Spaces:
Running
Running
File size: 85,242 Bytes
09801ca | 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 | """
๐ค PREDICT ENGINE v6.0 - SENIOR AIML ENGINEER ASSISTANT
=========================================================
A ChatGPT-like intelligent ML assistant designed by a Senior AIML Engineer.
Features:
- ๐ง Senior AIML Engineer level explanations
- ๐ Live integration with trained ML model charts
- ๐ฏ Real predictions from your trained AutoML model
- ๐ก Statistical insights and technical depth
- ๐ Connects with ML Predictions page for charts
- ๐ Conversational memory within session
NO HARDCODING - Everything based on real trained model + actual data!
Author: DataVision Team
Version: 6.0.0
"""
import logging
import pandas as pd
import numpy as np
from typing import Dict, Any, List, Optional, Tuple
from datetime import datetime
from enum import Enum
from dataclasses import dataclass
import re
import json
import os
logger = logging.getLogger(__name__)
# =============================================================================
# IMPORTS & DEPENDENCIES
# =============================================================================
# AutoML Engine
try:
from ml.automl_engine import automl_engine
AUTOML_AVAILABLE = True
except ImportError:
AUTOML_AVAILABLE = False
logger.warning("AutoML engine not available")
# LLM for intelligent explanations
try:
from core.llm import chat as llm_chat
LLM_AVAILABLE = True
except ImportError:
LLM_AVAILABLE = False
logger.warning("LLM not available")
# Chart Generator
try:
from ml.chart_generator import generate_ml_charts
CHARTS_AVAILABLE = True
except ImportError:
CHARTS_AVAILABLE = False
# Intelligent Query Processor (Claude-style)
try:
from core.intelligent_processor import IntelligentQueryProcessor, intelligent_process
INTELLIGENT_PROCESSOR_AVAILABLE = True
except ImportError:
INTELLIGENT_PROCESSOR_AVAILABLE = False
# =============================================================================
# QUERY TYPES - What the user is asking about
# =============================================================================
class QueryType(Enum):
"""All types of ML queries the assistant can handle"""
# Core ML Operations
PREDICTION = "prediction" # "Predict for age=60"
BATCH_PREDICTION = "batch_prediction" # "Predict all rows"
# Model Understanding
MODEL_INFO = "model_info" # "What model is trained?"
METRICS = "metrics" # "Show accuracy/F1"
FEATURE_IMPORTANCE = "importance" # "What features matter?"
FEATURE_SPECIFIC = "feature_specific" # "How does age affect..."
# Data Analysis
DATA_ANALYSIS = "data_analysis" # "Analyze the data"
DATA_DISTRIBUTION = "distribution" # "Show distribution of X"
CORRELATION = "correlation" # "What correlates with..."
# Visualizations - Direct Chart Requests
SHOW_CHART = "show_chart" # "Show confusion matrix"
CONFUSION_MATRIX = "confusion_matrix" # Direct confusion matrix
ROC_CURVE = "roc_curve" # ROC curve
PRECISION_RECALL = "precision_recall" # PR curve
ACTUAL_VS_PREDICTED = "actual_vs_pred" # Actual vs Predicted
# Advanced Analysis
WHAT_IF = "what_if" # Scenario analysis
EXPLAIN = "explain" # "Why did model predict X?"
SHAP = "shap" # SHAP explanation
THRESHOLD = "threshold" # Threshold tuning
# Meta
COMPARISON = "comparison" # Compare scenarios
DASHBOARD = "dashboard" # Full dashboard
HELP = "help" # How to use
GENERAL_ML = "general_ml" # General ML questions
# =============================================================================
# CONVERSATION MEMORY - Remember context within session
# =============================================================================
@dataclass
class ConversationContext:
"""Stores conversation context for multi-turn interactions"""
last_prediction: Dict = None
last_features: Dict = None
last_query_type: QueryType = None
chart_shown: List[str] = None
session_start: datetime = None
def __post_init__(self):
self.chart_shown = []
self.session_start = datetime.now()
# Session memory - persists during conversation
_session_memory: Dict[str, ConversationContext] = {}
def get_session(user_id: str) -> ConversationContext:
"""Get or create session for user"""
if user_id not in _session_memory:
_session_memory[user_id] = ConversationContext()
return _session_memory[user_id]
# =============================================================================
# SENIOR AIML ENGINEER PROMPTS
# =============================================================================
SENIOR_AIML_SYSTEM_PROMPT = """You are a Senior AI/ML Engineer at a top tech company.
When explaining ML concepts:
- Be precise with technical terminology
- Include statistical context (p-values, confidence intervals when relevant)
- Give actionable recommendations
- Acknowledge model limitations
- Use analogies for complex concepts
- Be concise but thorough
Your explanations should sound like a Staff ML Engineer in a code review,
not a marketing person. Be real about what the model can and cannot do."""
def generate_senior_explanation(
topic: str,
context: Dict[str, Any],
question: str = ""
) -> str:
"""Generate Senior AIML Engineer quality explanation"""
if not LLM_AVAILABLE:
return _fallback_explanation(topic, context)
prompt = f"""{SENIOR_AIML_SYSTEM_PROMPT}
Context:
- Model: {context.get('model_name', 'Unknown')}
- Task: {context.get('task_type', 'Unknown')}
- Target: {context.get('target', 'Unknown')}
- Key Metrics: {context.get('metrics', {})}
Topic to explain: {topic}
User question: {question}
Provide a concise but technically accurate explanation (2-4 sentences).
Focus on what a data scientist would want to know."""
try:
return llm_chat(prompt, temperature=0.3, max_tokens=200)
except Exception as e:
logger.error(f"LLM error: {e}")
return _fallback_explanation(topic, context)
def _fallback_explanation(topic: str, context: Dict) -> str:
"""Fallback when LLM is unavailable"""
target = context.get('target', 'the outcome')
model = context.get('model_name', 'the model')
explanations = {
'prediction': f"The {model} predicts {target} based on the input features. Higher confidence indicates stronger prediction certainty.",
'metrics': f"These metrics show how well {model} generalizes to unseen data. F1 balances precision and recall.",
'importance': f"Feature importance shows which inputs have the most predictive power for {target}.",
'default': f"This analysis helps understand how {model} makes predictions for {target}."
}
return explanations.get(topic, explanations['default'])
# =============================================================================
# QUERY DETECTION - Understand what user is asking
# =============================================================================
def detect_query_type(query: str, user_id: str = "") -> Tuple[QueryType, float, Dict]:
"""
Intelligently detect query type with confidence and extracted parameters.
Returns: (QueryType, confidence, extracted_params)
"""
q = query.lower().strip()
params = {}
# === CHART REQUESTS (High Priority) ===
chart_patterns = {
QueryType.CONFUSION_MATRIX: [
r'confusion\s*matrix', r'confusion\s+matrix', r'show\s+confusion'
],
QueryType.ROC_CURVE: [
r'roc\s*curve', r'roc\s+curve', r'receiver\s+operating'
],
QueryType.PRECISION_RECALL: [
r'precision.+recall', r'pr\s+curve', r'precision\s+recall\s+curve'
],
QueryType.ACTUAL_VS_PREDICTED: [
r'actual\s*(vs?|versus)\s*predicted', r'predicted\s+vs\s+actual',
r'scatter\s+plot', r'residual'
],
QueryType.SHOW_CHART: [
r'show\s+(me\s+)?(the\s+)?(chart|graph|plot|visualization)',
r'display\s+(the\s+)?(chart|graph)',
r'(feature\s+)?importance\s+(chart|plot)',
r'(class\s+)?distribution\s+(chart|plot)',
r'calibration\s+(curve|plot)',
r'correlation\s+heatmap',
# NEW: Generic ML chart requests
r'(give|show|display|generate|create)\s+(me\s+)?(two|2|some|the|any|my)?\s*(ml|machine\s*learning)\s*(chart|graph|plot|visualization)s?',
r'(ml|machine\s*learning)\s*(chart|graph|plot|visualization)s?',
r'(two|2|some|any)\s*(chart|graph|plot|visualization)s?',
r'(give|show|display)\s+(me\s+)?(chart|graph|plot|visualization)s?',
r'(trained|my)\s*(model|data)?\s*(chart|graph|plot|visualization)s?',
r'based\s+on\s+(my\s+)?(trained|ml|model)\s*(data)?\s*(give|show|display)?',
]
}
for qtype, patterns in chart_patterns.items():
for pattern in patterns:
if re.search(pattern, q):
# Extract specific chart name if present
chart_keywords = ['confusion', 'roc', 'precision', 'recall', 'feature',
'importance', 'distribution', 'calibration', 'correlation']
for kw in chart_keywords:
if kw in q:
params['chart_name'] = kw
break
return qtype, 0.95, params
# === PREDICTION REQUESTS ===
prediction_patterns = [
r'predict\s+(for|if|when|what|with)',
r'what\s+(is|will|would)\s+.+\s+(be|happen|predict)',
r'classify\s+',
r'estimate\s+',
r'will\s+.+\s+(survive|die|default|churn|happen)',
r'[a-z_]+\s*[=:]\s*\d+', # feature=value pattern
]
for pattern in prediction_patterns:
if re.search(pattern, q):
params['features'] = _extract_feature_values(q)
return QueryType.PREDICTION, 0.95, params
# === METRICS/PERFORMANCE ===
metrics_patterns = [
r'(accuracy|precision|recall|f1|f2|auc|performance)',
r'how\s+(good|accurate|well)',
r'model\s+(metrics|performance|score)',
r'(show|display)\s+.*(accuracy|metrics|performance)'
]
for pattern in metrics_patterns:
if re.search(pattern, q):
return QueryType.METRICS, 0.90, params
# === FEATURE IMPORTANCE ===
importance_patterns = [
r'(what|which)\s+(affects?|influences?|matters?|important)',
r'feature\s+(importance|ranking)',
r'(most|least)\s+important',
r'(key|top|main)\s+(factors?|features?)'
]
for pattern in importance_patterns:
if re.search(pattern, q):
return QueryType.FEATURE_IMPORTANCE, 0.90, params
# === WHAT-IF SCENARIOS ===
whatif_patterns = [
r'what\s+(if|would\s+happen)',
r'if\s+i\s+(change|increase|decrease)',
r'scenario',
r'suppose',
r'hypothetically'
]
for pattern in whatif_patterns:
if re.search(pattern, q):
params['scenario'] = q
return QueryType.WHAT_IF, 0.88, params
# === EXPLANATION REQUESTS ===
explain_patterns = [
r'why\s+(did|does|was|is)',
r'explain\s+(this|the|why|how)',
r'interpret',
r'understand\s+(why|this)',
r'what\s+does\s+this\s+(mean|show)'
]
for pattern in explain_patterns:
if re.search(pattern, q):
return QueryType.EXPLAIN, 0.85, params
# === MODEL INFO ===
model_patterns = [
r'(what|which)\s+(model|algorithm)',
r'model\s+(type|name|info)',
r'what\s+.+\s+trained',
r'features?\s+(used|available)'
]
for pattern in model_patterns:
if re.search(pattern, q):
return QueryType.MODEL_INFO, 0.85, params
# === DASHBOARD ===
if any(x in q for x in ['dashboard', 'all charts', 'full report', 'comprehensive']):
return QueryType.DASHBOARD, 0.88, params
# === HELP ===
if any(x in q for x in ['help', 'how to use', 'what can you do', 'example']):
return QueryType.HELP, 0.90, params
# Default to general ML
return QueryType.GENERAL_ML, 0.70, params
def _extract_feature_values(query: str) -> Dict[str, Any]:
"""Extract feature=value pairs from query"""
features = {}
# Pattern: feature_name=value or feature_name: value
pattern = r'([a-z_]+)\s*[=:]\s*([0-9.]+|true|false|yes|no|male|female|[a-z]+)'
matches = re.findall(pattern, query.lower())
for name, value in matches:
# Try to convert to appropriate type
try:
if '.' in value:
features[name] = float(value)
elif value.isdigit():
features[name] = int(value)
elif value in ['true', 'yes']:
features[name] = True
elif value in ['false', 'no']:
features[name] = False
else:
features[name] = value
except:
features[name] = value
return features
# =============================================================================
# CHART RETRIEVAL - Get stored charts from trained model
# =============================================================================
def get_stored_charts(user_id: str) -> Dict[str, Any]:
"""
Retrieve charts generated during model training.
Uses model_persistence for proper storage location.
"""
charts = {}
try:
# PRIMARY: Try model_persistence (correct location)
try:
from ml.model_persistence import model_persistence
charts = model_persistence.get_charts(user_id) or {}
if charts:
logger.info(f"๐ Retrieved {len(charts)} charts from model_persistence for user {user_id}")
return charts
except Exception as e:
logger.debug(f"model_persistence not available: {e}")
# FALLBACK: Try loading from automl engine saved state
try:
from ml.automl_engine import automl_engine
if automl_engine and hasattr(automl_engine, 'charts'):
engine_charts = automl_engine.charts or {}
if engine_charts:
charts.update(engine_charts)
logger.info(f"๐ Retrieved {len(charts)} charts from automl_engine")
except Exception as e:
logger.debug(f"automl_engine charts not available: {e}")
# FALLBACK: Try NLP engine charts
try:
from ml.nlp_engine import NLPEngine
nlp_engine = NLPEngine()
nlp_engine.load(user_id)
if hasattr(nlp_engine, 'charts') and nlp_engine.charts:
for key, chart in nlp_engine.charts.items():
charts[f"nlp_{key}"] = chart
logger.info(f"๐ Retrieved NLP charts for user {user_id}")
except Exception as e:
logger.debug(f"NLP charts not available: {e}")
# FALLBACK: Try Deep Learning engine charts
try:
from ml.deep_learning_engine import DeepLearningEngine
dl_engine = DeepLearningEngine()
dl_engine.load(user_id)
if hasattr(dl_engine, 'charts') and dl_engine.charts:
for key, chart in dl_engine.charts.items():
charts[f"dl_{key}"] = chart
logger.info(f"๐ Retrieved Deep Learning charts for user {user_id}")
except Exception as e:
logger.debug(f"DL charts not available: {e}")
# FALLBACK: Check legacy storage path (Docker-aware)
base_path = f"/app/backend/storage/automl/{user_id}" if os.path.exists("/app") else f"storage/automl/{user_id}"
if os.path.exists(base_path):
state_path = f"{base_path}/automl_state.pkl"
if os.path.exists(state_path):
import pickle
with open(state_path, 'rb') as f:
state = pickle.load(f)
if 'charts' in state:
charts.update(state['charts'])
logger.info(f"๐ Total charts retrieved for user {user_id}: {len(charts)}")
except Exception as e:
logger.error(f"Chart retrieval error: {e}")
return charts
def get_chart_by_name(user_id: str, chart_name: str) -> Optional[Dict]:
"""Get a specific chart by name (fuzzy matching)"""
charts = get_stored_charts(user_id)
# Direct match
if chart_name in charts:
return charts[chart_name]
# Fuzzy match
chart_name_lower = chart_name.lower().replace(' ', '_')
for key, value in charts.items():
if chart_name_lower in key.lower() or key.lower() in chart_name_lower:
return value
# Partial match
for key, value in charts.items():
if any(part in key.lower() for part in chart_name_lower.split('_')):
return value
return None
# =============================================================================
# MAIN RESPONSE FUNCTION
# =============================================================================
def predict_response_sync(user_id: str, query: str, context: str = "", df=None) -> Dict[str, Any]:
"""
๐ค SENIOR AIML ENGINEER ML ASSISTANT
Main entry point for all ML queries. Routes to appropriate handler
and returns ChatGPT-like responses with charts and explanations.
Uses DYNAMIC routing - ML queries use ML, general queries use AI.
"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.5,
"sources": [],
"ml_used": False,
"chart": None,
"charts": [] # Multiple charts support
}
start_time = datetime.now()
session = get_session(user_id)
# =================================================================
# ๐ DYNAMIC ROUTING: Check if query relates to ML/predictions
# =================================================================
q_lower = query.lower()
# Get column names from data
column_names = [col.lower() for col in df.columns] if df is not None and hasattr(df, 'columns') else []
column_names_spaced = [col.replace('_', ' ') for col in column_names]
# ML-related terms
ml_terms = [
'predict', 'forecast', 'model', 'train', 'accuracy', 'feature',
'classification', 'regression', 'ml', 'machine learning',
'precision', 'recall', 'f1', 'confusion', 'roc', 'auc'
]
# Data terms
data_terms = column_names + column_names_spaced + [
'my data', 'my ', 'our ', 'the data', 'uploaded', 'dataset',
'total', 'sum', 'average', 'count', 'revenue', 'sales', 'customer'
]
query_is_about_ml = any(term in q_lower for term in ml_terms)
query_is_about_data = any(term in q_lower for term in data_terms if term)
# If NOT about ML or data, use pure AI Knowledge
if not query_is_about_ml and not query_is_about_data:
logger.info("๐ Predict: Routing to AI KNOWLEDGE (query not about ML/data)")
if LLM_AVAILABLE:
try:
ai_prompt = f"""You are an AI with machine learning expertise.
Answer this question using your knowledge:
{query}
Provide a helpful, accurate response. If it's about ML concepts, explain clearly."""
llm_response = llm_chat(ai_prompt, temperature=0.7, max_tokens=600)
result["answer"] = f"""## ๐ฎ Predict
๐ **AI Knowledge**
{llm_response}
---
*๐ก This is general AI knowledge. For predictions on YOUR data, train a model first and ask about predictions.*"""
result["sources"] = ["AI Knowledge"]
exec_time = (datetime.now() - start_time).total_seconds()
result["execution_time"] = f"{exec_time:.2f}s"
return result
except Exception as e:
logger.error(f"AI Knowledge error: {e}")
# =================================================================
# ๐ ML PATH - Query IS about ML or user's data
# =================================================================
logger.info("๐ Predict: Routing to ML ANALYSIS")
# === CHECK AUTOML ===
if not AUTOML_AVAILABLE:
result["answer"] = "โ ๏ธ ML Engine not available. Please ensure the backend is properly configured."
return result
# === LOAD MODEL ===
try:
model_loaded = automl_engine.load(user_id)
if model_loaded:
logger.info(f"โ
Model loaded: {automl_engine.model_name}")
else:
logger.warning(f"โ ๏ธ No model found for user {user_id}")
except Exception as e:
logger.error(f"โ Model load error: {e}")
model_loaded = False
# === DETECT QUERY TYPE ===
query_type, type_confidence, params = detect_query_type(query, user_id)
logger.info(f"๐ Query: {query_type.value} (confidence: {type_confidence:.0%})")
result["confidence"] = type_confidence
# === ROUTE TO HANDLER ===
if not model_loaded:
result = _handle_no_model(query, query_type, df, user_id)
else:
handlers = {
# Core predictions
QueryType.PREDICTION: _handle_prediction,
QueryType.BATCH_PREDICTION: _handle_batch_prediction,
# Model understanding
QueryType.MODEL_INFO: _handle_model_info,
QueryType.METRICS: _handle_metrics,
QueryType.FEATURE_IMPORTANCE: _handle_feature_importance,
QueryType.FEATURE_SPECIFIC: _handle_feature_specific,
# Chart requests - pass user_id for stored charts
QueryType.SHOW_CHART: lambda q, d: _handle_show_chart(q, d, user_id),
QueryType.CONFUSION_MATRIX: lambda q, d: _handle_specific_chart(q, d, user_id, 'confusion_matrix'),
QueryType.ROC_CURVE: lambda q, d: _handle_specific_chart(q, d, user_id, 'roc_curve'),
QueryType.PRECISION_RECALL: lambda q, d: _handle_specific_chart(q, d, user_id, 'precision_recall'),
QueryType.ACTUAL_VS_PREDICTED: lambda q, d: _handle_specific_chart(q, d, user_id, 'actual_vs_predicted'),
# Advanced
QueryType.WHAT_IF: _handle_what_if,
QueryType.EXPLAIN: _handle_explain,
QueryType.DASHBOARD: lambda q, d: _handle_dashboard(q, d, user_id),
# Meta
QueryType.COMPARISON: _handle_comparison,
QueryType.HELP: _handle_help,
QueryType.GENERAL_ML: _handle_general_ml,
}
handler = handlers.get(query_type, _handle_general_ml)
try:
if query_type in [QueryType.CONFUSION_MATRIX, QueryType.ROC_CURVE,
QueryType.PRECISION_RECALL, QueryType.ACTUAL_VS_PREDICTED,
QueryType.DASHBOARD, QueryType.SHOW_CHART]:
result = handler(query, df)
elif query_type == QueryType.GENERAL_ML:
result = handler(query, df, context)
else:
result = handler(query, df)
except Exception as e:
logger.error(f"Handler error: {e}")
import traceback
traceback.print_exc()
result["answer"] = f"โ ๏ธ Error processing your request: {str(e)[:100]}"
# Update session memory
session.last_query_type = query_type
# Check for custom chart requests (radar, scatter, etc.) using LLM visualizer
q_lower = query.lower()
custom_chart_keywords = ['radar', 'spider', 'sunburst', 'treemap', 'bubble', 'scatter', 'heatmap', 'pink', 'blue', 'green']
wants_custom_chart = any(term in q_lower for term in custom_chart_keywords)
if wants_custom_chart and df is not None and not df.empty and "chart" not in result:
try:
from core.llm_visualizer import llm_visualize
import json
viz_result = llm_visualize(df, query, user_id)
if viz_result.get("success") and viz_result.get("chart"):
chart = viz_result.get("chart")
if isinstance(chart, dict) and 'data' in chart and 'layout' in chart:
chart_json = json.dumps(chart, default=str)
result["answer"] += f"\n\n```plotly_chart\n{chart_json}\n```"
result["chart"] = chart
result["visualization"] = chart
except Exception as e:
logger.warning(f"Predict custom chart failed: {e}")
# Add execution time
exec_time = (datetime.now() - start_time).total_seconds()
result["execution_time"] = f"{exec_time:.2f}s"
result["query_type"] = query_type.value
return result
# =============================================================================
# HANDLER: NO MODEL TRAINED
# =============================================================================
def _handle_no_model(query: str, query_type: QueryType, df=None, user_id: str = "") -> Dict[str, Any]:
"""Handle queries when no model is trained yet"""
response = """## ๐ค Senior AIML Engineer Assistant
Welcome! I'm your AI/ML assistant with senior engineer expertise.
### ๐ Getting Started
To unlock my full capabilities, you need to train a model first:
1. **Go to the AutoML tab** in the sidebar
2. **Upload your dataset** (CSV, Excel, or JSON)
3. **Select your target variable**
4. **Click Train** - I'll automatically:
- Clean and preprocess your data
- Try 15+ algorithms (including neural networks in Ultra mode)
- Optimize hyperparameters
- Select the best model
### ๐ก Once trained, ask me:
| Query Type | Example |
|------------|---------|
| **Predictions** | "Predict for age=60, income=50000" |
| **Model Info** | "What model was trained?" |
| **Performance** | "Show accuracy and confusion matrix" |
| **Feature Analysis** | "What features matter most?" |
| **Visualizations** | "Show me the ROC curve" |
| **What-If** | "What if age was 30 instead of 60?" |
| **Explanations** | "Why did the model predict this?" |
---
*Train a model and let's explore your data together!* ๐ฏ
"""
return {
"answer": response,
"mode": "predict",
"confidence": 0.5,
"sources": ["ML Assistant"],
"ml_used": False
}
# =============================================================================
# HANDLER: PREDICTIONS
# =============================================================================
def _handle_prediction(query: str, df=None) -> Dict[str, Any]:
"""Handle prediction requests with real ML model"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.85,
"sources": ["Trained ML Model", automl_engine.model_name],
"ml_used": True
}
# Extract features from query
feature_values = _extract_features_from_query(query, df)
if not feature_values:
result["answer"] = _get_prediction_help()
return result
try:
# Make real prediction
pred = automl_engine.predict(feature_values)
prediction = pred.get('prediction')
confidence = pred.get('confidence', 0.8)
probabilities = pred.get('probabilities', {})
# Get context
task_type = automl_engine.task_type
target = automl_engine.target_column
model_name = automl_engine.model_name
is_classification = 'classification' in task_type
# Generate senior-level explanation
ml_context = {
'model_name': model_name,
'task_type': task_type,
'target': target,
'prediction': prediction,
'confidence': confidence,
'features': feature_values
}
explanation = generate_senior_explanation('prediction', ml_context, query)
# Build response
conf_pct = int(confidence * 100) if confidence else 75
if is_classification:
pred_emoji = "โ
" if str(prediction) in ["0", "False", "No", "Negative", "Survived"] else "โ ๏ธ"
pred_display = _format_classification_prediction(prediction, target, probabilities)
else:
pred_emoji = "๐"
pred_display = f"**{target}**: {prediction:,.2f}" if isinstance(prediction, float) else f"**{target}**: {prediction}"
conf_bar = "๐ข" if conf_pct > 80 else "๐ก" if conf_pct > 60 else "๐ด"
response = f"""## {pred_emoji} ML Prediction Result
### {pred_display}
| Metric | Value |
|--------|-------|
| **Confidence** | {conf_bar} {conf_pct}% |
| **Model** | {model_name} |
| **Task** | {task_type.replace('_', ' ').title()} |
---
### ๐ก Senior ML Engineer Analysis
{explanation}
---
### ๐ฅ Input Features
"""
# Show input features
for feat, val in list(feature_values.items())[:10]:
if isinstance(val, float):
response += f"- `{feat}`: {val:,.2f}\n"
else:
response += f"- `{feat}`: {val}\n"
# Add probability breakdown for classification
if probabilities and is_classification:
response += "\n### ๐ Class Probabilities\n\n"
for cls, prob in probabilities.items():
bar_len = int(prob * 20)
bar = "โ" * bar_len + "โ" * (20 - bar_len)
response += f"- **{cls}**: `{bar}` {prob*100:.1f}%\n"
# Generate prediction chart
chart = _generate_prediction_confidence_chart(prediction, confidence, probabilities, target)
if chart:
response += f"\n\n```plotly_chart\n{json.dumps(chart)}\n```"
result["answer"] = response
result["confidence"] = confidence
except Exception as e:
logger.error(f"Prediction error: {e}")
import traceback
traceback.print_exc()
result["answer"] = f"โ ๏ธ Prediction failed: {str(e)}\n\nPlease check that your input features match the training data format."
return result
def _extract_features_from_query(query: str, df=None) -> Dict[str, Any]:
"""Extract feature values from natural language query"""
features = {}
# Get expected features
expected_features = automl_engine.feature_columns if hasattr(automl_engine, 'feature_columns') else []
metadata = automl_engine.feature_metadata if hasattr(automl_engine, 'feature_metadata') else []
meta_dict = {m.get('name', ''): m for m in metadata} if metadata else {}
q_lower = query.lower()
# Pattern: feature=value, feature:value, feature is value
patterns = [
r'([a-z_]+)\s*[=:]\s*([0-9.]+)',
r'([a-z_]+)\s*[=:]\s*([a-z]+)',
r'([a-z_]+)\s+is\s+([0-9.]+)',
r'([a-z_]+)\s+of\s+([0-9.]+)',
]
for pattern in patterns:
matches = re.findall(pattern, q_lower)
for name, value in matches:
# Match to expected feature
matched_feature = None
for feat in expected_features:
if name in feat.lower() or feat.lower() in name:
matched_feature = feat
break
if matched_feature:
# Convert value
meta = meta_dict.get(matched_feature, {})
if meta.get('type') == 'numeric':
try:
features[matched_feature] = float(value)
except:
features[matched_feature] = value
else:
features[matched_feature] = value
# Fill missing features with defaults
for feat in expected_features:
if feat not in features:
meta = meta_dict.get(feat, {})
if 'default' in meta:
features[feat] = meta['default']
elif 'mean' in meta:
features[feat] = meta['mean']
return features if len(features) >= len(expected_features) * 0.3 else {}
def _get_prediction_help() -> str:
"""Return help message for predictions"""
features = automl_engine.feature_columns[:8] if hasattr(automl_engine, 'feature_columns') else []
target = automl_engine.target_column if hasattr(automl_engine, 'target_column') else 'target'
response = f"""## ๐ฏ Making Predictions
To predict **{target}**, provide values for these features:
"""
for feat in features:
response += f"- `{feat}`\n"
response += f"""
### Example Queries:
1. **Simple**: "Predict for {features[0] if features else 'age'}=50"
2. **Multiple**: "{features[0] if features else 'age'}=50, {features[1] if len(features) > 1 else 'income'}=60000"
3. **Natural**: "What would happen if {features[0] if features else 'age'} is 30?"
---
*Tip: I'll fill in missing values with defaults from training data.*
"""
return response
def _format_classification_prediction(prediction, target: str, probabilities: Dict) -> str:
"""Format classification prediction for display"""
if 'death' in target.lower() or 'survive' in target.lower():
if str(prediction) in ['0', 'False', 'No']:
return "๐ข **Predicted: SURVIVED**"
else:
return "๐ด **Predicted: DEATH EVENT**"
elif 'fraud' in target.lower():
if str(prediction) in ['1', 'True', 'Yes']:
return "๐จ **Predicted: FRAUD DETECTED**"
else:
return "โ
**Predicted: LEGITIMATE**"
elif 'churn' in target.lower():
if str(prediction) in ['1', 'True', 'Yes']:
return "โ ๏ธ **Predicted: WILL CHURN**"
else:
return "โ
**Predicted: WILL STAY**"
else:
return f"**Predicted {target}**: {prediction}"
def _generate_prediction_confidence_chart(prediction, confidence, probabilities, target) -> Dict:
"""Generate confidence visualization chart"""
if probabilities and len(probabilities) > 0:
# Multi-class probability bar chart
classes = list(probabilities.keys())
probs = [probabilities[c] * 100 for c in classes]
return {
"data": [{
"type": "bar",
"x": classes,
"y": probs,
"marker": {"color": ["#22c55e" if str(prediction) == str(c) else "#94a3b8" for c in classes]}
}],
"layout": {
"title": {"text": f"Class Probabilities for {target}", "font": {"size": 14}},
"yaxis": {"title": "Probability (%)", "range": [0, 100]},
"paper_bgcolor": "#f8fafc",
"height": 300
}
}
else:
# Simple gauge for confidence
return {
"data": [{
"type": "indicator",
"mode": "gauge+number",
"value": confidence * 100,
"title": {"text": "Prediction Confidence"},
"gauge": {
"axis": {"range": [0, 100]},
"bar": {"color": "#3b82f6"},
"steps": [
{"range": [0, 60], "color": "#fee2e2"},
{"range": [60, 80], "color": "#fef3c7"},
{"range": [80, 100], "color": "#dcfce7"}
]
}
}],
"layout": {"paper_bgcolor": "#f8fafc", "height": 250}
}
# =============================================================================
# HANDLER: METRICS
# =============================================================================
def _handle_metrics(query: str, df=None) -> Dict[str, Any]:
"""Handle model metrics/performance queries"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.95,
"sources": ["Model Metrics", automl_engine.model_name],
"ml_used": True
}
try:
metrics_data = automl_engine.get_model_metrics()
except Exception as e:
result["answer"] = f"โ ๏ธ Error retrieving metrics: {e}"
return result
model = metrics_data.get('model_name', 'Unknown')
task = metrics_data.get('task_type', 'Unknown')
target = metrics_data.get('target', 'Unknown')
metrics = metrics_data.get('metrics', {})
cm = metrics_data.get('confusion_matrix')
response = f"""## ๐ Model Performance Dashboard
### ๐ค Model: **{model}**
| Property | Value |
|----------|-------|
| **Task Type** | {task.replace('_', ' ').title()} |
| **Target** | {target} |
| **Features** | {metrics_data.get('n_features', 'N/A')} |
---
### ๐ฏ Performance Metrics
"""
# Classification metrics
if metrics.get('accuracy'):
acc = metrics['accuracy'] * 100
acc_emoji = "๐ข" if acc > 85 else "๐ก" if acc > 70 else "๐ด"
response += f"| **Accuracy** | {acc_emoji} {acc:.1f}% |\n"
if metrics.get('f1'):
f1 = metrics['f1'] * 100
response += f"| **F1 Score** | {f1:.1f}% |\n"
if metrics.get('precision'):
response += f"| **Precision** | {metrics['precision']*100:.1f}% |\n"
if metrics.get('recall'):
response += f"| **Recall** | {metrics['recall']*100:.1f}% |\n"
if metrics.get('minority_recall'):
response += f"| **Minority Recall** | {metrics['minority_recall']*100:.1f}% |\n"
if metrics.get('auc_pr'):
response += f"| **AUC-PR** | {metrics['auc_pr']:.4f} |\n"
# Regression metrics
if metrics.get('r2'):
response += f"| **Rยฒ Score** | {metrics['r2']:.4f} |\n"
if metrics.get('mae'):
response += f"| **MAE** | {metrics['mae']:.4f} |\n"
if metrics.get('rmse'):
response += f"| **RMSE** | {metrics['rmse']:.4f} |\n"
# Senior ML Engineer interpretation
ml_context = {
'model_name': model,
'task_type': task,
'target': target,
'metrics': metrics
}
interpretation = generate_senior_explanation('metrics', ml_context, query)
response += f"\n---\n\n### ๐ก Senior ML Engineer Analysis\n\n{interpretation}\n"
# Add confusion matrix chart if available
if cm:
labels = ['Class 0', 'Class 1'] if len(cm) == 2 else [f'Class {i}' for i in range(len(cm))]
if 'death' in target.lower() and len(cm) == 2:
labels = ['Survived', 'Death']
chart = {
"data": [{
"type": "heatmap",
"z": cm,
"x": labels,
"y": labels,
"colorscale": [[0, "#dcfce7"], [0.5, "#22c55e"], [1, "#15803d"]],
"showscale": True,
"text": [[str(int(v)) for v in row] for row in cm],
"texttemplate": "%{text}",
"textfont": {"size": 18, "color": "#1f2937"}
}],
"layout": {
"title": {"text": "๐ Confusion Matrix", "font": {"size": 16}},
"xaxis": {"title": "Predicted"},
"yaxis": {"title": "Actual", "autorange": "reversed"},
"paper_bgcolor": "#f8fafc",
"height": 400, "width": 450
}
}
response += f"\n\n```plotly_chart\n{json.dumps(chart)}\n```"
result["answer"] = response
return result
# =============================================================================
# HANDLER: FEATURE IMPORTANCE
# =============================================================================
def _handle_feature_importance(query: str, df=None) -> Dict[str, Any]:
"""Handle feature importance queries"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.92,
"sources": ["Feature Analysis", automl_engine.model_name],
"ml_used": True
}
importance = _get_feature_importance_data()
target = automl_engine.target_column
model = automl_engine.model_name
if not importance:
result["answer"] = "โ ๏ธ Feature importance not available for this model type."
return result
response = f"""## ๐ Feature Importance Analysis
### What Drives **{target}** Predictions?
Based on the **{model}** model, these features have the highest predictive power:
"""
# Top features with bars
total_imp = sum(f['importance'] for f in importance)
for i, feat in enumerate(importance[:10], 1):
imp = feat['importance']
pct = (imp / total_imp * 100) if total_imp > 0 else 0
bar = "โ" * int(pct / 3) + "โ" * (30 - int(pct / 3))
response += f"**{i}. {feat['feature']}**\n`{bar}` {pct:.1f}%\n\n"
# Senior interpretation
top_3 = [f['feature'] for f in importance[:3]]
ml_context = {
'model_name': model,
'target': target,
'top_features': top_3
}
interpretation = generate_senior_explanation('importance', ml_context, query)
response += f"---\n\n### ๐ก Senior ML Engineer Analysis\n\n{interpretation}\n"
# Generate bar chart
feats = [f['feature'][:20] for f in importance[:8]]
vals = [f['importance'] * 100 for f in importance[:8]]
chart = {
"data": [{
"type": "bar",
"y": feats[::-1],
"x": vals[::-1],
"orientation": "h",
"marker": {"color": ["#3b82f6"] * (len(feats)-1) + ["#1d4ed8"]}
}],
"layout": {
"title": {"text": f"๐ Feature Importance", "font": {"size": 16}},
"xaxis": {"title": "Importance (%)"},
"paper_bgcolor": "#f8fafc",
"margin": {"l": 150},
"height": 400
}
}
response += f"\n\n```plotly_chart\n{json.dumps(chart)}\n```"
result["answer"] = response
return result
def _get_feature_importance_data() -> List[Dict]:
"""Get REAL feature importance from trained model - NO FAKE DATA"""
try:
model = automl_engine.model
feature_names = automl_engine.feature_columns
logger.info(f"Getting feature importance for model type: {type(model).__name__}")
importances = None
# 1. Direct feature_importances_ (tree-based models, HistGradientBoosting, etc.)
if hasattr(model, 'feature_importances_'):
importances = model.feature_importances_
logger.info(f"Got feature_importances_: {len(importances)} features")
# 2. Coefficients (linear models)
elif hasattr(model, 'coef_'):
importances = np.abs(model.coef_).flatten()
logger.info(f"Got coef_: {len(importances)} features")
# 3. Pipeline with final estimator
elif hasattr(model, 'named_steps'):
final_step = list(model.named_steps.values())[-1]
if hasattr(final_step, 'feature_importances_'):
importances = final_step.feature_importances_
logger.info(f"Got feature_importances_ from pipeline final step")
elif hasattr(final_step, 'coef_'):
importances = np.abs(final_step.coef_).flatten()
# 4. Ensemble with estimators_
elif hasattr(model, 'estimators_'):
valid_importances = []
for e in model.estimators_:
if hasattr(e, 'feature_importances_'):
valid_importances.append(e.feature_importances_)
if valid_importances:
importances = np.mean(valid_importances, axis=0)
logger.info(f"Got averaged importances from {len(valid_importances)} estimators")
# NO FALLBACK - Don't create fake/synthetic data
if importances is None:
logger.warning(f"No feature importance available for {type(model).__name__} - model doesn't support it")
return []
if len(importances) == 0:
logger.warning("Feature importance is empty")
return []
# Handle length mismatch
if len(importances) != len(feature_names):
logger.warning(f"Length mismatch: {len(importances)} importances vs {len(feature_names)} features")
min_len = min(len(importances), len(feature_names))
importances = importances[:min_len]
feature_names = feature_names[:min_len]
# Truncate or pad
if len(importances) > len(feature_names):
importances = importances[:len(feature_names)]
else:
# Use first N features
feature_names = feature_names[:len(importances)]
# Normalize
total = np.sum(importances)
if total > 0:
importances = importances / total
# Create list
result = [
{'feature': feat, 'importance': float(imp)}
for feat, imp in zip(feature_names, importances)
]
result.sort(key=lambda x: x['importance'], reverse=True)
logger.info(f"Returning {len(result)} feature importances, top: {result[0] if result else 'none'}")
return result
except Exception as e:
logger.error(f"Feature importance error: {e}")
import traceback
traceback.print_exc()
return []
# =============================================================================
# HANDLER: SPECIFIC CHARTS
# =============================================================================
def _handle_specific_chart(query: str, df, user_id: str, chart_type: str) -> Dict[str, Any]:
"""Handle requests for specific chart types - returns REAL charts from training"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.95,
"sources": ["Chart Generator", automl_engine.model_name],
"ml_used": True,
"charts": [],
"mlCharts": []
}
# Get stored charts first
stored_charts = get_stored_charts(user_id)
# Try to find matching chart from stored charts
chart = None
chart_data = None
# Map requested chart type to possible storage keys
chart_keys_mapping = {
'confusion_matrix': ['confusion_matrix', 'confusion', 'cm'],
'roc_curve': ['roc_curve', 'roc', 'auc_roc'],
'precision_recall': ['precision_recall', 'pr_curve', 'precision_recall_curve'],
'feature_importance': ['feature_importance', 'importance', 'features'],
'actual_vs_predicted': ['actual_vs_predicted', 'predictions', 'actual_predicted'],
'class_distribution': ['class_distribution', 'distribution', 'target_distribution'],
'residuals': ['residuals', 'residual_plot'],
'correlation': ['correlation', 'correlation_heatmap', 'corr'],
'metrics': ['metrics', 'performance', 'model_metrics']
}
# Find the chart
possible_keys = chart_keys_mapping.get(chart_type, [chart_type])
if stored_charts:
for key in possible_keys:
for stored_key, stored_data in stored_charts.items():
if key.lower() in stored_key.lower():
chart_data = stored_data
chart = stored_data if isinstance(stored_data, dict) else None
break
if chart_data:
break
# Fallback to get_chart_by_name
if not chart_data:
chart = get_chart_by_name(user_id, chart_type)
response = f"""## ๐ {chart_type.replace('_', ' ').title()}
"""
if chart_data:
# Check if it's a base64 image
if isinstance(chart_data, str):
if chart_data.startswith('data:image'):
# It's a base64 PNG image - add to mlCharts for frontend rendering
result["mlCharts"].append({
"type": f"ml_{chart_type}",
"image": chart_data,
"title": chart_type.replace('_', ' ').title()
})
response += f"Here is the **{chart_type.replace('_', ' ')}** from your trained model:\n\n"
response += f"*[Base64 chart rendered in dashboard - see ML Predictions page for interactive view]*\n"
elif chart_data.startswith('http'):
# It's a URL
response += f"\n"
else:
response += f"Chart data available but in unexpected format.\n"
elif isinstance(chart_data, dict):
# Check if Plotly chart
if 'data' in chart_data or 'type' in chart_data:
response += f"Here is the **{chart_type.replace('_', ' ')}** from your trained model:\n\n"
response += f"\n```plotly_chart\n{json.dumps(chart_data)}\n```"
result["charts"].append(chart_data)
else:
# It's metrics or other dict data
response += f"**{chart_type.replace('_', ' ').title()} Data:**\n\n"
for k, v in chart_data.items():
if isinstance(v, float):
response += f"- **{k}**: {v:.4f}\n"
else:
response += f"- **{k}**: {v}\n"
# Add interpretation
ml_context = {
'model_name': automl_engine.model_name,
'target': automl_engine.target_column,
'task_type': automl_engine.task_type,
'chart_type': chart_type
}
interpretation = generate_senior_explanation(chart_type, ml_context, query)
response += f"\n\n### ๐ก Interpretation\n\n{interpretation}"
elif chart:
# Got chart from get_chart_by_name
response += f"Here is the **{chart_type.replace('_', ' ')}** from your trained model:\n\n"
response += f"\n```plotly_chart\n{json.dumps(chart)}\n```"
result["charts"].append(chart)
ml_context = {
'model_name': automl_engine.model_name,
'target': automl_engine.target_column,
'chart_type': chart_type
}
interpretation = generate_senior_explanation(chart_type, ml_context, query)
response += f"\n\n### ๐ก Interpretation\n\n{interpretation}"
else:
# No chart found - generate one if possible
try:
metrics_data = automl_engine.get_model_metrics()
if chart_type == 'confusion_matrix' and metrics_data.get('confusion_matrix'):
cm = metrics_data['confusion_matrix']
labels = ['Class 0', 'Class 1'] if len(cm) == 2 else [f'Class {i}' for i in range(len(cm))]
target = automl_engine.target_column.lower()
if 'death' in target and len(cm) == 2:
labels = ['Survived', 'Death']
cm_chart = {
"data": [{
"type": "heatmap",
"z": cm,
"x": labels,
"y": labels,
"colorscale": [[0, "#dcfce7"], [0.5, "#22c55e"], [1, "#15803d"]],
"showscale": True,
"text": [[str(int(v)) for v in row] for row in cm],
"texttemplate": "%{text}",
"textfont": {"size": 18}
}],
"layout": {
"title": {"text": "Confusion Matrix"},
"xaxis": {"title": "Predicted"},
"yaxis": {"title": "Actual", "autorange": "reversed"},
"height": 400, "width": 450
}
}
response += f"\n```plotly_chart\n{json.dumps(cm_chart)}\n```"
result["charts"].append(cm_chart)
else:
response += f"""The **{chart_type.replace('_', ' ')}** chart was not found in storage.
**To get this chart:**
1. Go to **ML Predictions** page
2. Charts are generated during model training
3. Make sure you've trained a model with visualization enabled
**Available chart types:**
- Confusion Matrix
- ROC Curve
- Precision-Recall Curve
- Feature Importance
- Class Distribution
"""
except Exception as e:
logger.warning(f"Could not generate chart: {e}")
response += f"Chart not available. Visit **ML Predictions** page for full visualizations."
result["answer"] = response
return result
def _handle_show_chart(query: str, df=None, user_id: str = "") -> Dict[str, Any]:
"""Handle chart requests - shows REAL ML charts with explanations"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.95,
"sources": ["ML Charts", automl_engine.model_name],
"ml_used": True,
"charts": [],
"mlCharts": []
}
q = query.lower()
model = automl_engine.model_name
target = automl_engine.target_column
task = automl_engine.task_type
logger.info(f"๐ _handle_show_chart: model={model}, target={target}, task={task}, user_id={user_id}")
# Get model metrics (real from training)
try:
metrics_data = automl_engine.get_model_metrics()
metrics = metrics_data.get('metrics', {})
confusion_matrix = metrics_data.get('confusion_matrix')
except:
metrics = {}
confusion_matrix = None
# Get stored charts
stored_charts = get_stored_charts(user_id) if user_id else {}
logger.info(f"๐ Stored charts: {list(stored_charts.keys()) if stored_charts else 'None'}")
# Check what's available
has_confusion_matrix = confusion_matrix is not None and len(confusion_matrix) > 0
has_metrics = bool(metrics)
# Check feature importance support
try:
model_obj = automl_engine.model
has_feature_importance = hasattr(model_obj, 'feature_importances_') or hasattr(model_obj, 'coef_')
logger.info(f"๐ Model type: {type(model_obj).__name__}, has_feature_importances: {has_feature_importance}")
except:
has_feature_importance = False
# Build response
response = f"""## ๐ ML Charts - {model}
**Model**: {model}
**Target**: {target}
**Task**: {task.replace('_', ' ').title()}
---
"""
charts_added = 0
max_charts = 2 if ('two' in q or '2' in q) else 4
# =========================================================================
# CHART 1: Feature Importance
# =========================================================================
if charts_added < max_charts and has_feature_importance:
try:
importance = _get_feature_importance_data()
if importance and len(importance) > 0:
feats = [f['feature'][:25] for f in importance[:10]]
vals = [f['importance'] * 100 for f in importance[:10]]
fi_chart = {
"data": [{
"type": "bar",
"y": feats[::-1],
"x": vals[::-1],
"orientation": "h",
"marker": {"color": "#3b82f6"},
"text": [f"{v:.1f}%" for v in vals[::-1]],
"textposition": "outside"
}],
"layout": {
"title": {"text": "๐ Feature Importance", "font": {"size": 16}},
"xaxis": {"title": "Importance (%)"},
"margin": {"l": 180, "r": 60, "t": 50, "b": 50},
"height": 400,
"paper_bgcolor": "rgba(248,250,252,1)",
"plot_bgcolor": "rgba(248,250,252,1)"
}
}
response += f"### ๐ Feature Importance\n\n"
response += f"**Top features driving {target} predictions:**\n"
for i, feat in enumerate(importance[:5]):
response += f"- **{feat['feature']}**: {feat['importance']*100:.1f}%\n"
response += f"\n```plotly_chart\n{json.dumps(fi_chart)}\n```\n\n"
result["charts"].append(fi_chart)
charts_added += 1
logger.info(f"โ
Added Feature Importance chart with {len(importance)} features")
except Exception as e:
logger.error(f"Feature importance failed: {e}")
import traceback
traceback.print_exc()
# =========================================================================
# CHART 2: Metrics Chart
# =========================================================================
if charts_added < max_charts and has_metrics:
try:
metric_names = []
metric_values = []
metric_colors = []
if 'regression' in task.lower():
if 'r2' in metrics:
r2_val = metrics['r2']
metric_names.append('Rยฒ Score')
metric_values.append(r2_val * 100)
metric_colors.append('#22c55e' if r2_val > 0.7 else '#eab308' if r2_val > 0.5 else '#ef4444')
response += f"### ๐ Regression Metrics\n\n"
response += f"| Metric | Value | Interpretation |\n|--------|-------|----------------|\n"
if 'r2' in metrics:
r2 = metrics['r2']
interp = "Excellent" if r2 > 0.8 else "Good" if r2 > 0.6 else "Moderate" if r2 > 0.4 else "Weak"
response += f"| **Rยฒ Score** | {r2:.4f} | {interp} fit |\n"
if 'mae' in metrics:
response += f"| **MAE** | {metrics['mae']:,.2f} | Average error |\n"
if 'rmse' in metrics:
response += f"| **RMSE** | {metrics['rmse']:,.2f} | Root mean squared error |\n"
response += "\n"
else:
if 'accuracy' in metrics:
metric_names.append('Accuracy')
metric_values.append(metrics['accuracy'] * 100)
metric_colors.append('#22c55e')
if 'f1' in metrics:
metric_names.append('F1 Score')
metric_values.append(metrics['f1'] * 100)
metric_colors.append('#3b82f6')
if 'precision' in metrics:
metric_names.append('Precision')
metric_values.append(metrics['precision'] * 100)
metric_colors.append('#8b5cf6')
if 'recall' in metrics:
metric_names.append('Recall')
metric_values.append(metrics['recall'] * 100)
metric_colors.append('#ec4899')
response += f"### ๐ Classification Metrics\n\n"
if 'accuracy' in metrics:
acc = metrics['accuracy'] * 100
response += f"- **Accuracy**: {acc:.1f}%\n"
if 'f1' in metrics:
response += f"- **F1 Score**: {metrics['f1']*100:.1f}%\n"
response += "\n"
if metric_names:
metrics_chart = {
"data": [{
"type": "bar",
"x": metric_names,
"y": metric_values,
"marker": {"color": metric_colors},
"text": [f"{v:.1f}%" for v in metric_values],
"textposition": "outside"
}],
"layout": {
"title": {"text": "๐ Model Performance", "font": {"size": 16}},
"yaxis": {"title": "Score (%)", "range": [0, 110]},
"height": 350,
"paper_bgcolor": "rgba(248,250,252,1)",
"plot_bgcolor": "rgba(248,250,252,1)"
}
}
response += f"```plotly_chart\n{json.dumps(metrics_chart)}\n```\n\n"
result["charts"].append(metrics_chart)
charts_added += 1
logger.info(f"โ
Added Metrics chart")
except Exception as e:
logger.error(f"Metrics chart failed: {e}")
# =========================================================================
# CHART 3: Confusion Matrix (classification only)
# =========================================================================
if charts_added < max_charts and has_confusion_matrix and 'classification' in task.lower():
try:
cm = confusion_matrix
labels = ['Class 0', 'Class 1'] if len(cm) == 2 else [f'Class {i}' for i in range(len(cm))]
cm_chart = {
"data": [{
"type": "heatmap",
"z": cm,
"x": labels,
"y": labels,
"colorscale": [[0, "#dcfce7"], [0.5, "#22c55e"], [1, "#15803d"]],
"showscale": True,
"text": [[str(int(v)) for v in row] for row in cm],
"texttemplate": "%{text}",
"textfont": {"size": 18}
}],
"layout": {
"title": {"text": "๐ Confusion Matrix", "font": {"size": 16}},
"xaxis": {"title": "Predicted"},
"yaxis": {"title": "Actual", "autorange": "reversed"},
"height": 400, "width": 450,
"paper_bgcolor": "rgba(248,250,252,1)"
}
}
response += f"### ๐ Confusion Matrix\n\n"
response += f"```plotly_chart\n{json.dumps(cm_chart)}\n```\n\n"
result["charts"].append(cm_chart)
charts_added += 1
logger.info(f"โ
Added Confusion Matrix chart")
except Exception as e:
logger.error(f"Confusion matrix failed: {e}")
# =========================================================================
# CHART 4: Try stored charts
# =========================================================================
if stored_charts and charts_added < max_charts:
for chart_key, chart_data in stored_charts.items():
if charts_added >= max_charts:
break
try:
if isinstance(chart_data, str) and chart_data.startswith('data:image'):
result["mlCharts"].append({
"type": chart_key,
"image": chart_data,
"title": chart_key.replace('_', ' ').title()
})
response += f"### {chart_key.replace('_', ' ').title()}\n\n*[Stored chart from training]*\n\n"
charts_added += 1
elif isinstance(chart_data, dict) and 'data' in chart_data:
response += f"### {chart_key.replace('_', ' ').title()}\n\n"
response += f"```plotly_chart\n{json.dumps(chart_data)}\n```\n\n"
result["charts"].append(chart_data)
charts_added += 1
except:
continue
# =========================================================================
# SUMMARY
# =========================================================================
if charts_added == 0:
response = f"""## โ ๏ธ Charts Not Available for Your Model
**Model**: {model}
**Target**: {target}
**Task**: {task.replace('_', ' ').title()}
---
Your current model/data configuration does not support the requested charts:
"""
if not has_feature_importance:
response += f"- โ **Feature Importance**: The {model} model doesn't expose feature importances\n"
if not has_metrics:
response += "- โ **Metrics**: No performance metrics saved during training\n"
if 'regression' in task.lower():
response += "- โน๏ธ **Note**: Regression models don't have confusion matrix or ROC curves\n"
response += """
### ๐ก What You Can Do:
1. **View metrics**: Ask `Show model metrics` for Rยฒ, MAE, RMSE values
2. **Make predictions**: Ask `Predict for [feature]=value`
3. **Retrain**: Use AutoML tab with a different model that supports feature importance
"""
else:
response += f"\n---\n\nโ
**{charts_added} chart(s)** generated from your **{model}** model.\n\n"
# Add interpretation
ml_context = {
'model_name': model,
'task_type': task,
'target': target,
'metrics': metrics
}
interpretation = generate_senior_explanation('charts', ml_context, query)
response += f"### ๐ก Analysis\n\n{interpretation}\n"
result["answer"] = response
logger.info(f"๐ _handle_show_chart completed: {charts_added} charts")
return result
# =============================================================================
# HANDLER: DASHBOARD
# =============================================================================
def _handle_dashboard(query: str, df, user_id: str) -> Dict[str, Any]:
"""Generate comprehensive dashboard with REAL charts and metrics from trained model"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.98,
"sources": ["Full Dashboard", automl_engine.model_name],
"ml_used": True,
"charts": [], # Return charts array for frontend
"mlCharts": [] # For matplotlib base64 charts
}
model = automl_engine.model_name
target = automl_engine.target_column
task = automl_engine.task_type
# Get metrics
try:
metrics_data = automl_engine.get_model_metrics()
metrics = metrics_data.get('metrics', {})
cm = metrics_data.get('confusion_matrix')
except:
metrics = {}
cm = None
response = f"""## ๐ Complete ML Dashboard
### ๐ค Model Overview
| Property | Value |
|----------|-------|
| **Model** | {model} |
| **Target** | {target} |
| **Task** | {task.replace('_', ' ').title()} |
| **Features** | {len(automl_engine.feature_columns)} |
---
### ๐ฏ Key Metrics
"""
# Show all available metrics
if metrics.get('accuracy'):
acc = metrics['accuracy'] * 100
emoji = "๐ข" if acc > 85 else "๐ก" if acc > 70 else "๐ด"
response += f"| **Accuracy** | {emoji} {acc:.2f}% |\n"
if metrics.get('precision'):
response += f"| **Precision** | {metrics['precision']*100:.2f}% |\n"
if metrics.get('recall'):
response += f"| **Recall** | {metrics['recall']*100:.2f}% |\n"
if metrics.get('f1'):
response += f"| **F1 Score** | {metrics['f1']*100:.2f}% |\n"
if metrics.get('r2'):
response += f"| **Rยฒ Score** | {metrics['r2']:.4f} |\n"
if metrics.get('rmse'):
response += f"| **RMSE** | {metrics['rmse']:.4f} |\n"
if metrics.get('mae'):
response += f"| **MAE** | {metrics['mae']:.4f} |\n"
response += "\n---\n\n### ๐ Real Charts from Your Trained Model\n\n"
# Get real charts from storage
stored_charts = get_stored_charts(user_id)
chart_count = 0
if stored_charts:
# Priority charts to show
priority_charts = ['confusion_matrix', 'roc_curve', 'feature_importance', 'metrics',
'precision_recall', 'actual_vs_predicted', 'residuals', 'class_distribution']
for chart_name in priority_charts:
for stored_name, chart_data in stored_charts.items():
if chart_name in stored_name.lower() and chart_count < 4:
# Check if it's a base64 image or plotly chart
if isinstance(chart_data, str) and chart_data.startswith('data:image'):
# Base64 image chart
result["mlCharts"].append({
"type": f"ml_{chart_name}",
"image": chart_data,
"title": chart_name.replace('_', ' ').title()
})
response += f"๐ **{chart_name.replace('_', ' ').title()}** - Loaded from training\n\n"
chart_count += 1
elif isinstance(chart_data, dict) and ('data' in chart_data or 'type' in chart_data):
# Plotly chart
response += f"\n```plotly_chart\n{json.dumps(chart_data)}\n```\n"
result["charts"].append(chart_data)
chart_count += 1
break
# Show available charts summary
response += f"\n**{len(stored_charts)} charts available** from your training session.\n"
available_charts = list(stored_charts.keys())[:10]
response += "\nAvailable: " + ", ".join([c.replace('_', ' ').title() for c in available_charts])
else:
response += "โ ๏ธ No stored charts found. Visit **ML Predictions** page to view charts.\n"
# Add confusion matrix if available in metrics
if cm and chart_count < 4:
labels = ['Class 0', 'Class 1'] if len(cm) == 2 else [f'Class {i}' for i in range(len(cm))]
if 'death' in target.lower() and len(cm) == 2:
labels = ['Survived', 'Death']
cm_chart = {
"data": [{
"type": "heatmap",
"z": cm,
"x": labels,
"y": labels,
"colorscale": [[0, "#dcfce7"], [0.5, "#22c55e"], [1, "#15803d"]],
"showscale": True,
"text": [[str(int(v)) for v in row] for row in cm],
"texttemplate": "%{text}",
"textfont": {"size": 18, "color": "#1f2937"}
}],
"layout": {
"title": {"text": "๐ Confusion Matrix", "font": {"size": 16}},
"xaxis": {"title": "Predicted"},
"yaxis": {"title": "Actual", "autorange": "reversed"},
"paper_bgcolor": "rgba(0,0,0,0)",
"height": 400, "width": 450
}
}
response += f"\n\n```plotly_chart\n{json.dumps(cm_chart)}\n```"
result["charts"].append(cm_chart)
response += """
---
### ๐ก Quick Actions
- **Make Prediction**: "Predict for [feature]=value"
- **Show Chart**: "Show confusion matrix" or "Show ROC curve"
- **What-If Analysis**: "What if [feature] was [value]?"
- **Explain**: "Why did the model predict this?"
"""
result["answer"] = response
return result
# =============================================================================
# HANDLER: MODEL INFO
# =============================================================================
def _handle_model_info(query: str, df=None) -> Dict[str, Any]:
"""Handle model information queries"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.95,
"sources": ["Model Info"],
"ml_used": True
}
model = automl_engine.model_name
task = automl_engine.task_type
target = automl_engine.target_column
features = automl_engine.feature_columns
n_classes = getattr(automl_engine, 'n_classes', None)
response = f"""## ๐ค Trained Model Information
### Model Architecture
| Property | Value |
|----------|-------|
| **Algorithm** | {model} |
| **Task Type** | {task.replace('_', ' ').title()} |
| **Target Variable** | `{target}` |
| **Number of Features** | {len(features)} |
"""
if n_classes and 'classification' in task:
response += f"| **Number of Classes** | {n_classes} |\n"
response += f"""
---
### ๐ Features ({len(features)} total)
"""
for i, feat in enumerate(features[:15], 1):
response += f"{i}. `{feat}`\n"
if len(features) > 15:
response += f"\n*...and {len(features) - 15} more features*\n"
response += f"""
---
### ๐ก How to Use This Model
| Query Type | Example |
|------------|---------|
| **Predict** | "Predict for {features[0] if features else 'feature'}=50" |
| **Importance** | "What features matter most?" |
| **Metrics** | "Show model accuracy" |
| **What-If** | "What if {features[0] if features else 'feature'} was 100?" |
"""
result["answer"] = response
return result
# =============================================================================
# HANDLER: WHAT-IF SCENARIOS
# =============================================================================
def _handle_what_if(query: str, df=None) -> Dict[str, Any]:
"""Handle what-if scenario analysis using Causal Agent"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.95,
"sources": ["Causal Digital Twin Simulator", automl_engine.model_name],
"ml_used": True
}
if df is None or df.empty:
result["answer"] = "Data is required for What-If simulation."
return result
target = automl_engine.target_column
try:
from agents.causal_simulator import simulate_what_if
success, sim_df, params = simulate_what_if(query, df)
if not success:
result["answer"] = f"โ ๏ธ Causal Simulation failed. Could not parse valid parameters from your query. Error: {params.get('error', 'Unknown')}"
return result
col = params.get('manipulated_variable')
op = params.get('operation')
val = params.get('value')
# We predict on the mean/mode of the baseline vs the simulated dataset
# For a truly accurate simulation, we should aggregate the predictions across all rows
# 1. Baseline Predictions
baseline_preds = automl_engine.batch_predict(df)
baseline_mean = baseline_preds['prediction'].mean() if pd.api.types.is_numeric_dtype(baseline_preds['prediction']) else baseline_preds['prediction'].mode()[0]
# 2. Simulated Predictions
sim_preds = automl_engine.batch_predict(sim_df)
sim_mean = sim_preds['prediction'].mean() if pd.api.types.is_numeric_dtype(sim_preds['prediction']) else sim_preds['prediction'].mode()[0]
# Calculate Delta
if isinstance(baseline_mean, (int, float)):
delta = sim_mean - baseline_mean
delta_pct = (delta / baseline_mean) * 100 if baseline_mean != 0 else 0
change_str = f"{'+' if delta > 0 else ''}{delta:.2f} ({delta_pct:.1f}%)"
else:
change_str = f"Changed to {sim_mean}"
response = f"""## ๐ฎ Causal AI "What-If" Simulation
### Simulating: {op.title()} **{col}** by **{val}**
I have built a Digital Twin of your dataset and applied your requested manipulation. Here is how it affects the target variable (**{target}**) across your entire dataset:
| Scenario | Average {target} Prediction | Delta |
|----------|------------|------------|
| **Baseline (Original)** | {baseline_mean:.2f} | - |
| **Simulated (What-If)** | {sim_mean:.2f} | **{change_str}** |
---
### ๐ก Causal Insight
"""
# LLM interpretation
ml_context = {
'manipulation': f"{op} {col} by {val}",
'target': target,
'baseline_pred': baseline_mean,
'simulated_pred': sim_mean,
'delta': change_str
}
interpretation = generate_senior_explanation('what_if', ml_context, query)
response += interpretation
# Chart
chart = {
"data": [{
"type": "bar",
"x": ["Baseline", "Simulated"],
"y": [baseline_mean, sim_mean],
"marker": {"color": ["#64748b", "#3b82f6"]}
}],
"layout": {
"title": {"text": f"Simulation Impact on {target}", "font": {"size": 14}},
"yaxis": {"title": f"Predicted {target}"},
"paper_bgcolor": "#f8fafc",
"height": 300
}
}
response += f"\n\n```plotly_chart\n{json.dumps(chart)}\n```"
result["answer"] = response
except Exception as e:
result["answer"] = f"โ ๏ธ What-If analysis failed: {e}"
return result
# =============================================================================
# HANDLER: EXPLAIN
# =============================================================================
def _handle_explain(query: str, df=None) -> Dict[str, Any]:
"""Handle explanation requests"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.85,
"sources": ["Model Explainer", automl_engine.model_name],
"ml_used": True
}
model = automl_engine.model_name
target = automl_engine.target_column
# Get feature importance for context
importance = _get_feature_importance_data()[:5]
response = f"""## ๐ Model Explanation
### How does **{model}** predict **{target}**?
"""
if importance:
response += "### Key Decision Factors\n\n"
for i, feat in enumerate(importance, 1):
response += f"{i}. **{feat['feature']}**: {feat['importance']*100:.1f}% influence\n"
response += "\n---\n\n### ๐ก Senior ML Engineer Analysis\n\n"
ml_context = {
'model_name': model,
'target': target,
'top_features': [f['feature'] for f in importance] if importance else []
}
explanation = generate_senior_explanation('explain', ml_context, query)
response += explanation
response += """
---
### ๐ฌ For Deeper Analysis
- Ask: "Show SHAP values" (if available)
- Ask: "Why did the model predict X?"
- View the **ML Predictions** page for interactive explanations
"""
result["answer"] = response
return result
# =============================================================================
# HANDLER: COMPARISON
# =============================================================================
def _handle_comparison(query: str, df=None) -> Dict[str, Any]:
"""Handle comparison queries between scenarios"""
# Reuse what-if handler
return _handle_what_if(query, df)
# =============================================================================
# HANDLER: HELP
# =============================================================================
def _handle_help(query: str, df=None) -> Dict[str, Any]:
"""Handle help queries"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.95,
"sources": ["Help"],
"ml_used": False
}
target = automl_engine.target_column if hasattr(automl_engine, 'target_column') else 'outcome'
response = f"""## ๐ค ML Assistant Help
I'm your Senior AIML Engineer assistant. Here's what I can do:
### ๐ฏ Make Predictions
```
"Predict for age=60, income=50000"
"What would happen if credit_score is 750?"
```
### ๐ View Model Performance
```
"Show accuracy"
"Show confusion matrix"
"How well does the model perform?"
```
### ๐ Understand Features
```
"What features are most important?"
"How does age affect {target}?"
```
### ๐ View Charts
```
"Show ROC curve"
"Show precision-recall curve"
"Display feature importance chart"
```
### ๐ฎ Scenario Analysis
```
"What if age was 30 instead of 60?"
"Compare high vs low income scenarios"
```
### ๐ก Get Explanations
```
"Why did the model predict this?"
"Explain how the model works"
```
---
*I provide senior ML engineer level insights with statistical depth!*
"""
result["answer"] = response
return result
# =============================================================================
# HANDLER: GENERAL ML
# =============================================================================
def _handle_general_ml(query: str, df=None, context: str = "") -> Dict[str, Any]:
"""Handle general ML questions with LLM"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.75,
"sources": ["ML Knowledge"],
"ml_used": True
}
if not LLM_AVAILABLE:
result["answer"] = "LLM not available for general questions. Try specific queries like 'show accuracy' or 'predict for age=50'."
return result
# Build context
model_context = ""
if hasattr(automl_engine, 'model_name'):
model_context = f"""
Current trained model: {automl_engine.model_name}
Target: {automl_engine.target_column}
Task: {automl_engine.task_type}
Features: {len(automl_engine.feature_columns)}
"""
prompt = f"""{SENIOR_AIML_SYSTEM_PROMPT}
{model_context}
User question: {query}
Provide a helpful, technically accurate response. If the question relates to the trained model, incorporate that context. Be concise but thorough."""
try:
response = llm_chat(prompt, temperature=0.4, max_tokens=400)
result["answer"] = f"## ๐ก ML Insight\n\n{response}"
except Exception as e:
result["answer"] = f"โ ๏ธ Error generating response: {e}"
return result
# =============================================================================
# HANDLER: BATCH PREDICTION
# =============================================================================
def _handle_batch_prediction(query: str, df=None) -> Dict[str, Any]:
"""Handle batch prediction requests"""
result = {
"answer": "Batch prediction feature coming soon! For now, use the ML Predictions page to make predictions on uploaded data.",
"mode": "predict",
"confidence": 0.7,
"sources": [],
"ml_used": False
}
return result
# =============================================================================
# HANDLER: FEATURE SPECIFIC
# =============================================================================
def _handle_feature_specific(query: str, df=None) -> Dict[str, Any]:
"""Handle specific feature analysis"""
result = {
"answer": "",
"mode": "predict",
"confidence": 0.90,
"sources": ["Feature Analysis"],
"ml_used": True
}
# Find mentioned feature
features = automl_engine.feature_columns
q = query.lower()
target_feat = None
for feat in features:
if feat.lower() in q or feat.lower().replace('_', ' ') in q:
target_feat = feat
break
if not target_feat:
result["answer"] = f"Please specify which feature you want to analyze. Available: {', '.join(features[:10])}"
return result
# Get metadata
metadata = {m['name']: m for m in automl_engine.feature_metadata}
meta = metadata.get(target_feat, {})
target = automl_engine.target_column
response = f"""## ๐ Feature Analysis: **{target_feat}**
### ๐ Statistics
"""
if meta.get('type') == 'numeric':
response += f"""| Statistic | Value |
|-----------|-------|
| **Type** | Numeric |
| **Min** | {meta.get('min', 'N/A'):,.1f} |
| **Max** | {meta.get('max', 'N/A'):,.1f} |
| **Mean** | {meta.get('mean', 'N/A'):,.1f} |
"""
else:
options = meta.get('options', [])[:5]
response += f"""| Property | Value |
|----------|-------|
| **Type** | Categorical |
| **Unique Values** | {len(meta.get('options', []))} |
| **Examples** | {', '.join(str(o) for o in options)} |
"""
# Get importance rank
importance = _get_feature_importance_data()
rank = next((i+1 for i, f in enumerate(importance) if f['feature'] == target_feat), None)
if rank:
response += f"\n### ๐ Importance Rank: #{rank} of {len(features)}\n"
# LLM explanation
ml_context = {
'feature': target_feat,
'target': target,
'stats': meta
}
explanation = generate_senior_explanation('feature_analysis', ml_context, query)
response += f"\n---\n\n### ๐ก Analysis\n\n{explanation}"
result["answer"] = response
return result
# =============================================================================
# EXPORT - For use by other modules
# =============================================================================
__all__ = [
'predict_response_sync',
'detect_query_type',
'QueryType',
'get_stored_charts',
'get_chart_by_name'
]
|