Spaces:
Running
Running
File size: 123,549 Bytes
c47ec10 | 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 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 | import base64
import json
import mimetypes
import os
import random
import re
import time
import urllib.error
import urllib.request
from concurrent.futures import ThreadPoolExecutor
from dataclasses import dataclass
from io import BytesIO
from pathlib import Path
from collections.abc import Callable
from typing import Any, Dict, Iterator, Optional
from urllib.parse import unquote, urlparse
from curl_cffi import requests
from PIL import Image
from services.account_service import account_service
from services.config import config
from services.proxy_service import proxy_settings
from utils.helper import UpstreamHTTPError, ensure_ok, iter_sse_payloads, new_uuid, split_image_model
from utils.log import logger
from utils.pow import build_legacy_requirements_token, build_proof_token, parse_pow_resources
from utils.turnstile import solve_turnstile_token
class InvalidAccessTokenError(RuntimeError):
pass
class ImagePollTimeoutError(RuntimeError):
pass
class ImageContentPolicyError(RuntimeError):
"""Raised when image generation is blocked by content policy moderation."""
pass
@dataclass
class ChatRequirements:
"""保存一次对话请求所需的 sentinel token。"""
token: str
proof_token: str = ""
turnstile_token: str = ""
so_token: str = ""
raw_finalize: Optional[Dict[str, Any]] = None
DEFAULT_CLIENT_VERSION = "prod-a194cd50d4416d3c0b47c740f206b12ce60f5887"
DEFAULT_CLIENT_BUILD_NUMBER = "6708908"
DEFAULT_POW_SCRIPT = "https://chatgpt.com/backend-api/sentinel/sdk.js"
CODEX_IMAGE_MODEL = "codex-gpt-image-2"
CODEX_RESPONSES_MODEL = "gpt-5.5"
SEARCH_MODEL = "gpt-5-5"
SEARCH_TIMEOUT_SECS = 300.0
SEARCH_POLL_INTERVAL_SECS = 3.0
SEARCH_DONE_STATUS = {"finished_successfully", "finished_partial_completion"}
SEARCH_CONVERSATION_ID_RE = re.compile(r'"conversation_id"\s*:\s*"([^"]+)"')
SEARCH_URL_RE = re.compile(r"https?://[^\s\"'<>）)\]}]+")
EDITABLE_FILE_MODEL = "gpt-5-5-thinking"
EDITABLE_FILE_THINKING_EFFORT = "extended"
EDITABLE_FILE_TIMEOUT_SECS = 1200.0
EDITABLE_FILE_POLL_INTERVAL_SECS = 5.0
EDITABLE_FILE_CLIENT_VERSION = "prod-bede35f9dcd856d080e012478f0c1031faa2588e"
EDITABLE_FILE_CLIENT_BUILD_NUMBER = "6631702"
EDITABLE_FILE_PSD_OUTPUT_DIR = "data/files/psd"
EDITABLE_FILE_PPT_OUTPUT_DIR = "data/files/ppt"
EDITABLE_FILE_PPT_PROMPT = """我需要你根据用户的需求，来制作一个可以编辑的PPT，你可以使用Agent来做，你不要再继续询问用户问题，内容风格、版式、配色、内容结构和页面信息你可以自行补充并直接执行。整体的流程如下：
1. 用生图的方式，帮我生成一个精美的产品介绍ppt，5-6个页面
2. 帮我把以上涉及到的所有图像和形状素材拆分成单独png，每个素材单独一张图片，不要有遗漏，让我可以直接在ppt里拼接素材还原，不要文字
3. 利用以上所有图片和形状素材，帮我还原你第一次生成的展示ppt，我需要是可编辑的ppt格式，主要部分需要你单独还原插入，文字需要可以编辑
最后只需要给我生成一个PPT文件，以及生成中遇到的各种素材压缩包zip文件就行。"""
EDITABLE_FILE_PSD_PROMPT = "帮我生成这个图像，把这张海报分成若干图像，包括背景图，每个元素不要改位置，这样子我可以直接在 平时里无需拖动，底色为白色，不要伪透明底。再帮我将以上拆分的图像拼合成一个psd文件，去除白色底，不要改变每个图层的相应位置，保留每个元素所在图层的相应位置，保留每个元素的图层，最后只需要给我输出psd文件，以及每个图层的zip文件"
EDITABLE_ASSET_POINTER_RE = re.compile(r"(?:file-service|sediment)://([A-Za-z0-9_-]+)")
EDITABLE_ZIP_MIME_TYPES = {"application/zip", "application/x-zip-compressed"}
EDITABLE_PSD_MIME_TYPES = {"image/vnd.adobe.photoshop", "application/vnd.adobe.photoshop"}
EDITABLE_PPT_MIME_TYPES = {
"application/vnd.openxmlformats-officedocument.presentationml.presentation",
"application/vnd.ms-powerpoint",
}
EDITABLE_PSD_EXPORT_FILE_RE = re.compile(r"(?:sandbox:)?(/mnt/data/[^\s\"'\)\]]+\.(?:psd|zip))", re.IGNORECASE)
EDITABLE_PPT_EXPORT_FILE_RE = re.compile(r"(?:sandbox:)?(/mnt/data/[^\s\"'\)\]]+\.(?:pptx?|zip))", re.IGNORECASE)
FILE_SERVICE_ID_RE = re.compile(r"file-service://([A-Za-z0-9_-]+)")
FILE_ID_RE = re.compile(r"\b(file[-_](?!service\b)[A-Za-z0-9_-]+)\b")
# 真正的图片文件 ID 格式：file_00000000 + 24位十六进制字符（共32字符）
REAL_IMAGE_FILE_ID_RE = re.compile(r"\bfile_00000000[a-f0-9]{24}\b")
SEDIMENT_ID_RE = re.compile(r"sediment://([A-Za-z0-9_-]+)")
IMAGE_POLL_SETTLE_SECS = 2.0
CODEX_RESPONSES_INSTRUCTIONS = (
"Use the image_generation tool to create exactly one image for the user's request. "
"Return the generated image result."
)
# 内容政策违规错误关键词（上游拒绝生成图片的各种表述）
_CONTENT_POLICY_KEYWORDS = (
# 明确的内容政策违规
"内容政策", "防护限制", "违反", "moderation", "policy", "blocked",
# 拒绝生成类
"不能生成", "无法生成", "不能帮助", "无法帮助",
# 敏感内容类
"裸体", "裸露", "色情", "性内容", "未成年",
# 通用拒绝
"抱歉，我不能",
)
def _is_content_policy_error(error_msg: str) -> bool:
"""检查错误消息是否为内容政策违规。"""
if not error_msg:
return False
msg_lower = error_msg.lower()
return any(keyword in msg_lower for keyword in _CONTENT_POLICY_KEYWORDS)
@dataclass
class EditableFileArtifact:
attachment_id: str = ""
file_id: str = ""
name: str = ""
mime_type: str = ""
create_time: float = 0.0
author_role: str = ""
sandbox_path: str = ""
message_id: str = ""
@dataclass
class EditableFileExportResult:
conversation_id: str
primary_path: Path
zip_path: Path
class OpenAIBackendAPI:
"""ChatGPT Web 后端封装。
说明：
- 传入 `access_token` 时，聊天和模型列表都会走已登录链路
例如 `/backend-api/sentinel/chat-requirements`、`/backend-api/conversation`
- 不传 `access_token` 时，会走未登录链路
例如 `/backend-anon/sentinel/chat-requirements`、`/backend-anon/conversation`
- `stream_conversation()` 是底层统一流式入口
- 协议兼容转换放在 `services.protocol`
"""
def __init__(self, access_token: str = "") -> None:
"""初始化后端客户端。
参数：
- `access_token`：可选。传入后表示使用已登录链路；不传则使用未登录链路。
"""
self.base_url = "https://chatgpt.com"
self.client_version = DEFAULT_CLIENT_VERSION
self.client_build_number = DEFAULT_CLIENT_BUILD_NUMBER
self.access_token = access_token
self.account = account_service.get_account(self.access_token) if self.access_token else {}
self.account = self.account if isinstance(self.account, dict) else {}
self.fp = self._build_fp()
self.user_agent = self.fp["user-agent"]
self.device_id = self.fp["oai-device-id"]
self.session_id = self.fp["oai-session-id"]
self.pow_script_sources: list[str] = []
self.pow_data_build = ""
self.progress_callback: Callable[[str], None] | None = None
self.session = requests.Session(**proxy_settings.build_session_kwargs(
account=self.account,
impersonate=self.fp["impersonate"],
verify=True,
))
self.session.headers.update({
"User-Agent": self.user_agent,
"Origin": self.base_url,
"Referer": self.base_url + "/",
"Accept-Language": "zh-CN,zh;q=0.9,en;q=0.8,en-US;q=0.7",
"Cache-Control": "no-cache",
"Pragma": "no-cache",
"Priority": "u=1, i",
"Sec-Ch-Ua": self.fp["sec-ch-ua"],
"Sec-Ch-Ua-Arch": '"x86"',
"Sec-Ch-Ua-Bitness": '"64"',
"Sec-Ch-Ua-Full-Version": '"143.0.3650.96"',
"Sec-Ch-Ua-Full-Version-List": '"Microsoft Edge";v="143.0.3650.96", "Chromium";v="143.0.7499.147", "Not A(Brand";v="24.0.0.0"',
"Sec-Ch-Ua-Mobile": self.fp["sec-ch-ua-mobile"],
"Sec-Ch-Ua-Model": '""',
"Sec-Ch-Ua-Platform": self.fp["sec-ch-ua-platform"],
"Sec-Ch-Ua-Platform-Version": '"19.0.0"',
"Sec-Fetch-Dest": "empty",
"Sec-Fetch-Mode": "cors",
"Sec-Fetch-Site": "same-origin",
"OAI-Device-Id": self.device_id,
"OAI-Session-Id": self.session_id,
"OAI-Language": "zh-CN",
"OAI-Client-Version": self.client_version,
"OAI-Client-Build-Number": self.client_build_number,
})
if self.access_token:
self.session.headers["Authorization"] = f"Bearer {self.access_token}"
def close(self) -> None:
if getattr(self, "_closed", False):
return
self._closed = True
session = getattr(self, "session", None)
if session:
try:
session.close()
except Exception:
pass
def __del__(self):
self.close()
def __enter__(self):
return self
def __exit__(self, *args):
self.close()
return False
def _build_fp(self) -> Dict[str, str]:
account = self.account
raw_fp = account.get("fp")
fp = {str(k).lower(): str(v) for k, v in raw_fp.items()} if isinstance(raw_fp, dict) else {}
for key in (
"user-agent",
"impersonate",
"oai-device-id",
"oai-session-id",
"sec-ch-ua",
"sec-ch-ua-mobile",
"sec-ch-ua-platform",
):
value = str(account.get(key) or "").strip()
if value:
fp[key] = value
fp.setdefault(
"user-agent",
"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
"(KHTML, like Gecko) Chrome/143.0.0.0 Safari/537.36 Edg/143.0.0.0",
)
fp.setdefault("impersonate", "chrome110")
fp.setdefault("oai-device-id", new_uuid())
fp.setdefault("oai-session-id", new_uuid())
fp.setdefault("sec-ch-ua", '"Microsoft Edge";v="143", "Chromium";v="143", "Not A(Brand";v="24"')
fp.setdefault("sec-ch-ua-mobile", "?0")
fp.setdefault("sec-ch-ua-platform", '"Windows"')
return fp
def _headers(self, path: str, extra: Optional[Dict[str, str]] = None) -> Dict[str, str]:
"""构造请求头，并补上 web 端要求的 target path/route。"""
headers = dict(self.session.headers)
headers["X-OpenAI-Target-Path"] = path
headers["X-OpenAI-Target-Route"] = path
if extra:
headers.update(extra)
return headers
@staticmethod
def _extract_quota_and_restore_at(limits_progress: list[Any]) -> tuple[int, str | None]:
for item in limits_progress:
if isinstance(item, dict) and item.get("feature_name") == "image_gen":
return int(item.get("remaining") or 0), str(item.get("reset_after") or "") or None
return 0, None
def _raise_on_error(self, response: Any, path: str) -> None:
if response.status_code == 401:
raise InvalidAccessTokenError(f"token invalidated ({path})")
raise RuntimeError(f"{path} failed: HTTP {response.status_code}")
def _get_me(self) -> Dict[str, Any]:
path = "/backend-api/me"
response = self.session.get(self.base_url + path, headers=self._headers(path), timeout=20)
if response.status_code != 200:
self._raise_on_error(response, path)
return response.json()
def _get_conversation_init(self) -> Dict[str, Any]:
path = "/backend-api/conversation/init"
response = self.session.post(
self.base_url + path,
headers=self._headers(path, {"Content-Type": "application/json"}),
json={
"gizmo_id": None,
"requested_default_model": None,
"conversation_id": None,
"timezone_offset_min": -480,
},
timeout=20,
)
if response.status_code != 200:
self._raise_on_error(response, path)
return response.json()
def _get_default_account(self) -> Dict[str, Any]:
path = "/backend-api/accounts/check/v4-2023-04-27"
response = self.session.get(self.base_url + path + "?timezone_offset_min=-480", headers=self._headers(path),
timeout=20)
if response.status_code != 200:
self._raise_on_error(response, path)
payload = response.json()
default_account = ((payload.get("accounts") or {}).get("default") or {}).get("account") or {}
logger.debug({
"event": "backend_user_info_account_payload",
"plan_type": default_account.get("plan_type"),
"account_user_role": default_account.get("account_user_role"),
"account_id": default_account.get("account_id"),
"is_deactivated": default_account.get("is_deactivated"),
"has_active_subscription": (payload.get("accounts") or {}).get("default", {}).get("entitlement", {}).get("has_active_subscription"),
"subscription_plan": (payload.get("accounts") or {}).get("default", {}).get("entitlement", {}).get("subscription_plan"),
})
return default_account
def get_user_info(self) -> Dict[str, Any]:
"""获取当前 token 的账号信息。"""
if not self.access_token:
raise RuntimeError("access_token is required")
executor = ThreadPoolExecutor(max_workers=3)
try:
me_future = executor.submit(self._get_me)
init_future = executor.submit(self._get_conversation_init)
account_future = executor.submit(self._get_default_account)
me_payload, init_payload, default_account = me_future.result(), init_future.result(), account_future.result()
except (KeyboardInterrupt, SystemExit):
executor.shutdown(wait=False, cancel_futures=True)
raise
except BaseException:
executor.shutdown(wait=False, cancel_futures=True)
raise
else:
executor.shutdown(wait=True, cancel_futures=True)
plan_type = str(default_account.get("plan_type") or "free")
limits_progress = init_payload.get("limits_progress")
limits_progress = limits_progress if isinstance(limits_progress, list) else []
quota, restore_at = self._extract_quota_and_restore_at(limits_progress)
result = {
"email": me_payload.get("email"),
"user_id": me_payload.get("id"),
"type": plan_type,
"quota": quota,
"limits_progress": limits_progress,
"default_model_slug": init_payload.get("default_model_slug"),
"restore_at": restore_at,
"status": "限流" if quota == 0 else "正常",
}
logger.debug({
"event": "backend_user_info_result",
"email": result.get("email"),
"user_id": result.get("user_id"),
"type": result.get("type"),
"quota": result.get("quota"),
"default_model_slug": result.get("default_model_slug"),
"restore_at": result.get("restore_at"),
"status": result.get("status"),
})
return result
def _bootstrap_headers(self) -> Dict[str, str]:
"""构造首页预热请求头。"""
return {
"User-Agent": self.user_agent,
"Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,*/*;q=0.8",
"Accept-Language": "zh-CN,zh;q=0.9,en;q=0.8",
"Sec-Ch-Ua": self.session.headers["Sec-Ch-Ua"],
"Sec-Ch-Ua-Mobile": self.session.headers["Sec-Ch-Ua-Mobile"],
"Sec-Ch-Ua-Platform": self.session.headers["Sec-Ch-Ua-Platform"],
"Sec-Fetch-Dest": "document",
"Sec-Fetch-Mode": "navigate",
"Sec-Fetch-Site": "none",
"Sec-Fetch-User": "?1",
"Upgrade-Insecure-Requests": "1",
}
def _build_requirements(self, data: Dict[str, Any], source_p: str = "") -> ChatRequirements:
"""把 sentinel 响应整理成后续对话需要的 token 集合。"""
if (data.get("arkose") or {}).get("required"):
raise RuntimeError("chat requirements requires arkose token, which is not implemented")
proof_token = ""
proof_info = data.get("proofofwork") or {}
if proof_info.get("required"):
proof_token = build_proof_token(
proof_info.get("seed", ""),
proof_info.get("difficulty", ""),
self.user_agent,
script_sources=self.pow_script_sources,
data_build=self.pow_data_build,
)
turnstile_token = ""
turnstile_info = data.get("turnstile") or {}
if turnstile_info.get("required") and turnstile_info.get("dx"):
turnstile_token = solve_turnstile_token(turnstile_info["dx"], source_p) or ""
return ChatRequirements(
token=data.get("token", ""),
proof_token=proof_token,
turnstile_token=turnstile_token,
so_token=data.get("so_token", ""),
raw_finalize=data,
)
def _conversation_headers(self, path: str, requirements: ChatRequirements) -> Dict[str, str]:
"""根据当前 requirements 构造对话 SSE 请求头。"""
headers = {
"Accept": "text/event-stream",
"Content-Type": "application/json",
"OpenAI-Sentinel-Chat-Requirements-Token": requirements.token,
}
if requirements.proof_token:
headers["OpenAI-Sentinel-Proof-Token"] = requirements.proof_token
if requirements.turnstile_token:
headers["OpenAI-Sentinel-Turnstile-Token"] = requirements.turnstile_token
if requirements.so_token:
headers["OpenAI-Sentinel-SO-Token"] = requirements.so_token
return self._headers(path, headers)
def _api_messages_to_conversation_messages(self, messages: list[Dict[str, Any]]) -> list[Dict[str, Any]]:
"""把标准 chat messages 转成 web conversation 所需的 messages。"""
conversation_messages = []
for item in messages:
role = item.get("role", "user")
content = item.get("content", "")
if isinstance(content, str):
conversation_messages.append({
"id": new_uuid(),
"author": {"role": role},
"content": {"content_type": "text", "parts": [content]},
})
continue
if not isinstance(content, list):
raise RuntimeError("only string or list message content is supported")
text_parts: list[str] = []
image_inputs: list[tuple[bytes, str]] = []
for part in content:
if not isinstance(part, dict):
continue
part_type = str(part.get("type") or "")
if part_type == "text":
text_parts.append(str(part.get("text") or ""))
elif part_type == "image":
data = part.get("data")
mime = str(part.get("mime") or "image/png")
if isinstance(data, (bytes, bytearray)):
image_inputs.append((bytes(data), mime))
if not image_inputs:
conversation_messages.append({
"id": new_uuid(),
"author": {"role": role},
"content": {"content_type": "text", "parts": ["".join(text_parts)]},
})
continue
if not self.access_token:
raise RuntimeError("authenticated upstream account required for image input")
uploaded: list[Dict[str, Any]] = []
for idx, (data, mime) in enumerate(image_inputs, start=1):
ext_part = mime.split("/", 1)[1].split("+")[0] if "/" in mime else "png"
extension = "jpg" if ext_part == "jpeg" else (ext_part or "png")
b64 = base64.b64encode(data).decode("ascii")
uploaded.append(self._upload_image(f"data:{mime};base64,{b64}", f"image_{idx}.{extension}"))
parts: list[Any] = []
for ref in uploaded:
parts.append({
"content_type": "image_asset_pointer",
"asset_pointer": f"file-service://{ref['file_id']}",
"width": ref["width"],
"height": ref["height"],
"size_bytes": ref["file_size"],
})
text = "".join(text_parts)
if text:
parts.append(text)
conversation_messages.append({
"id": new_uuid(),
"author": {"role": role},
"content": {"content_type": "multimodal_text", "parts": parts},
"metadata": {
"attachments": [{
"id": ref["file_id"],
"mimeType": ref["mime_type"],
"name": ref["file_name"],
"size": ref["file_size"],
"width": ref["width"],
"height": ref["height"],
} for ref in uploaded],
},
})
return conversation_messages
@staticmethod
def _normalize_thinking_effort(value: str) -> str:
normalized = str(value or "").strip().lower()
if normalized in {"", "none"}:
return ""
if normalized in {"low", "medium", "high"}:
return normalized
if normalized in {"xhigh", "extended"}:
return "extended"
return ""
def _conversation_payload(
self,
messages: list[Dict[str, Any]],
model: str,
timezone: str,
thinking_effort: str = "",
) -> Dict[str, Any]:
"""把标准 messages 构造成 web 对话请求体。"""
payload = {
"action": "next",
"messages": self._api_messages_to_conversation_messages(messages),
"model": model,
"parent_message_id": new_uuid(),
"conversation_mode": {"kind": "primary_assistant"},
"conversation_origin": None,
"force_paragen": False,
"force_paragen_model_slug": "",
"force_rate_limit": False,
"force_use_sse": True,
"history_and_training_disabled": True,
"reset_rate_limits": False,
"suggestions": [],
"supported_encodings": [],
"system_hints": [],
"timezone": timezone,
"timezone_offset_min": -480,
"variant_purpose": "comparison_implicit",
"websocket_request_id": new_uuid(),
"client_contextual_info": {
"is_dark_mode": False,
"time_since_loaded": 120,
"page_height": 900,
"page_width": 1400,
"pixel_ratio": 2,
"screen_height": 1440,
"screen_width": 2560,
},
}
normalized_effort = self._normalize_thinking_effort(thinking_effort)
if normalized_effort:
payload["thinking_effort"] = normalized_effort
return payload
def _image_model_slug(self, model: str) -> str:
"""把标准图片模型名映射到底层 model slug。"""
_, base_model = split_image_model(model)
if not base_model:
return "auto"
if base_model == "gpt-image-2":
return "gpt-5-3"
if base_model == CODEX_IMAGE_MODEL:
return base_model
return "auto"
def _image_headers(self, path: str, requirements: ChatRequirements, conduit_token: str = "", accept: str = "*/*") -> \
Dict[str, str]:
"""构造图片链路请求头。"""
headers = {
"Content-Type": "application/json",
"Accept": accept,
"OpenAI-Sentinel-Chat-Requirements-Token": requirements.token,
}
if requirements.proof_token:
headers["OpenAI-Sentinel-Proof-Token"] = requirements.proof_token
if conduit_token:
headers["X-Conduit-Token"] = conduit_token
if accept == "text/event-stream":
headers["X-Oai-Turn-Trace-Id"] = new_uuid()
return self._headers(path, headers)
def _codex_responses_headers(self) -> Dict[str, str]:
return {
"Authorization": f"Bearer {self.access_token}",
"Content-Type": "application/json",
}
def _ensure_codex_source_account(self) -> None:
account = account_service.get_account(self.access_token)
source_type = str((account or {}).get("source_type") or "web").strip().lower()
if source_type != "codex":
raise RuntimeError("codex responses endpoint requires a codex source account")
@staticmethod
def _codex_image_input(prompt: str, images: list[str]) -> list[Dict[str, Any]]:
content: list[Dict[str, Any]] = [{"type": "input_text", "text": prompt}]
for image in images:
payload = image if image.startswith("data:image/") else f"data:image/png;base64,{image}"
content.append({"type": "input_image", "image_url": payload})
return [{"role": "user", "content": content}]
@staticmethod
def _codex_body_preview(body: Any, limit: int = 4000) -> str:
if isinstance(body, (dict, list)):
try:
text = json.dumps(body, ensure_ascii=False)
except Exception:
text = repr(body)
else:
text = str(body or "")
return text if len(text) <= limit else text[:limit] + "...[truncated]"
@staticmethod
def _codex_event_image_result_lengths(value: Any) -> list[int]:
if isinstance(value, dict):
lengths: list[int] = []
if value.get("type") == "image_generation_call" and isinstance(value.get("result"), str):
lengths.append(len(value["result"]))
for item in value.values():
lengths.extend(OpenAIBackendAPI._codex_event_image_result_lengths(item))
return lengths
if isinstance(value, list):
lengths: list[int] = []
for item in value:
lengths.extend(OpenAIBackendAPI._codex_event_image_result_lengths(item))
return lengths
return []
@staticmethod
def _codex_event_summary(event: Dict[str, Any]) -> Dict[str, Any]:
summary: Dict[str, Any] = {
"type": str(event.get("type") or ""),
"keys": list(event.keys())[:30],
}
for key in ("id", "status", "sequence_number", "response_id", "item_id", "output_index", "content_index"):
value = event.get(key)
if isinstance(value, (str, int, float, bool)) or value is None:
summary[key] = value
for key in ("response", "item", "output"):
value = event.get(key)
if isinstance(value, dict):
summary[f"{key}_type"] = value.get("type")
summary[f"{key}_status"] = value.get("status")
summary[f"{key}_keys"] = list(value.keys())[:30]
elif isinstance(value, list):
summary[f"{key}_len"] = len(value)
summary[f"{key}_types"] = [
item.get("type") for item in value[:10] if isinstance(item, dict)
]
error = event.get("error")
if isinstance(error, dict):
summary["error"] = {
key: error.get(key)
for key in ("type", "code", "message")
if error.get(key) is not None
}
delta = event.get("delta")
if isinstance(delta, str):
summary["delta_len"] = len(delta)
summary["delta_preview"] = delta[:200]
result_lengths = OpenAIBackendAPI._codex_event_image_result_lengths(event)
if result_lengths:
summary["image_result_lengths"] = result_lengths[:10]
return summary
def _log_codex_response_failure(
self,
path: str,
status_code: int,
headers: Any,
payload: Dict[str, Any],
body: Any,
) -> None:
request_headers = self._codex_responses_headers()
safe_request_headers = {
key: value for key, value in request_headers.items() if key.lower() != "authorization"
}
response_headers = dict(headers.items()) if hasattr(headers, "items") else dict(headers or {})
tool = ((payload.get("tools") or [{}])[0]) if isinstance(payload.get("tools"), list) else {}
logger.warning({
"event": "codex_responses_http_error",
"path": path,
"status_code": status_code,
"request": {
"model": payload.get("model"),
"tool_model": tool.get("model"),
"tool_action": tool.get("action"),
"size": tool.get("size"),
"quality": tool.get("quality"),
"image_input_count": max(len((payload.get("input") or [{}])[0].get("content") or []) - 1, 0),
"prompt_preview": self._codex_body_preview(
(((payload.get("input") or [{}])[0].get("content") or [{}])[0].get("text") or ""),
500,
),
"headers": safe_request_headers,
},
"response": {
"headers": response_headers,
"body_preview": self._codex_body_preview(body),
},
})
@staticmethod
def _iter_codex_response_events(raw: Any) -> Iterator[Dict[str, Any]]:
content_type = str(raw.headers.get("content-type") or "").lower()
text = raw.read().decode("utf-8", "replace")
status_code = getattr(raw, "status", None)
parse_errors: list[str] = []
events: list[Dict[str, Any]] = []
if "application/json" in content_type:
try:
data = json.loads(text)
if isinstance(data, dict):
events.append(data)
except Exception as exc:
parse_errors.append(str(exc))
else:
lines: list[str] = []
for line in text.splitlines() + [""]:
if not line:
if lines:
payload_text = "\n".join(lines).strip()
if payload_text and payload_text != "[DONE]":
try:
data = json.loads(payload_text)
except Exception as exc:
parse_errors.append(str(exc))
data = None
if isinstance(data, dict):
events.append(data)
lines = []
elif line.startswith("data:"):
lines.append(line[5:].lstrip())
event_types: Dict[str, int] = {}
image_result_lengths: list[int] = []
for event in events:
event_type = str(event.get("type") or "<missing>")
event_types[event_type] = event_types.get(event_type, 0) + 1
image_result_lengths.extend(OpenAIBackendAPI._codex_event_image_result_lengths(event))
logger.info({
"event": "codex_responses_response_debug",
"status_code": status_code,
"content_type": content_type,
"response_text_len": len(text),
"event_count": len(events),
"event_types": event_types,
"image_result_lengths": image_result_lengths[:10],
"parse_error_count": len(parse_errors),
"parse_errors": parse_errors[:5],
"event_summaries": [OpenAIBackendAPI._codex_event_summary(event) for event in events[:30]],
"event_previews": [
OpenAIBackendAPI._codex_body_preview(event, 1500)
for event in events[:10]
] if not image_result_lengths else [],
"body_preview": text[:1000] if not events else "",
})
for event in events:
yield event
def iter_codex_image_response_events(
self,
prompt: str,
images: list[str] | None = None,
size: str | None = None,
quality: str = "auto",
) -> Iterator[Dict[str, Any]]:
if not self.access_token:
raise RuntimeError("access_token is required for codex image endpoints")
self._ensure_codex_source_account()
path = "/backend-api/codex/responses"
payload = {
"model": CODEX_RESPONSES_MODEL,
"instructions": CODEX_RESPONSES_INSTRUCTIONS,
"store": False,
"input": self._codex_image_input(prompt, images or []),
"tools": [{
"type": "image_generation",
"model": "gpt-image-2",
"action": "edit" if images else "generate",
"size": str(size or "1024x1024"),
"quality": str(quality or "auto"),
"output_format": "png",
}],
"tool_choice": {"type": "image_generation"},
"stream": True,
}
request = urllib.request.Request(
self.base_url + path,
json.dumps(payload).encode(),
self._codex_responses_headers(),
method="POST",
)
account = account_service.get_account(self.access_token) or {}
token_payload = account_service._decode_jwt_payload(self.access_token)
auth_claim = token_payload.get("https://api.openai.com/auth")
auth_claim = auth_claim if isinstance(auth_claim, dict) else {}
tool = payload["tools"][0]
logger.info({
"event": "codex_responses_request_debug",
"url": self.base_url + path,
"transport": "urllib.request",
"timeout_secs": 1200,
"account_email": str(account.get("email") or "").strip(),
"source_type": str(account.get("source_type") or "").strip(),
"account_type": str(account.get("type") or "").strip(),
"token_claims": {
"jti": token_payload.get("jti"),
"iat": token_payload.get("iat"),
"exp": token_payload.get("exp"),
"client_id": token_payload.get("client_id"),
"chatgpt_account_id": auth_claim.get("chatgpt_account_id"),
"chatgpt_plan_type": auth_claim.get("chatgpt_plan_type"),
"localhost": auth_claim.get("localhost"),
},
"request": {
"model": payload.get("model"),
"tool_model": tool.get("model"),
"tool_action": tool.get("action"),
"size": tool.get("size"),
"quality": tool.get("quality"),
"output_format": tool.get("output_format"),
"stream": payload.get("stream"),
"image_input_count": max(len((payload.get("input") or [{}])[0].get("content") or []) - 1, 0),
"prompt_preview": self._codex_body_preview(
(((payload.get("input") or [{}])[0].get("content") or [{}])[0].get("text") or ""),
500,
),
},
"headers": {
key: value for key, value in self._codex_responses_headers().items()
if key.lower() != "authorization"
},
})
try:
with urllib.request.urlopen(request, timeout=1200) as raw:
yield from self._iter_codex_response_events(raw)
except urllib.error.HTTPError as error:
body_text = error.read().decode("utf-8", "replace")
body: Any = body_text
try:
body = json.loads(body_text)
except Exception:
pass
self._log_codex_response_failure(path, error.code, error.headers, payload, body)
retry_after_header = error.headers.get("Retry-After") if error.headers else None
retry_after = int(retry_after_header) if str(retry_after_header or "").isdigit() else None
raise UpstreamHTTPError(path, error.code, body, retry_after=retry_after) from error
def _prepare_image_conversation(self, prompt: str, requirements: ChatRequirements, model: str) -> str:
"""为图片生成准备 conduit token。"""
path = "/backend-api/f/conversation/prepare"
payload = {
"action": "next",
"fork_from_shared_post": False,
"parent_message_id": new_uuid(),
"model": self._image_model_slug(model),
"client_prepare_state": "success",
"timezone_offset_min": -480,
"timezone": "Asia/Shanghai",
"conversation_mode": {"kind": "primary_assistant"},
"system_hints": ["picture_v2"],
"partial_query": {
"id": new_uuid(),
"author": {"role": "user"},
"content": {"content_type": "text", "parts": [prompt]},
},
"supports_buffering": True,
"supported_encodings": ["v1"],
"client_contextual_info": {"app_name": "chatgpt.com"},
}
response = self.session.post(
self.base_url + path,
headers=self._image_headers(path, requirements),
json=payload,
timeout=60,
)
ensure_ok(response, path)
return response.json().get("conduit_token", "")
def _decode_image_base64(self, image: str) -> bytes:
"""把 base64 图片字符串或本地路径解码成二进制。"""
if (
image
and len(image) < 512
and not image.startswith("data:")
and "\n" not in image
and "\r" not in image
):
file_path = Path(os.path.expanduser(image))
if file_path.exists() and file_path.is_file():
return file_path.read_bytes()
payload = image.split(",", 1)[1] if image.startswith("data:") and "," in image else image
return base64.b64decode(payload)
def _upload_image(self, image: str, file_name: str = "image.png") -> Dict[str, Any]:
"""上传一张 base64 图片，返回底层文件元数据。"""
data = self._decode_image_base64(image)
if (
image
and len(image) < 512
and not image.startswith("data:")
and "\n" not in image
and "\r" not in image
):
candidate_path = Path(os.path.expanduser(image))
if candidate_path.exists() and candidate_path.is_file():
file_name = candidate_path.name
image = Image.open(BytesIO(data))
width, height = image.size
mime_type = Image.MIME.get(image.format, "image/png")
path = "/backend-api/files"
response = self.session.post(
self.base_url + path,
headers=self._headers(path, {"Content-Type": "application/json", "Accept": "application/json"}),
json={"file_name": file_name, "file_size": len(data), "use_case": "multimodal", "width": width,
"height": height},
timeout=60,
)
ensure_ok(response, path)
upload_meta = response.json()
response = self.session.put(
upload_meta["upload_url"],
headers={
"Content-Type": mime_type,
"x-ms-blob-type": "BlockBlob",
"x-ms-version": "2020-04-08",
"Origin": self.base_url,
"Referer": self.base_url + "/",
"User-Agent": self.user_agent,
"Accept": "application/json, text/plain, */*",
"Accept-Language": "en-US,en;q=0.8",
},
data=data,
timeout=120,
)
ensure_ok(response, "image_upload")
path = f"/backend-api/files/{upload_meta['file_id']}/uploaded"
response = self.session.post(
self.base_url + path,
headers=self._headers(path, {"Content-Type": "application/json", "Accept": "application/json"}),
data="{}",
timeout=60,
)
ensure_ok(response, path)
return {
"file_id": upload_meta["file_id"],
"file_name": file_name,
"file_size": len(data),
"mime_type": mime_type,
"width": width,
"height": height,
}
def _start_image_generation(self, prompt: str, requirements: ChatRequirements, conduit_token: str, model: str,
references: Optional[list[Dict[str, Any]]] = None) -> requests.Response:
"""启动图片生成或编辑的 SSE 请求。"""
references = references or []
parts = [{
"content_type": "image_asset_pointer",
"asset_pointer": f"file-service://{item['file_id']}",
"width": item["width"],
"height": item["height"],
"size_bytes": item["file_size"],
} for item in references]
parts.append(prompt)
content = {"content_type": "multimodal_text", "parts": parts} if references else {"content_type": "text",
"parts": [prompt]}
metadata = {
"developer_mode_connector_ids": [],
"selected_github_repos": [],
"selected_all_github_repos": False,
"system_hints": ["picture_v2"],
"serialization_metadata": {"custom_symbol_offsets": []},
}
if references:
metadata["attachments"] = [{
"id": item["file_id"],
"mimeType": item["mime_type"],
"name": item["file_name"],
"size": item["file_size"],
"width": item["width"],
"height": item["height"],
} for item in references]
payload = {
"action": "next",
"messages": [{
"id": new_uuid(),
"author": {"role": "user"},
"create_time": time.time(),
"content": content,
"metadata": metadata,
}],
"parent_message_id": new_uuid(),
"model": self._image_model_slug(model),
"client_prepare_state": "sent",
"timezone_offset_min": -480,
"timezone": "Asia/Shanghai",
"conversation_mode": {"kind": "primary_assistant"},
"enable_message_followups": True,
"system_hints": ["picture_v2"],
"supports_buffering": True,
"supported_encodings": ["v1"],
"client_contextual_info": {
"is_dark_mode": False,
"time_since_loaded": 1200,
"page_height": 1072,
"page_width": 1724,
"pixel_ratio": 1.2,
"screen_height": 1440,
"screen_width": 2560,
"app_name": "chatgpt.com",
},
"paragen_cot_summary_display_override": "allow",
"force_parallel_switch": "auto",
}
path = "/backend-api/f/conversation"
response = self.session.post(
self.base_url + path,
headers=self._image_headers(path, requirements, conduit_token, "text/event-stream"),
json=payload,
timeout=300,
stream=True,
)
ensure_ok(response, path)
return response
def _get_conversation(self, conversation_id: str) -> Dict[str, Any]:
"""获取完整 conversation 详情。"""
path = f"/backend-api/conversation/{conversation_id}"
response = self.session.get(self.base_url + path, headers=self._headers(path, {"Accept": "application/json"}),
timeout=60)
ensure_ok(response, path)
return response.json()
def delete_conversation(self, conversation_id: str) -> Dict[str, Any]:
"""删除本地对话记录。"""
path = f"/backend-api/conversation/{conversation_id}"
headers = self._headers(path, {
"Accept": "*/*",
"Content-Type": "application/json",
"Referer": f"{self.base_url}/c/{conversation_id}",
"X-OpenAI-Target-Route": "/backend-api/conversation/{conversation_id}",
})
response = self.session.patch(
self.base_url + path,
headers=headers,
json={"is_visible": False},
timeout=60,
)
ensure_ok(response, path)
return response.json()
def _list_recent_conversations(self, limit: int = 5, timeout_secs: float = 10.0) -> list[Dict[str, Any]]:
"""列出最近的对话列表，按更新时间倒序。
当 SSE 流太短导致 conversation_id 丢失时，可以通过此方法
查找最近创建的对话来恢复 conversation_id。
"""
path = f"/backend-api/conversations?offset=0&limit={limit}&order=updated&conversation_filter=all"
try:
response = self.session.get(
self.base_url + path,
headers=self._headers(path, {"Accept": "application/json"}),
timeout=timeout_secs,
)
ensure_ok(response, path)
data = response.json()
return data.get("items") or data.get("conversations") or []
except Exception as exc:
logger.debug({"event": "list_conversations_failed", "error": str(exc)})
return []
def find_conversation_by_prompt(self, prompt: str, started_at: float, timeout_secs: float = 10.0) -> str:
"""根据 prompt 和开始时间，从最近对话列表中查找匹配的 conversation_id。
当 SSE 流太短导致 conversation_id 丢失时，使用此方法恢复。
通过对比 prompt 关键词和时间戳来匹配最可能的对话。
参数：
prompt: 用户输入的 prompt 文本
started_at: 请求开始的时间戳（epoch seconds）
timeout_secs: 请求超时秒数
返回：
匹配的 conversation_id，如果未找到返回空字符串
"""
items = self._list_recent_conversations(limit=10, timeout_secs=timeout_secs)
if not items:
return ""
# 筛选在 started_at 之前或附近创建的对话（最多往前 5 分钟）
# ChatGPT 的 updated_at 通常晚于实际请求时间
prompt_lower = str(prompt or "").lower().strip()
best_match = ""
best_score = 0.0
for item in items:
# item 可能是完整的 conversation 对象或摘要
conv_id = str(item.get("id") or item.get("conversation_id") or "")
if not conv_id:
continue
# 检查时间范围：对话的 updated_at 应该在请求开始时间之后（或附近）
updated_at = float(item.get("update_time") or item.get("updated_at") or 0)
if updated_at and started_at and (updated_at < started_at - 30 or updated_at > started_at + 600):
continue
# 匹配 prompt 关键词
title = str(item.get("title") or "").lower()
# 计算匹配分数
score = 0.0
if prompt_lower and title:
# 简单的关键词匹配
prompt_words = set(prompt_lower.split())
title_words = set(title.split())
common = prompt_words & title_words
if common:
score = len(common) / max(len(prompt_words), 1)
# 图生图通常标题为 "Image" 开头
if title.startswith("image"):
score += 0.3
if score > best_score:
best_score = score
best_match = conv_id
if best_match and best_score > 0.1:
logger.info({
"event": "conversation_prompt_match_found",
"conversation_id": best_match,
"match_score": round(best_score, 2),
})
return best_match
# 如果没有标题匹配，返回最新的对话（时间最近的）
for item in items:
conv_id = str(item.get("id") or item.get("conversation_id") or "")
updated_at = float(item.get("update_time") or item.get("updated_at") or 0)
if conv_id and updated_at and started_at and updated_at >= started_at - 30:
logger.info({
"event": "conversation_latest_match",
"conversation_id": conv_id,
"updated_at": updated_at,
})
return conv_id
return ""
@staticmethod
def _editable_prompt(fixed_prompt: str, user_prompt_text: str) -> str:
extra = str(user_prompt_text or "").strip()
return fixed_prompt if not extra else fixed_prompt + "\n\n以下是用户补充需求，请直接结合执行：\n" + extra
def export_ppt_zip(
self,
base64_images: list[str] | None,
prompt: str,
output_dir: str | Path = EDITABLE_FILE_PPT_OUTPUT_DIR,
timeout_secs: float = EDITABLE_FILE_TIMEOUT_SECS,
poll_interval_secs: float = EDITABLE_FILE_POLL_INTERVAL_SECS,
) -> EditableFileExportResult:
return self._export_editable_file_zip(
base64_images or [],
self._editable_prompt(EDITABLE_FILE_PPT_PROMPT, prompt),
output_dir,
primary_label="ppt",
primary_suffixes=(".ppt", ".pptx"),
primary_mime_types=EDITABLE_PPT_MIME_TYPES,
primary_mime_keywords=("presentationml.presentation", "ms-powerpoint"),
primary_default_extension=".pptx",
export_file_re=EDITABLE_PPT_EXPORT_FILE_RE,
timeout_secs=timeout_secs,
poll_interval_secs=poll_interval_secs,
)
def export_psd_zip(
self,
base64_images: list[str],
prompt: str,
output_dir: str | Path = EDITABLE_FILE_PSD_OUTPUT_DIR,
timeout_secs: float = EDITABLE_FILE_TIMEOUT_SECS,
poll_interval_secs: float = EDITABLE_FILE_POLL_INTERVAL_SECS,
) -> EditableFileExportResult:
if not base64_images:
raise ValueError("base64_images is empty")
return self._export_editable_file_zip(
base64_images,
self._editable_prompt(EDITABLE_FILE_PSD_PROMPT, prompt),
output_dir,
primary_label="psd",
primary_suffixes=(".psd",),
primary_mime_types=EDITABLE_PSD_MIME_TYPES,
primary_mime_keywords=("photoshop",),
primary_default_extension=".psd",
export_file_re=EDITABLE_PSD_EXPORT_FILE_RE,
timeout_secs=timeout_secs,
poll_interval_secs=poll_interval_secs,
)
def _export_editable_file_zip(
self,
base64_images: list[str],
prompt: str,
output_dir: str | Path,
*,
primary_label: str,
primary_suffixes: tuple[str, ...],
primary_mime_types: set[str],
primary_mime_keywords: tuple[str, ...],
primary_default_extension: str,
export_file_re: re.Pattern[str],
timeout_secs: float,
poll_interval_secs: float,
) -> EditableFileExportResult:
if not self.access_token:
raise RuntimeError("access_token is required for editable file export")
self.client_version = EDITABLE_FILE_CLIENT_VERSION
self.client_build_number = EDITABLE_FILE_CLIENT_BUILD_NUMBER
self.session.headers["OAI-Client-Version"] = EDITABLE_FILE_CLIENT_VERSION
self.session.headers["OAI-Client-Build-Number"] = EDITABLE_FILE_CLIENT_BUILD_NUMBER
output_path = Path(output_dir).expanduser().resolve()
output_path.mkdir(parents=True, exist_ok=True)
uploaded = [self._upload_editable_base64_image(item, index) for index, item in enumerate(base64_images, start=1)]
conduit_token = self._prepare_editable_conversation(prompt, [item["mime_type"] for item in uploaded])
conversation_id = self._run_editable_conversation(prompt, uploaded, conduit_token)
artifacts = self._wait_editable_output_artifacts(
conversation_id,
primary_label,
primary_suffixes,
primary_mime_types,
primary_mime_keywords,
export_file_re,
timeout_secs,
poll_interval_secs,
)
downloaded = [self._download_editable_artifact(
conversation_id,
item,
output_path,
primary_mime_types,
primary_mime_keywords,
primary_default_extension,
) for item in artifacts]
primary_path = next((item for item in downloaded if item.suffix.lower() in primary_suffixes), None)
zip_path = next((item for item in downloaded if item.suffix.lower() == ".zip"), None)
if not primary_path or not zip_path:
raise RuntimeError(f"download finished but did not get both {primary_label} and zip files: {downloaded}")
return EditableFileExportResult(conversation_id=conversation_id, primary_path=primary_path, zip_path=zip_path)
def _upload_editable_base64_image(self, base64_image: str, index: int) -> Dict[str, Any]:
data, file_name, mime_type, width, height = self._decode_editable_base64_image(base64_image, index)
path = "/backend-api/files"
response = self.session.post(
self.base_url + path,
headers=self._headers(path, {"Accept": "*/*", "Content-Type": "application/json"}),
json={
"file_name": file_name,
"file_size": len(data),
"use_case": "multimodal",
"timezone_offset_min": -480,
"reset_rate_limits": False,
"store_in_library": True,
"library_persistence_mode": "opportunistic",
},
timeout=60,
)
ensure_ok(response, path)
payload = response.json()
upload_url = str(payload.get("upload_url") or "")
file_id = str(payload.get("file_id") or "")
if not upload_url or not file_id:
raise RuntimeError(f"invalid upload response: {payload}")
response = self.session.put(
upload_url,
headers={
"Content-Type": mime_type,
"x-ms-blob-type": "BlockBlob",
"x-ms-version": "2020-04-08",
"Origin": self.base_url,
"Referer": self.base_url + "/",
"User-Agent": self.user_agent,
"Accept": "application/json, text/plain, */*",
"Accept-Language": "en-US,en;q=0.8",
},
data=data,
timeout=120,
)
ensure_ok(response, "image_upload")
path = f"/backend-api/files/{file_id}/uploaded"
response = self.session.post(
self.base_url + path,
headers=self._headers(path, {"Accept": "*/*", "Content-Type": "application/json"}),
data="{}",
timeout=60,
)
ensure_ok(response, path)
return {
"file_id": file_id,
"library_file_id": str(payload.get("library_file_id") or ""),
"file_name": file_name,
"file_size": len(data),
"mime_type": mime_type,
"width": width,
"height": height,
}
def _decode_editable_base64_image(self, base64_image: str, index: int) -> tuple[bytes, str, str, int, int]:
raw = str(base64_image or "").strip()
if not raw:
raise ValueError("base64 image is empty")
mime_type = ""
payload = raw
match = re.match(r"^data:([^;]+);base64,(.*)$", raw, re.IGNORECASE | re.DOTALL)
if match:
mime_type = str(match.group(1) or "").strip().lower()
payload = str(match.group(2) or "").strip()
data = base64.b64decode(payload)
image = Image.open(BytesIO(data))
image.load()
width, height = image.size
mime_type = Image.MIME.get(image.format, mime_type or "image/png")
extension = mimetypes.guess_extension(mime_type) or ".png"
return data, f"image_{index}{extension}", mime_type, width, height
def _prepare_editable_conversation(self, prompt: str, attachment_mime_types: list[str]) -> str:
path = "/backend-api/f/conversation/prepare"
payload: Dict[str, Any] = {
"action": "next",
"fork_from_shared_post": False,
"parent_message_id": "client-created-root",
"model": EDITABLE_FILE_MODEL,
"client_prepare_state": "success",
"timezone_offset_min": -480,
"timezone": "Asia/Shanghai",
"conversation_mode": {"kind": "primary_assistant"},
"system_hints": [],
"partial_query": {"id": new_uuid(), "author": {"role": "user"}, "content": {"content_type": "text", "parts": [prompt]}},
"supports_buffering": True,
"supported_encodings": ["v1"],
"client_contextual_info": {"app_name": "chatgpt.com"},
"thinking_effort": EDITABLE_FILE_THINKING_EFFORT,
}
if attachment_mime_types:
payload["attachment_mime_types"] = attachment_mime_types
response = self.session.post(
self.base_url + path,
headers=self._headers(path, {"Accept": "*/*", "Content-Type": "application/json", "X-Conduit-Token": "no-token"}),
json=payload,
timeout=60,
)
ensure_ok(response, path)
conduit_token = str(response.json().get("conduit_token") or "")
if not conduit_token:
raise RuntimeError(f"missing conduit_token: {response.text}")
return conduit_token
def _run_editable_conversation(self, prompt: str, uploaded: list[Dict[str, Any]], conduit_token: str) -> str:
self._bootstrap()
requirements = self._get_chat_requirements()
message: Dict[str, Any] = {"id": new_uuid(), "author": {"role": "user"}, "create_time": time.time()}
if uploaded:
parts = [{
"content_type": "image_asset_pointer",
"asset_pointer": f"sediment://{item['file_id']}",
"size_bytes": item["file_size"],
"width": item["width"],
"height": item["height"],
} for item in uploaded]
parts.append(prompt)
message["content"] = {"content_type": "multimodal_text", "parts": parts}
message["metadata"] = {
"attachments": [{
"id": item["file_id"],
"size": item["file_size"],
"name": item["file_name"],
"mime_type": item["mime_type"],
"width": item["width"],
"height": item["height"],
"source": "library",
"library_file_id": item["library_file_id"],
"is_big_paste": False,
} for item in uploaded],
"developer_mode_connector_ids": [],
"selected_sources": [],
"selected_github_repos": [],
"selected_all_github_repos": False,
"serialization_metadata": {"custom_symbol_offsets": []},
}
else:
message["content"] = {"content_type": "text", "parts": [prompt]}
path = "/backend-api/f/conversation"
response = self.session.post(
self.base_url + path,
headers=self._image_headers(path, requirements, conduit_token, "text/event-stream"),
json={
"action": "next",
"messages": [message],
"parent_message_id": "client-created-root",
"model": EDITABLE_FILE_MODEL,
"client_prepare_state": "sent",
"timezone_offset_min": -480,
"timezone": "Asia/Shanghai",
"conversation_mode": {"kind": "primary_assistant"},
"enable_message_followups": True,
"system_hints": [],
"supports_buffering": True,
"supported_encodings": ["v1"],
"client_contextual_info": {
"is_dark_mode": False,
"time_since_loaded": 401,
"page_height": 1138,
"page_width": 803,
"pixel_ratio": 2,
"screen_height": 1440,
"screen_width": 2560,
"app_name": "chatgpt.com",
},
"paragen_cot_summary_display_override": "allow",
"force_parallel_switch": "auto",
"thinking_effort": EDITABLE_FILE_THINKING_EFFORT,
},
timeout=300,
stream=True,
)
ensure_ok(response, path)
conversation_id = ""
try:
for payload in iter_sse_payloads(response):
if payload == "[DONE]":
break
conversation_id = conversation_id or self._find_editable_value(payload, "conversation_id")
finally:
response.close()
if not conversation_id:
raise RuntimeError("conversation_id not found in stream")
return conversation_id
def _wait_editable_output_artifacts(
self,
conversation_id: str,
primary_label: str,
primary_suffixes: tuple[str, ...],
primary_mime_types: set[str],
primary_mime_keywords: tuple[str, ...],
export_file_re: re.Pattern[str],
timeout_secs: float,
poll_interval_secs: float,
) -> list[EditableFileArtifact]:
deadline = time.time() + timeout_secs
while time.time() < deadline:
try:
conversation = self._get_editable_conversation_detail(conversation_id)
except UpstreamHTTPError as exc:
if exc.status_code in {404, 409, 423, 429, 500, 502, 503, 504}:
time.sleep(poll_interval_secs)
continue
raise
targeted = self._pick_editable_target_artifacts(
self._extract_editable_artifacts(conversation, export_file_re),
primary_suffixes,
primary_mime_types,
primary_mime_keywords,
)
if targeted:
return targeted
time.sleep(poll_interval_secs)
raise RuntimeError(f"timed out waiting for {primary_label}/zip outputs")
def _get_editable_conversation_detail(self, conversation_id: str) -> Dict[str, Any]:
path = f"/backend-api/conversation/{conversation_id}"
response = self.session.get(self.base_url + path, headers=self._editable_conversation_document_headers(path, conversation_id), timeout=60)
ensure_ok(response, path)
return response.json()
def _editable_browser_headers(self, path: str, conversation_id: str) -> Dict[str, str]:
headers = self._headers(path, {"Accept": "*/*"})
headers["Referer"] = f"{self.base_url}/c/{conversation_id}"
return headers
def _editable_conversation_document_headers(self, path: str, conversation_id: str) -> Dict[str, str]:
headers = self._editable_browser_headers(path, conversation_id)
headers["X-OpenAI-Target-Route"] = "/backend-api/conversation/{conversation_id}"
return headers
def _extract_editable_artifacts(self, conversation: Dict[str, Any], export_file_re: re.Pattern[str]) -> list[EditableFileArtifact]:
artifacts: dict[str, EditableFileArtifact] = {}
for node in sorted((conversation.get("mapping") or {}).values(), key=lambda item: float(((item or {}).get("message") or {}).get("create_time") or 0.0)):
message = (node or {}).get("message") or {}
message_id = str(message.get("id") or "")
author_role = str(((message.get("author") or {}).get("role") or "")).strip()
if author_role not in {"assistant", "tool"}:
continue
create_time = float(message.get("create_time") or 0.0)
message_text = self._editable_message_text(message)
for artifact in self._extract_editable_message_artifacts(message, message_id, author_role, create_time, export_file_re):
key = artifact.attachment_id or artifact.file_id or artifact.name or artifact.sandbox_path
if key:
artifacts[key] = self._merge_editable_artifact(artifacts.get(key), artifact)
for export_path in self._extract_editable_export_paths(message_text, export_file_re):
inferred = EditableFileArtifact(name=Path(export_path).name, create_time=create_time, author_role=author_role, sandbox_path=export_path, message_id=message_id)
artifacts[export_path] = self._merge_editable_artifact(artifacts.get(export_path), inferred)
return sorted(artifacts.values(), key=lambda item: item.create_time)
def _extract_editable_message_artifacts(
self,
message: Dict[str, Any],
message_id: str,
author_role: str,
create_time: float,
export_file_re: re.Pattern[str],
) -> list[EditableFileArtifact]:
artifacts: list[EditableFileArtifact] = []
for item in (message.get("metadata") or {}).get("attachments") or []:
artifact = self._editable_artifact_from_dict(item, message_id, author_role, create_time, export_file_re)
if artifact:
artifacts.append(artifact)
for obj in self._walk_search_dicts(message):
artifact = self._editable_artifact_from_dict(obj, message_id, author_role, create_time, export_file_re)
if artifact:
artifacts.append(artifact)
return artifacts
def _editable_artifact_from_dict(
self,
payload: Dict[str, Any],
message_id: str,
author_role: str,
create_time: float,
export_file_re: re.Pattern[str],
) -> EditableFileArtifact | None:
if not ({"id", "file_id", "asset_pointer", "name", "file_name", "filename", "mime_type", "mimeType"} & set(payload.keys())):
return None
attachment_id = self._match_editable_file_id(str(payload.get("id") or ""))
file_id = self._match_editable_file_id(str(payload.get("file_id") or ""))
name = self._sanitize_editable_filename(str(payload.get("name") or payload.get("file_name") or payload.get("filename") or payload.get("title") or "").strip())
mime_type = self._clean_editable_mime_type(payload.get("mime_type") or payload.get("mimeType") or "")
for asset_id in EDITABLE_ASSET_POINTER_RE.findall(str(payload.get("asset_pointer") or "")):
attachment_id = attachment_id or asset_id
file_id = file_id or asset_id
if not attachment_id or not file_id:
ids = self._extract_editable_file_ids(json.dumps(payload, ensure_ascii=False))
attachment_id = attachment_id or (ids[0] if ids else "")
file_id = file_id or (ids[0] if ids else "")
if not attachment_id and not file_id:
return None
return EditableFileArtifact(
attachment_id=attachment_id,
file_id=file_id,
name=name,
mime_type=mime_type,
create_time=create_time,
author_role=author_role,
sandbox_path=(self._extract_editable_export_paths(payload, export_file_re) or [""])[0],
message_id=message_id,
)
def _pick_editable_target_artifacts(
self,
artifacts: list[EditableFileArtifact],
primary_suffixes: tuple[str, ...],
primary_mime_types: set[str],
primary_mime_keywords: tuple[str, ...],
) -> list[EditableFileArtifact]:
primary = next((item for item in reversed(artifacts) if self._looks_like_editable_primary(item, primary_suffixes, primary_mime_types, primary_mime_keywords)), None)
zip_item = next((item for item in reversed(artifacts) if self._looks_like_editable_zip(item)), None)
return [primary, zip_item] if primary and zip_item else []
def _download_editable_artifact(
self,
conversation_id: str,
artifact: EditableFileArtifact,
output_dir: Path,
primary_mime_types: set[str],
primary_mime_keywords: tuple[str, ...],
primary_default_extension: str,
) -> Path:
download_url = self._resolve_editable_download_url(conversation_id, artifact)
if not download_url:
raise RuntimeError(f"download url not found for artifact: {artifact}")
response = self.session.get(download_url, timeout=300)
ensure_ok(response, "artifact_download")
content_type = self._clean_editable_mime_type(response.headers.get("Content-Type") or artifact.mime_type)
file_name = self._resolve_editable_output_name(artifact, response.url, response.headers.get("Content-Disposition"), content_type, primary_mime_types, primary_mime_keywords, primary_default_extension)
target_path = self._unique_editable_path(output_dir / file_name)
target_path.write_bytes(response.content)
return target_path
def _resolve_editable_download_url(self, conversation_id: str, artifact: EditableFileArtifact) -> str:
ids: list[str] = []
for item in (artifact.attachment_id, artifact.file_id):
if item and item not in ids:
ids.append(item)
if artifact.sandbox_path and artifact.message_id:
path = f"/backend-api/conversation/{conversation_id}/interpreter/download"
response = self.session.get(
self.base_url + path,
headers=self._editable_download_headers(path, conversation_id, "/backend-api/conversation/{conversation_id}/interpreter/download"),
params={"message_id": artifact.message_id, "sandbox_path": artifact.sandbox_path},
timeout=60,
)
if 200 <= response.status_code < 300:
url = self._download_url_from_response(response)
if url:
return url
for attachment_id in ids:
path = f"/backend-api/conversation/{conversation_id}/attachment/{attachment_id}/download"
response = self.session.get(
self.base_url + path,
headers=self._editable_download_headers(path, conversation_id, "/backend-api/conversation/{conversation_id}/attachment/{attachment_id}/download"),
timeout=60,
)
if 200 <= response.status_code < 300:
url = self._download_url_from_response(response)
if url:
return url
for file_id in ids:
path = f"/backend-api/files/download/{file_id}"
response = self.session.get(
self.base_url + path,
headers=self._editable_download_headers(path, conversation_id, "/backend-api/files/download/{file_id}"),
params={"post_id": "", "inline": "false"},
timeout=60,
)
if 200 <= response.status_code < 300:
url = self._download_url_from_response(response)
if url:
return url
for file_id in ids:
path = f"/backend-api/files/{file_id}/download"
response = self.session.get(
self.base_url + path,
headers=self._editable_download_headers(path, conversation_id, "/backend-api/files/download/{file_id}"),
timeout=60,
)
if 200 <= response.status_code < 300:
url = self._download_url_from_response(response)
if url:
return url
return ""
def _editable_download_headers(self, path: str, conversation_id: str, route: str) -> Dict[str, str]:
headers = self._editable_browser_headers(path, conversation_id)
headers["X-OpenAI-Target-Route"] = route
return headers
@staticmethod
def _download_url_from_response(response: Any) -> str:
try:
payload = response.json()
except Exception:
payload = {}
return str(payload.get("download_url") or payload.get("url") or "")
def _resolve_editable_output_name(
self,
artifact: EditableFileArtifact,
final_url: str,
content_disposition: str | None,
content_type: str,
primary_mime_types: set[str],
primary_mime_keywords: tuple[str, ...],
primary_default_extension: str,
) -> str:
file_name = self._sanitize_editable_filename(artifact.name)
if not file_name and artifact.sandbox_path:
file_name = self._sanitize_editable_filename(Path(artifact.sandbox_path).name)
if not file_name:
file_name = self._sanitize_editable_filename(self._editable_filename_from_content_disposition(content_disposition or ""))
if not file_name:
file_name = self._sanitize_editable_filename(Path(urlparse(final_url).path).name)
extension = self._editable_extension_from_mime_type(content_type, primary_mime_types, primary_mime_keywords, primary_default_extension)
return f"artifact{extension}" if not file_name else (file_name if Path(file_name).suffix else file_name + extension)
def _find_editable_value(self, payload: Any, key: str) -> str:
if isinstance(payload, str):
match = SEARCH_CONVERSATION_ID_RE.search(payload) if key == "conversation_id" else None
if match:
return match.group(1)
try:
payload = json.loads(payload)
except json.JSONDecodeError:
return ""
if isinstance(payload, dict):
value = payload.get(key)
if isinstance(value, str) and value:
return value
return next((found for item in payload.values() if (found := self._find_editable_value(item, key))), "")
if isinstance(payload, list):
return next((found for item in payload if (found := self._find_editable_value(item, key))), "")
return ""
def _extract_editable_file_ids(self, text: str) -> list[str]:
values: list[str] = []
for item in EDITABLE_ASSET_POINTER_RE.findall(text):
if item not in values:
values.append(item)
for item in FILE_ID_RE.findall(text):
if item not in values:
values.append(item)
return values
@staticmethod
def _match_editable_file_id(value: str) -> str:
match = FILE_ID_RE.search(value)
return match.group(1) if match else ""
@staticmethod
def _clean_editable_mime_type(value: Any) -> str:
text = str(value or "").strip().lower()
return text.split(";", 1)[0] if "/" in text else ""
def _looks_like_editable_primary(
self,
artifact: EditableFileArtifact,
primary_suffixes: tuple[str, ...],
primary_mime_types: set[str],
primary_mime_keywords: tuple[str, ...],
) -> bool:
path, name, mime = artifact.sandbox_path.lower(), artifact.name.lower(), artifact.mime_type
return name.endswith(primary_suffixes) or path.endswith(primary_suffixes) or mime in primary_mime_types or any(keyword in mime for keyword in primary_mime_keywords)
@staticmethod
def _looks_like_editable_zip(artifact: EditableFileArtifact) -> bool:
path, name, mime = artifact.sandbox_path.lower(), artifact.name.lower(), artifact.mime_type
return name.endswith(".zip") or path.endswith(".zip") or mime in EDITABLE_ZIP_MIME_TYPES or mime.endswith("/zip")
@staticmethod
def _editable_extension_from_mime_type(
mime_type: str,
primary_mime_types: set[str],
primary_mime_keywords: tuple[str, ...],
primary_default_extension: str,
) -> str:
if mime_type in primary_mime_types or any(keyword in mime_type for keyword in primary_mime_keywords):
return primary_default_extension
if mime_type in EDITABLE_ZIP_MIME_TYPES or mime_type.endswith("/zip"):
return ".zip"
return mimetypes.guess_extension(mime_type) or ""
@staticmethod
def _editable_filename_from_content_disposition(content_disposition: str) -> str:
extended_match = re.search(r"filename\*=UTF-8''([^;]+)", content_disposition, re.IGNORECASE)
if extended_match:
return unquote(extended_match.group(1)).strip()
plain_match = re.search(r'filename="([^"]+)"', content_disposition, re.IGNORECASE)
return plain_match.group(1).strip() if plain_match else ""
@staticmethod
def _sanitize_editable_filename(value: str) -> str:
return Path(str(value or "").strip()).name.replace("\x00", "").strip()
@staticmethod
def _unique_editable_path(path: Path) -> Path:
if not path.exists():
return path
for index in range(1, 1000):
candidate = path.with_name(f"{path.stem}_{index}{path.suffix}")
if not candidate.exists():
return candidate
raise RuntimeError(f"failed to allocate output path for {path}")
@staticmethod
def _merge_editable_artifact(current: EditableFileArtifact | None, latest: EditableFileArtifact) -> EditableFileArtifact:
if current is None:
return latest
return EditableFileArtifact(
attachment_id=latest.attachment_id or current.attachment_id,
file_id=latest.file_id or current.file_id,
name=latest.name or current.name,
mime_type=latest.mime_type or current.mime_type,
create_time=max(current.create_time, latest.create_time),
author_role=latest.author_role or current.author_role,
sandbox_path=latest.sandbox_path or current.sandbox_path,
message_id=latest.message_id or current.message_id,
)
@staticmethod
def _editable_message_text(message: Any) -> str:
if not isinstance(message, dict):
return ""
content = message.get("content") or {}
parts: list[str] = []
if isinstance(content, dict):
if isinstance(content.get("text"), str):
parts.append(content["text"])
for part in content.get("parts") or []:
if isinstance(part, str):
parts.append(part)
elif isinstance(part, dict):
parts.extend(str(part.get(key) or "") for key in ("text", "asset_pointer", "model_set_context") if part.get(key))
if isinstance(message.get("content"), str):
parts.append(str(message["content"]))
return "\n".join(part for part in parts if part).strip()
@staticmethod
def _extract_editable_export_paths(payload: Any, export_file_re: re.Pattern[str]) -> list[str]:
if isinstance(payload, str):
text = payload
else:
try:
text = json.dumps(payload, ensure_ascii=False)
except Exception:
text = str(payload)
values: list[str] = []
for item in export_file_re.findall(text):
path = str(item or "").strip()
if path and path not in values:
values.append(path)
return values
def search(self, prompt: str, model: str = SEARCH_MODEL, timeout_secs: float = SEARCH_TIMEOUT_SECS,
poll_interval_secs: float = SEARCH_POLL_INTERVAL_SECS) -> Dict[str, Any]:
if not self.access_token:
raise RuntimeError("access_token is required for search")
conduit_token = self._prepare_search_conversation(prompt, model)
self._bootstrap()
conversation_id = self._run_search_conversation(prompt, conduit_token, model)
return self._wait_search_result(conversation_id, timeout_secs, poll_interval_secs)
def _prepare_search_conversation(self, prompt: str, model: str) -> str:
path = "/backend-api/f/conversation/prepare"
response = self.session.post(
self.base_url + path,
headers=self._headers(path, {"Accept": "*/*", "Content-Type": "application/json", "X-Conduit-Token": "no-token"}),
json={
"action": "next",
"fork_from_shared_post": False,
"parent_message_id": "client-created-root",
"model": model,
"client_prepare_state": "success",
"timezone_offset_min": -480,
"timezone": "Asia/Shanghai",
"conversation_mode": {"kind": "primary_assistant"},
"system_hints": ["search"],
"partial_query": {"id": new_uuid(), "author": {"role": "user"}, "content": {"content_type": "text", "parts": [prompt]}},
"supports_buffering": True,
"supported_encodings": ["v1"],
"client_contextual_info": {"app_name": "chatgpt.com"},
},
timeout=60,
)
ensure_ok(response, path)
token = str(response.json().get("conduit_token") or "")
if not token:
raise RuntimeError("missing conduit_token")
return token
def _run_search_conversation(self, prompt: str, conduit_token: str, model: str) -> str:
requirements = self._get_chat_requirements()
path = "/backend-api/f/conversation"
response = self.session.post(
self.base_url + path,
headers=self._image_headers(path, requirements, conduit_token, "text/event-stream"),
json={
"action": "next",
"messages": [{
"id": new_uuid(),
"author": {"role": "user"},
"create_time": time.time(),
"content": {"content_type": "text", "parts": [prompt]},
"metadata": {
"developer_mode_connector_ids": [],
"selected_github_repos": [],
"selected_all_github_repos": False,
"system_hints": ["search"],
"serialization_metadata": {"custom_symbol_offsets": []},
},
}],
"parent_message_id": "client-created-root",
"model": model,
"client_prepare_state": "success",
"timezone_offset_min": -480,
"timezone": "Asia/Shanghai",
"conversation_mode": {"kind": "primary_assistant"},
"enable_message_followups": True,
"system_hints": [],
"supports_buffering": True,
"supported_encodings": ["v1"],
"force_use_search": True,
"client_reported_search_source": "conversation_composer_web_icon",
"client_contextual_info": {"is_dark_mode": False, "time_since_loaded": 36, "page_height": 925, "page_width": 886, "pixel_ratio": 2, "screen_height": 1440, "screen_width": 2560, "app_name": "chatgpt.com"},
"paragen_cot_summary_display_override": "allow",
"force_parallel_switch": "auto",
},
timeout=300,
stream=True,
)
ensure_ok(response, path)
conversation_id = ""
try:
for payload in iter_sse_payloads(response):
conversation_id = conversation_id or self._find_search_value(payload, "conversation_id")
if payload == "[DONE]":
break
finally:
response.close()
if not conversation_id:
raise RuntimeError("conversation_id not found in stream")
return conversation_id
def _wait_search_result(self, conversation_id: str, timeout_secs: float, poll_interval_secs: float) -> Dict[str, Any]:
deadline = time.time() + timeout_secs
last_result: Dict[str, Any] | None = None
last_answer = ""
stable_hits = 0
while time.time() < deadline:
try:
last_result = self._extract_search_result(conversation_id, self._get_search_conversation(conversation_id))
except UpstreamHTTPError as exc:
if exc.status_code not in {404, 409, 423, 429, 500, 502, 503, 504}:
raise
if last_result and last_result.get("answer"):
if last_result.get("status") in SEARCH_DONE_STATUS:
return last_result
answer = str(last_result.get("answer") or "")
stable_hits = stable_hits + 1 if answer == last_answer else 0
last_answer = answer
if stable_hits >= 2:
return last_result
time.sleep(poll_interval_secs)
if last_result:
return last_result
raise RuntimeError(f"timed out waiting for search result: {conversation_id}")
def _get_search_conversation(self, conversation_id: str) -> Dict[str, Any]:
path = f"/backend-api/conversation/{conversation_id}"
headers = self._headers(path, {"Accept": "*/*"})
headers["Referer"] = f"{self.base_url}/c/{conversation_id}"
headers["X-OpenAI-Target-Route"] = "/backend-api/conversation/{conversation_id}"
response = self.session.get(self.base_url + path, headers=headers, timeout=60)
ensure_ok(response, path)
return response.json()
def _extract_search_result(self, conversation_id: str, conversation: Dict[str, Any]) -> Dict[str, Any]:
messages = []
for node in (conversation.get("mapping") or {}).values():
message = (node or {}).get("message") or {}
if ((message.get("author") or {}).get("role") or "") == "assistant":
messages.append(message)
message = max(messages, key=lambda item: float(item.get("create_time") or 0.0)) if messages else {}
metadata = message.get("metadata") if isinstance(message.get("metadata"), dict) else {}
finish_details = metadata.get("finish_details") if isinstance(metadata.get("finish_details"), dict) else {}
answer = self._search_message_text(message)
sources = self._extract_search_sources(message)
for url in SEARCH_URL_RE.findall(answer):
url = self._clean_search_url(url)
if url and all(item["url"] != url for item in sources):
sources.append({"title": "", "url": url, "snippet": "", "source_type": ""})
return {
"conversation_id": conversation_id,
"status": str(finish_details.get("type") or metadata.get("status") or self._find_search_value(message, "status") or "").strip(),
"answer": answer,
"sources": sources,
"assistant_message_id": str(message.get("id") or ""),
"create_time": float(message.get("create_time") or 0.0),
}
def _extract_search_sources(self, payload: Any) -> list[Dict[str, str]]:
sources: list[Dict[str, str]] = []
for obj in self._walk_search_dicts(payload):
metadata = obj.get("metadata") if isinstance(obj.get("metadata"), dict) else {}
url = self._clean_search_url(obj.get("url") or obj.get("link") or obj.get("source_url") or metadata.get("url"))
if url and all(item["url"] != url for item in sources):
sources.append({
"title": str(obj.get("title") or obj.get("name") or obj.get("source") or "").strip(),
"url": url,
"snippet": str(obj.get("snippet") or obj.get("text") or obj.get("description") or "").strip(),
"source_type": str(obj.get("type") or obj.get("source_type") or "").strip(),
})
return sources
def _search_message_text(self, message: Any) -> str:
content = message.get("content") if isinstance(message, dict) else {}
parts = []
if isinstance(content, dict):
if isinstance(content.get("text"), str):
parts.append(content["text"])
for part in content.get("parts") or []:
if isinstance(part, str):
parts.append(part)
elif isinstance(part, dict):
parts.extend(str(part.get(key) or "") for key in ("text", "summary", "content") if part.get(key))
elif isinstance(content, str):
parts.append(content)
return "\n".join(part.strip() for part in parts if str(part).strip()).strip()
def _find_search_value(self, payload: Any, key: str) -> str:
if isinstance(payload, str):
match = SEARCH_CONVERSATION_ID_RE.search(payload) if key == "conversation_id" else None
if match:
return match.group(1)
try:
payload = json.loads(payload)
except json.JSONDecodeError:
return ""
if isinstance(payload, dict):
value = payload.get(key)
if isinstance(value, str) and value:
return value
return next((found for item in payload.values() if (found := self._find_search_value(item, key))), "")
if isinstance(payload, list):
return next((found for item in payload if (found := self._find_search_value(item, key))), "")
return ""
def _walk_search_dicts(self, payload: Any) -> list[Dict[str, Any]]:
if isinstance(payload, dict):
return [payload, *(item for value in payload.values() for item in self._walk_search_dicts(value))]
if isinstance(payload, list):
return [item for value in payload for item in self._walk_search_dicts(value)]
return []
def _clean_search_url(self, value: Any) -> str:
return str(value or "").strip().rstrip(".,;，。；")
@staticmethod
def _add_unique(values: list[str], candidates: list[str]) -> None:
for candidate in candidates:
if candidate and candidate not in values:
values.append(candidate)
@classmethod
def _extract_image_reference_ids(cls, payload: Any) -> tuple[list[str], list[str]]:
file_ids: list[str] = []
sediment_ids: list[str] = []
def walk(value: Any) -> None:
if isinstance(value, str):
# 只提取真正的图片文件 ID（file_00000000... 格式）和 file-service:// URI
cls._add_unique(file_ids, FILE_SERVICE_ID_RE.findall(value))
cls._add_unique(file_ids, REAL_IMAGE_FILE_ID_RE.findall(value))
cls._add_unique(sediment_ids, SEDIMENT_ID_RE.findall(value))
return
if isinstance(value, dict):
for item in value.values():
walk(item)
return
if isinstance(value, list):
for item in value:
walk(item)
walk(payload)
return file_ids, sediment_ids
@classmethod
def _has_image_asset_pointer(cls, payload: Any) -> bool:
if isinstance(payload, dict):
if str(payload.get("content_type") or "") == "image_asset_pointer":
return True
asset_pointer = str(payload.get("asset_pointer") or "")
if asset_pointer.startswith(("file-service://", "sediment://")):
return True
return any(cls._has_image_asset_pointer(item) for item in payload.values())
if isinstance(payload, list):
return any(cls._has_image_asset_pointer(item) for item in payload)
return False
def _extract_image_tool_records(self, data: Dict[str, Any]) -> list[Dict[str, Any]]:
"""从 conversation 明细里提取图片工具输出记录。"""
mapping = data.get("mapping") or {}
records = []
for message_id, node in mapping.items():
message = (node or {}).get("message") or {}
author = message.get("author") or {}
metadata = message.get("metadata") or {}
content = message.get("content") or {}
role = str(author.get("role") or "").strip().lower()
if role not in {"tool", "assistant"}:
continue
is_image_gen = metadata.get("async_task_type") == "image_gen"
has_asset_pointer = self._has_image_asset_pointer(content) or self._has_image_asset_pointer(metadata)
if role == "assistant" and not (is_image_gen or has_asset_pointer):
continue
file_ids, sediment_ids = self._extract_image_reference_ids({"content": content, "metadata": metadata})
if not is_image_gen and not has_asset_pointer and not file_ids and not sediment_ids:
continue
records.append(
{"message_id": message_id, "create_time": message.get("create_time") or 0, "file_ids": file_ids,
"sediment_ids": sediment_ids})
return sorted(records, key=lambda item: item["create_time"])
@staticmethod
def _find_content_policy_error_in_conversation(data: Dict[str, Any]) -> str:
"""从对话文档中查找内容政策违规错误消息。
上游拒绝生成图片时，错误消息会出现在 assistant 消息的文本中。
本方法遍历所有 assistant/tool 消息，检查是否包含内容政策违规关键词，
如果匹配则返回该消息文本（截断至 500 字符），否则返回空字符串。
"""
mapping = data.get("mapping") or {}
for node in mapping.values():
message = (node or {}).get("message") or {}
author = message.get("author") or {}
role = str(author.get("role") or "").strip().lower()
if role not in {"assistant", "tool"}:
continue
content = message.get("content") or {}
# 提取消息文本
text_parts: list[str] = []
if isinstance(content, dict):
msg_parts = content.get("parts") or []
if isinstance(msg_parts, list):
for part in msg_parts:
if isinstance(part, str) and part.strip():
text_parts.append(part.strip())
text_field = str(content.get("text") or "")
if text_field.strip():
text_parts.append(text_field.strip())
elif isinstance(content, str) and content.strip():
text_parts.append(content.strip())
msg_text = "\n".join(text_parts)
if msg_text and _is_content_policy_error(msg_text):
return msg_text[:500]
return ""
def _poll_image_results(
self,
conversation_id: str,
timeout_secs: float = 120.0,
initial_file_ids: list[str] | None = None,
initial_sediment_ids: list[str] | None = None,
) -> tuple[list[str], list[str]]:
"""Poll the conversation document until image file ids appear or budget runs out.
- Sleeps image_poll_initial_wait_secs first (default 10s, +jitter). ChatGPT
image generation takes ~30s; polling immediately wastes requests and trips
a transient 429 the upstream returns within ~200ms of the SSE stream
closing (the conversation document is not yet committed).
- Subsequent polls are image_poll_interval_secs apart (default 10s).
- On upstream 429 / 5xx or network errors, backs off exponentially
(capped at 16s, +jitter) honoring Retry-After when present.
- All sleeps stay within timeout_secs; on exhaustion raises ImagePollTimeoutError.
"""
start = time.time()
attempt = 0
interval = float(config.image_poll_interval_secs)
initial_wait = float(config.image_poll_initial_wait_secs)
file_ids: list[str] = []
sediment_ids: list[str] = []
self._add_unique(file_ids, initial_file_ids or [])
self._add_unique(sediment_ids, initial_sediment_ids or [])
has_initial_ids = bool(file_ids or sediment_ids)
last_hit_key: tuple[tuple[str, ...], tuple[str, ...]] | None = (
(tuple(file_ids), tuple(sediment_ids)) if has_initial_ids else None
)
logger.info({
"event": "image_poll_start",
"conversation_id": conversation_id,
"timeout_secs": timeout_secs,
"initial_wait_secs": initial_wait,
"interval_secs": interval,
"initial_file_ids": file_ids,
"initial_sediment_ids": sediment_ids,
})
def _remaining() -> float:
return timeout_secs - (time.time() - start)
if has_initial_ids and config.image_settle_enabled:
settle_for = min(config.image_settle_secs, max(0.0, _remaining()))
if settle_for > 0:
time.sleep(settle_for)
elif initial_wait > 0:
jitter = random.uniform(0, min(2.0, initial_wait * 0.2))
sleep_for = min(initial_wait + jitter, max(0.0, _remaining()))
if sleep_for > 0:
time.sleep(sleep_for)
def _retry_sleep(reason: str, status_code: int | None, error: str | None, retry_after: int | None) -> bool:
# retry_after=0 means "retry immediately" — must not be coerced via falsy check.
base = retry_after if retry_after is not None else min(2 ** min(attempt, 4), 16)
backoff = base + random.uniform(0, 0.5)
remaining = _remaining()
if remaining <= 0:
return False
sleep_for = min(backoff, remaining)
log_payload: Dict[str, Any] = {
"event": "image_poll_retry",
"conversation_id": conversation_id,
"attempt": attempt,
"reason": reason,
"sleep_secs": round(sleep_for, 2),
}
if status_code is not None:
log_payload["status_code"] = status_code
if error is not None:
log_payload["error"] = error
logger.warning(log_payload)
time.sleep(sleep_for)
return True
last_task_error = ""
while _remaining() > 0:
attempt += 1
# 在每次轮询时，检查 /backend-api/tasks/ 是否有错误（仅记录，不中断）
# 内容政策违规检测通过对话文本进行（在 _find_content_policy_error_in_conversation 中）
last_task_error = ""
try:
tasks = self._query_backend_tasks(conversation_id=conversation_id, timeout_secs=5.0)
for task in tasks:
is_error, error_msg, metadata = self.check_task_error(task)
if is_error and error_msg:
last_task_error = error_msg
logger.info({
"event": "image_poll_task_error_not_blocking",
"conversation_id": conversation_id,
"attempt": attempt,
"error_msg": error_msg,
"metadata": metadata,
})
except Exception as exc:
# tasks 查询失败不影响正常轮询流程
logger.debug({
"event": "image_poll_task_check_failed",
"conversation_id": conversation_id,
"attempt": attempt,
"error": str(exc),
})
try:
conversation = self._get_conversation(conversation_id)
except UpstreamHTTPError as exc:
if exc.status_code in (429, 500, 502, 503, 504):
if _retry_sleep("upstream_status", exc.status_code, None, exc.retry_after):
continue
break
raise
except requests.exceptions.RequestException as exc:
if _retry_sleep("network", None, str(exc), None):
continue
break
for record in self._extract_image_tool_records(conversation):
for file_id in record["file_ids"]:
if file_id not in file_ids:
file_ids.append(file_id)
for sediment_id in record["sediment_ids"]:
if sediment_id not in sediment_ids:
sediment_ids.append(sediment_id)
# 检查对话文本中是否包含内容政策违规错误
# 当上游拒绝生成图片时，错误消息会出现在对话文档的 assistant 消息中，
# 而非 /backend-api/tasks/ 的 task error 结构中。
# 如果在没有找到图片文件 ID 的同时检测到内容政策违规，立即中断轮询。
if not file_ids and not sediment_ids:
policy_msg = self._find_content_policy_error_in_conversation(conversation)
if policy_msg:
logger.warning({
"event": "image_poll_conversation_text_policy_violation",
"conversation_id": conversation_id,
"attempt": attempt,
"error_msg": policy_msg[:200],
})
raise ImageContentPolicyError(policy_msg)
logger.debug({"event": "image_poll_check", "conversation_id": conversation_id, "attempt": attempt,
"file_ids": file_ids, "sediment_ids": sediment_ids})
if file_ids or sediment_ids:
if not config.image_check_before_hit_enabled:
# 先check再hit 机制关闭：直接返回首次发现的 file_ids
logger.info({"event": "image_poll_hit_no_settle", "conversation_id": conversation_id,
"file_ids": file_ids, "sediment_ids": sediment_ids})
return file_ids, sediment_ids
hit_key = (tuple(file_ids), tuple(sediment_ids))
if last_hit_key == hit_key:
logger.info({"event": "image_poll_hit", "conversation_id": conversation_id, "file_ids": file_ids,
"sediment_ids": sediment_ids})
return file_ids, sediment_ids
last_hit_key = hit_key
if not config.image_settle_enabled:
# 二次确认机制关闭：直接返回首次发现的 file_ids
logger.info({"event": "image_poll_hit_settle_disabled", "conversation_id": conversation_id,
"file_ids": file_ids, "sediment_ids": sediment_ids})
return file_ids, sediment_ids
logger.info({"event": "image_poll_hit_pending_settle", "conversation_id": conversation_id,
"file_ids": file_ids, "sediment_ids": sediment_ids,
"settle_secs": config.image_settle_secs})
wait = min(config.image_settle_secs, max(0.0, _remaining()))
if wait > 0:
time.sleep(wait)
continue
return file_ids, sediment_ids
logger.debug({"event": "image_poll_wait", "conversation_id": conversation_id,
"elapsed_secs": round(time.time() - start, 1)})
wait = min(interval, max(0.0, _remaining()))
if wait > 0:
time.sleep(wait)
logger.info({
"event": "image_poll_timeout",
"conversation_id": conversation_id,
"timeout_secs": timeout_secs,
"attempts_made": attempt,
# attempts_made == 0 means the initial_wait consumed the entire budget — no HTTP attempted.
"initial_wait_exhausted_budget": attempt == 0,
"last_task_error": last_task_error if last_task_error else None,
})
exc = ImagePollTimeoutError(
f"ChatGPT 生图超时（已等待 {timeout_secs} 秒）。"
f"当前超时阈值可在 config.json 中调大 image_poll_timeout_secs，"
f"也可能是账号被限流或生图队列拥堵导致。"
)
if last_task_error:
setattr(exc, "task_error", last_task_error)
setattr(exc, "conversation_id", conversation_id or "")
raise exc
def _get_file_download_url(self, file_id: str) -> str:
"""获取文件下载地址。"""
path = f"/backend-api/files/{file_id}/download"
response = self.session.get(self.base_url + path, headers=self._headers(path, {"Accept": "application/json"}),
timeout=60)
ensure_ok(response, path)
data = response.json()
return data.get("download_url") or data.get("url") or ""
def _get_attachment_download_url(self, conversation_id: str, attachment_id: str) -> str:
"""通过 conversation 附件接口获取下载地址。"""
path = f"/backend-api/conversation/{conversation_id}/attachment/{attachment_id}/download"
response = self.session.get(self.base_url + path, headers=self._headers(path, {"Accept": "application/json"}),
timeout=60)
ensure_ok(response, path)
data = response.json()
return data.get("download_url") or data.get("url") or ""
def _query_backend_tasks(
self,
conversation_id: str = "",
task_id: str = "",
timeout_secs: float = 30.0,
) -> list[Dict[str, Any]]:
"""查询 /backend-api/tasks/ 接口获取异步任务状态和错误信息。
参数：
- `conversation_id`：可选。按 conversation_id 过滤任务。
- `task_id`：可选。按 task_id 过滤任务。
- `timeout_secs`：请求超时秒数。
返回：
- 任务列表，每个任务包含 image_gen_message 等字段。
"""
path = "/backend-api/tasks"
response = self.session.get(
self.base_url + path,
headers=self._headers(path, {"Accept": "application/json"}),
timeout=timeout_secs,
)
ensure_ok(response, path)
data = response.json()
tasks = data.get("tasks", [])
if not isinstance(tasks, list):
return []
# 按 conversation_id 或 task_id 过滤
if conversation_id:
tasks = [
t for t in tasks
if isinstance(t, dict) and (
t.get("conversation_id") == conversation_id
or t.get("original_conversation_id") == conversation_id
)
]
if task_id:
tasks = [t for t in tasks if isinstance(t, dict) and t.get("task_id") == task_id]
return tasks
def check_task_error(self, task: Dict[str, Any]) -> tuple[bool, str, Dict[str, Any]]:
"""检查单个任务是否包含结构化错误。
通过以下字段判断（不依赖文本匹配）：
- image_gen_message.metadata.is_error == True
- image_gen_message.author.role == "assistant" (而非 "tool")
- image_gen_message.content.content_type == "text" (而非 "multimodal_text")
返回：
- (is_error, error_msg, metadata)
"""
img_msg = task.get("image_gen_message") or {}
if not img_msg:
return False, "", {}
metadata = img_msg.get("metadata") or {}
content = img_msg.get("content") or {}
author = img_msg.get("author") or {}
is_error = metadata.get("is_error", False)
is_text_only = content.get("content_type") == "text"
is_assistant_role = author.get("role") == "assistant"
# 提取错误文本
error_msg = ""
if is_error and is_text_only:
parts = content.get("parts", [])
error_msg = "".join(p for p in parts if isinstance(p, str))
return is_error, error_msg, metadata
def _resolve_image_urls(self, conversation_id: str, file_ids: list[str], sediment_ids: list[str]) -> list[str]:
"""把图片结果 id 解析成可下载 URL。"""
urls = []
skip_patterns = {"file_upload"}
for file_id in file_ids:
if file_id in skip_patterns:
logger.debug({
"event": "image_file_id_skipped",
"source": "file",
"conversation_id": conversation_id,
"id": file_id,
})
continue
try:
url = self._get_file_download_url(file_id)
except Exception as exc:
logger.debug({
"event": "image_download_url_failed",
"source": "file",
"conversation_id": conversation_id,
"id": file_id,
"error": repr(exc),
})
continue
if url:
if url not in urls:
urls.append(url)
else:
logger.debug({
"event": "image_download_url_empty",
"source": "file",
"conversation_id": conversation_id,
"id": file_id,
})
if not conversation_id or not sediment_ids:
logger.debug({
"event": "image_urls_resolved",
"conversation_id": conversation_id,
"file_ids": file_ids,
"sediment_ids": sediment_ids,
"urls": urls,
})
return urls
for sediment_id in sediment_ids:
try:
url = self._get_attachment_download_url(conversation_id, sediment_id)
except Exception as exc:
logger.debug({
"event": "image_download_url_failed",
"source": "sediment",
"conversation_id": conversation_id,
"id": sediment_id,
"error": repr(exc),
})
continue
if url:
if url not in urls:
urls.append(url)
else:
logger.debug({
"event": "image_download_url_empty",
"source": "sediment",
"conversation_id": conversation_id,
"id": sediment_id,
})
logger.debug({
"event": "image_urls_resolved",
"conversation_id": conversation_id,
"file_ids": file_ids,
"sediment_ids": sediment_ids,
"urls": urls,
})
return urls
def resolve_conversation_image_urls(
self,
conversation_id: str,
file_ids: list[str],
sediment_ids: list[str],
poll: bool = True,
poll_timeout_secs: float | None = None,
) -> list[str]:
file_ids = [item for item in file_ids if item != "file_upload"]
sediment_ids = list(sediment_ids)
timeout = poll_timeout_secs if poll_timeout_secs is not None else config.image_poll_timeout_secs
# 当 check-before-hit 和 settle 均已关闭，且 SSE 已给出 file_ids 时，
# 跳过轮询直接解析 URL，省去 initial_wait + 轮询耗时。
if poll and conversation_id and (file_ids or sediment_ids):
if not config.image_check_before_hit_enabled and not config.image_settle_enabled:
logger.info({
"event": "image_resolve_skip_poll_direct_resolve",
"conversation_id": conversation_id,
"file_ids": file_ids,
"sediment_ids": sediment_ids,
})
return self._resolve_image_urls(conversation_id, file_ids, sediment_ids)
if poll and conversation_id:
logger.info({
"event": "image_resolve_poll_needed",
"conversation_id": conversation_id,
"initial_file_ids": file_ids,
"initial_sediment_ids": sediment_ids,
"poll_timeout_secs": timeout,
})
try:
polled_file_ids, polled_sediment_ids = self._poll_image_results(
conversation_id,
timeout,
file_ids,
sediment_ids,
)
except ImagePollTimeoutError as exc:
# 如果轮询超时且有 task error（如 moderation 拦截），抛出 ImageContentPolicyError
# 而非 ImagePollTimeoutError，让调用方能区分真正的超时和上游拒绝
task_error = getattr(exc, "task_error", "")
if not file_ids and not sediment_ids:
if task_error:
raise ImageContentPolicyError(task_error) from exc
raise
logger.warning({
"event": "image_resolve_poll_partial_timeout",
"conversation_id": conversation_id,
"file_ids": file_ids,
"sediment_ids": sediment_ids,
})
except Exception as exc:
if not file_ids and not sediment_ids:
raise
logger.warning({
"event": "image_resolve_poll_partial_error",
"conversation_id": conversation_id,
"file_ids": file_ids,
"sediment_ids": sediment_ids,
"error": repr(exc),
})
else:
file_ids.extend(item for item in polled_file_ids if item and item not in file_ids)
sediment_ids.extend(item for item in polled_sediment_ids if item and item not in sediment_ids)
return self._resolve_image_urls(conversation_id, file_ids, sediment_ids)
def download_image_bytes(self, urls: list[str]) -> list[bytes]:
images = []
for url in urls:
response = self.session.get(url, timeout=120)
ensure_ok(response, "image_download")
if response.content not in images:
images.append(response.content)
return images
def stream_conversation(
self,
messages: Optional[list[Dict[str, Any]]] = None,
model: str = "auto",
prompt: str = "",
images: Optional[list[str]] = None,
system_hints: Optional[list[str]] = None,
thinking_effort: str = "",
) -> Iterator[str]:
system_hints = system_hints or []
if "picture_v2" in system_hints:
yield from self._stream_picture_conversation(prompt, model, images or [])
return
normalized = messages or [{"role": "user", "content": prompt}]
self._bootstrap()
requirements = self._get_chat_requirements()
path, timezone = self._chat_target()
payload = self._conversation_payload(normalized, model, timezone, thinking_effort=thinking_effort)
response = self.session.post(
self.base_url + path,
headers=self._conversation_headers(path, requirements),
json=payload,
timeout=300,
stream=True,
)
ensure_ok(response, path)
try:
yield from iter_sse_payloads(response)
finally:
response.close()
def _report_progress(self, step: str) -> None:
"""Report progress step to the callback if set."""
if self.progress_callback:
try:
self.progress_callback(step)
except Exception:
pass
def _stream_picture_conversation(
self,
prompt: str,
model: str,
images: list[str],
) -> Iterator[str]:
if not self.access_token:
raise RuntimeError("access_token is required for image endpoints")
self._report_progress("uploading")
references = [self._upload_image(image, f"image_{idx}.png") for idx, image in enumerate(images, start=1)]
self._report_progress("bootstrapping")
self._bootstrap()
self._report_progress("getting_token")
requirements = self._get_chat_requirements()
self._report_progress("preparing_conversation")
conduit_token = self._prepare_image_conversation(prompt, requirements, model)
self._report_progress("starting_generation")
response = self._start_image_generation(prompt, requirements, conduit_token, model, references)
self._report_progress("generating")
try:
yield from iter_sse_payloads(response)
finally:
response.close()
def _bootstrap(self) -> None:
"""预热首页，并提取 PoW 相关脚本引用。"""
response = self.session.get(
self.base_url + "/",
headers=self._bootstrap_headers(),
timeout=30,
)
ensure_ok(response, "bootstrap")
self.pow_script_sources, self.pow_data_build = parse_pow_resources(response.text)
if not self.pow_script_sources:
self.pow_script_sources = [DEFAULT_POW_SCRIPT]
def _get_chat_requirements(self) -> ChatRequirements:
"""获取当前模式对话所需的 sentinel token（prepare + finalize 两步流程）。"""
base = "/backend-api/sentinel/chat-requirements" if self.access_token else "/backend-anon/sentinel/chat-requirements"
p_token = build_legacy_requirements_token(self.user_agent, self.pow_script_sources, self.pow_data_build)
prepare_path = base + "/prepare"
response = self.session.post(
self.base_url + prepare_path,
headers=self._headers(prepare_path, {"Content-Type": "application/json"}),
json={"p": p_token},
timeout=30,
)
ensure_ok(response, "chat_requirements_prepare")
prepare_data = response.json()
if (prepare_data.get("arkose") or {}).get("required"):
raise RuntimeError("chat requirements requires arkose token, which is not implemented")
proof_token = ""
proof_info = prepare_data.get("proofofwork") or {}
if proof_info.get("required"):
proof_token = build_proof_token(
proof_info.get("seed", ""),
proof_info.get("difficulty", ""),
self.user_agent,
script_sources=self.pow_script_sources,
data_build=self.pow_data_build,
)
turnstile_token = ""
turnstile_info = prepare_data.get("turnstile") or {}
if turnstile_info.get("required") and turnstile_info.get("dx"):
turnstile_token = solve_turnstile_token(turnstile_info["dx"], p_token) or ""
finalize_path = base + "/finalize"
response = self.session.post(
self.base_url + finalize_path,
headers=self._headers(finalize_path, {"Content-Type": "application/json"}),
json={
"prepare_token": prepare_data.get("prepare_token", ""),
"proof_token": proof_token,
"turnstile_token": turnstile_token,
},
timeout=30,
)
ensure_ok(response, "chat_requirements_finalize")
data = response.json()
token = data.get("token", "")
if not token:
message = "missing auth chat requirements token" if self.access_token else "missing chat requirements token"
raise RuntimeError(f"{message}: {data}")
return ChatRequirements(
token=token,
proof_token=proof_token,
turnstile_token=turnstile_token,
so_token=data.get("so_token", ""),
raw_finalize=data,
)
def _chat_target(self) -> tuple[str, str]:
if self.access_token:
return "/backend-api/conversation", "Asia/Shanghai"
return "/backend-anon/conversation", "America/Los_Angeles"
def list_models(self) -> Dict[str, Any]:
"""返回当前模式下可用模型，格式对齐 OpenAI `/v1/models`。"""
self._bootstrap()
path = "/backend-api/models?history_and_training_disabled=false" if self.access_token else (
"/backend-anon/models?iim=false&is_gizmo=false"
)
route = "/backend-api/models" if self.access_token else "/backend-anon/models"
context = "auth_models" if self.access_token else "anon_models"
response = self.session.get(
self.base_url + path,
headers=self._headers(route),
timeout=30,
)
ensure_ok(response, context)
data = []
seen = set()
for item in response.json().get("models", []):
if not isinstance(item, dict):
continue
slug = str(item.get("slug", "")).strip()
if not slug or slug in seen:
continue
seen.add(slug)
data.append({
"id": slug,
"object": "model",
"created": int(item.get("created") or 0),
"owned_by": str(item.get("owned_by") or "chatgpt"),
"permission": [],
"root": slug,
"parent": None,
})
data.sort(key=lambda item: item["id"])
return {"object": "list", "data": data}
|