx, n = 0, len(s)
# objs = []
# while idx < n:
# # 找下一个 { 或 [
# m = re.search(r'[{\[]', s[idx:])
# if not m:
# break
# idx += m.start()
# try:
# obj, end = dec.raw_decode(s, idx)
# objs.append(obj)
# idx = end
# except JSONDecodeError:
# idx += 1
# return objs
# ========================================================================
# For Paper2ExpFigure
# ========================================================================
def pdf_to_pil_images(
pdf_path: Union[str, Path],
dpi: int = 300
) -> List[Image.Image]:
"""
将 PDF 文件的每一页转换为 PIL Image 对象。
修复了 CMYK/灰度/透明背景导致的白图或花屏问题。
"""
pdf_path = Path(pdf_path)
if not pdf_path.exists():
raise FileNotFoundError(f"PDF 文件不存在: {pdf_path}")
# 计算缩放比例
zoom = dpi / 72.0
matrix = fitz.Matrix(zoom, zoom)
images: List[Image.Image] = []
# 这里的 flags 用于处理一些复杂的渲染情况（可选，但在某些图表 PDF 中很有用）
# fitz.pdf.Page.get_pixmap 默认会自动处理大多情况
doc = fitz.open(pdf_path)
try:
for page_num in range(len(doc)):
page = doc[page_num]
# 1. 获取初始 Pixmap
# alpha=False: 强制背景不透明（默认白色），解决透明底变黑或变白的问题
pix = page.get_pixmap(matrix=matrix, alpha=False)
# 2. 关键修复：检查色彩空间并转换
# pix.n 是通道数。如果不是 3 (RGB)，则强制转换为 RGB
if pix.n != 3:
# fitz.csRGB 是 PyMuPDF 内置的 RGB 色彩空间定义
temp_pix = fitz.Pixmap(fitz.csRGB, pix)
pix = temp_pix # 替换为转换后的 RGB pixmap
# 注意：temp_pix 只是引用，赋值给 pix 后后续逻辑一致，且会自动管理内存
# 3. 安全转换为 PIL
# 现在我们可以 100% 确定数据是 RGB 格式的
img = Image.frombytes("RGB", (pix.width, pix.height), pix.samples)
images.append(img)
log.info(f"[pdf_to_pil_images] 已转换第 {page_num + 1} 页 (模式:{pix.n}通道)，尺寸: {pix.width}x{pix.height}")
except Exception as e:
log.error(f"[pdf_to_pil_images] 转换过程出错: {e}")
raise e
finally:
doc.close()
log.info(f"[pdf_to_pil_images] 完成，共生成 {len(images)} 张图片")
return images
def _parse_html_table(html_content: str) -> tuple[List[str], List[List[str]]]:
"""
解析 HTML 表格内容，提取表头和数据行。
参数
----
html_content : str
HTML 表格字符串，如 <table>...</table>
返回
----
tuple[List[str], List[List[str]]]
(headers, rows) - 表头列名和数据行
"""
try:
from bs4 import BeautifulSoup
except ImportError:
log.warning("[_parse_html_table] BeautifulSoup 未安装，使用简单解析")
return _parse_html_table_simple(html_content)
try:
soup = BeautifulSoup(html_content, 'html.parser')
table = soup.find('table')
if not table:
log.warning("[_parse_html_table] 未找到 table 标签")
return [], []
# 提取所有行
rows_data = []
for tr in table.find_all('tr'):
row = []
for cell in tr.find_all(['td', 'th']):
# 处理 colspan
colspan = int(cell.get('colspan', 1))
cell_text = cell.get_text(strip=True)
row.append(cell_text)
# 如果有 colspan，添加空单元格
for _ in range(colspan - 1):
row.append('')
if row: # 只添加非空行
rows_data.append(row)
if not rows_data:
return [], []
# 第一行作为表头
headers = rows_data[0]
data_rows = rows_data[1:] if len(rows_data) > 1 else []
return headers, data_rows
except Exception as e:
log.error(f"[_parse_html_table] 解析失败: {e}")
return _parse_html_table_simple(html_content)
def _parse_html_table_simple(html_content: str) -> tuple[List[str], List[List[str]]]:
"""
简单的 HTML 表格解析（不依赖 BeautifulSoup）。
仅用于备用，可能不够健壮。
"""
try:
# 简单正则提取所有 <tr>...</tr> 内容
import re
tr_pattern = re.compile(r'<tr>(.*?)</tr>', re.DOTALL | re.IGNORECASE)
td_pattern = re.compile(r'<t[dh][^>]*>(.*?)</t[dh]>', re.DOTALL | re.IGNORECASE)
rows_data = []
for tr_match in tr_pattern.finditer(html_content):
tr_content = tr_match.group(1)
row = []
for td_match in td_pattern.finditer(tr_content):
cell_text = td_match.group(1).strip()
# 移除 HTML 标签
cell_text = re.sub(r'<[^>]+>', '', cell_text)
row.append(cell_text)
if row:
rows_data.append(row)
if not rows_data:
return [], []
headers = rows_data[0]
data_rows = rows_data[1:] if len(rows_data) > 1 else []
return headers, data_rows
except Exception as e:
log.error(f"[_parse_html_table_simple] 简单解析失败: {e}")
return [], []
def extract_tables_from_mineru_results(
mineru_items: List[Dict[str, Any]],
min_rows: int = 2,
min_cols: int = 2,
) -> List[Dict[str, Any]]:
"""
从 MinerU 识别结果中提取表格数据。
参数
----
mineru_items : List[Dict[str, Any]]
MinerU 返回的 items 列表，每个 item 包含 type, bbox, content 等字段
min_rows : int, default 2
最小行数，少于此值的表格会被过滤
min_cols : int, default 2
最小列数，少于此值的表格会被过滤
返回
----
List[Dict[str, Any]]
提取的表格列表，每个表格格式:
{
"table_id": str, # 表格唯一标识
"headers": List[str], # 表头列名
"rows": List[List[str]], # 数据行
"caption": str, # 表格标题/说明
"bbox": List[int], # 原始坐标
"content": str, # 原始 HTML 内容（如果是 HTML 表格）
}
"""
tables = []
table_idx = 0
# 用于关联 caption 和 table 的临时存储
pending_caption = ""
for item in mineru_items:
item_type = item.get("type", "")
content = item.get("content", "")
bbox = item.get("bbox", [])
# 处理 table_caption
if item_type == "table_caption" and content:
pending_caption = content
continue
# 处理表格
if item_type == "table" and content:
# MinerU 返回的 content 是 HTML 字符串（如 <table>...</table>）
headers, rows = _parse_html_table(content)
# 过滤小表格
if len(headers) < min_cols or len(rows) < min_rows:
log.debug(f"[extract_tables] 跳过小表格: {len(headers)} 列, {len(rows)} 行")
continue
table_id = f"table_{table_idx}"
table_idx += 1
tables.append({
"table_id": table_id,
"headers": headers,
"rows": rows,
"caption": pending_caption,
"bbox": bbox,
"content": content, # 保留原始 HTML
})
pending_caption = "" # 重置
log.info(f"[extract_tables] 提取表格 {table_id}: {len(headers)} 列, {len(rows)} 行, caption: {pending_caption[:50] if pending_caption else 'N/A'}")
log.info(f"[extract_tables] 共提取 {len(tables)} 个表格")
return tables
def extract_text_from_mineru_results(
mineru_items: List[Dict[str, Any]],
max_chars: int = 10000,
) -> str:
"""
从 MinerU 识别结果中提取纯文本内容（用于 paper_idea_extractor）。
参数
----
mineru_items : List[Dict[str, Any]]
MinerU 返回的 items 列表
max_chars : int, default 10000
最大提取字符数，避免内容过长
返回
----
str
提取的文本内容
"""
text_parts = []
total_chars = 0
for item in mineru_items:
if total_chars >= max_chars:
break
item_type = item.get("type", "")
content = item.get("content", "")
# 提取文本类型的内容（text, title, table_caption 等）
if item_type in ["text", "title", "table_caption"] and content:
text_parts.append(content)
total_chars += len(content)
result = "\n\n".join(text_parts)
if total_chars > max_chars:
result = result[:max_chars] + "..."
log.info(f"[extract_text] 提取了 {len(text_parts)} 段文本，共 {len(result)} 字符")
return result
def execute_matplotlib_code(
code: str,
output_path: Union[str, Path],
timeout: int = 30,
allowed_modules: Optional[List[str]] = None,
) -> Dict[str, Any]:
"""
安全执行 matplotlib 代码并保存图表。
参数
----
code : str
要执行的 matplotlib Python 代码
output_path : str | Path
图表输出路径（包含文件名，如 /tmp/chart.png）
timeout : int, default 30
执行超时时间（秒）
allowed_modules : List[str], optional
允许导入的模块列表，默认为 ["matplotlib", "numpy", "pandas"]
返回
----
Dict[str, Any]
{
"success": bool,
"output_path": str, # 成功时返回图片路径
"error": str, # 失败时返回错误信息
}
"""
import subprocess
import tempfile
if allowed_modules is None:
allowed_modules = ["matplotlib", "numpy", "pandas", "math"]
output_path = Path(output_path)
output_path.parent.mkdir(parents=True, exist_ok=True)
# 安全性检查：禁止危险操作
dangerous_patterns = [
r'\bos\.system\b',
r'\bsubprocess\b',
r'\beval\b',
r'\bexec\b',
r'\bopen\s*\(',
r'\b__import__\b',
r'\bimport\s+os\b',
r'\bimport\s+sys\b',
r'\bimport\s+subprocess\b',
r'\bfrom\s+os\b',
r'\bfrom\s+sys\b',
]
for pattern in dangerous_patterns:
if re.search(pattern, code):
return {
"success": False,
"output_path": "",
"error": f"检测到危险操作: {pattern}",
}
# 构建完整的执行代码
# full_code = f'''
# import matplotlib
# matplotlib.use('Agg') # 非交互式后端
# import matplotlib.pyplot as plt
# import numpy as np
# # 用户代码
# {code}
# # 保存图表
# plt.tight_layout()
# plt.savefig(r"{str(output_path)}", dpi=150, bbox_inches='tight')
# plt.close('all')
# print("SUCCESS")
# '''
full_code = f'''
{code}
'''
# 写入临时文件并执行
try:
with tempfile.NamedTemporaryFile(mode='w', suffix='.py', delete=False, encoding='utf-8') as f:
f.write(full_code)
temp_script = f.name
result = subprocess.run(
['python', temp_script],
capture_output=True,
text=True,
timeout=timeout,
cwd=str(output_path.parent),
)
# 清理临时文件
try:
os.unlink(temp_script)
except Exception:
pass
# 判断执行是否成功：返回码为 0 且图片文件存在
if result.returncode == 0:
if output_path.exists():
log.info(f"[execute_matplotlib] 图表生成成功: {output_path}")
return {
"success": True,
"output_path": str(output_path),
"error": "",
}
else:
log.warning(f"[execute_matplotlib] 代码执行成功但图片未生成")
return {
"success": False,
"output_path": "",
"error": "代码执行成功但图片未生成",
}
else:
error_msg = result.stderr or result.stdout or "未知错误"
log.warning(f"[execute_matplotlib] 执行失败 (返回码={result.returncode}): {error_msg}")
return {
"success": False,
"output_path": "",
"error": error_msg[:500], # 截断过长的错误信息
}
except subprocess.TimeoutExpired:
log.warning(f"[execute_matplotlib] 执行超时 ({timeout}s)")
return {
"success": False,
"output_path": "",
"error": f"代码执行超时 ({timeout} 秒)",
}
except Exception as e:
log.error(f"[execute_matplotlib] 执行异常: {e}")
return {
"success": False,
"output_path": "",
"final_code": full_code,
"error": str(e),
}
|