Spaces:
Sleeping
Sleeping
File size: 65,151 Bytes
7e184bf 2e1d581 7cf5031 4f3da4f 7e184bf 7cf5031 bed229a 7e184bf bed229a 7cf5031 bf8f953 4f3da4f 4a2c0eb 7cf5031 5386e26 1160b9b 71621f7 158215f 36dcbf4 5386e26 7cf5031 2e1d581 5386e26 bed229a 5386e26 5a47613 5386e26 4f3da4f 7e184bf 85c9819 bef09db ea37944 7cf5031 1ba6ce2 7cf5031 bed229a 7cf5031 158215f 1ae9a08 ea37944 1ae9a08 775fb04 85c9819 4a2c0eb 7904f75 1160b9b 158215f bed229a 71621f7 7cf5031 bed229a 97d9a2e bed229a 7cf5031 97d9a2e 7cf5031 97d9a2e 29a29d5 5386e26 4a2c0eb 158215f 29a29d5 7cf5031 f4b3c44 7cf5031 5386e26 7cf5031 4a325e1 7cf5031 4a325e1 bf8f953 7cf5031 7904f75 1ba6ce2 8e7137e 7e184bf 7904f75 7e184bf 7138767 f4b3c44 7e184bf 7cf5031 85c9819 7cf5031 bf8f953 85c9819 29a29d5 5a47613 85c9819 4f3da4f 4a2c0eb 4f3da4f 5386e26 1160b9b 7cf5031 7e184bf 7cf5031 bf8f953 29a29d5 5a47613 7cf5031 7e184bf 7cf5031 4a325e1 4a2c0eb 4a325e1 7e184bf 7cf5031 1e2aaef dfcb128 bed229a 4a325e1 dfcb128 4a325e1 7cf5031 7e184bf 4a325e1 dfcb128 4a325e1 7cf5031 4a325e1 7cf5031 080a6ff dfcb128 080a6ff 7cf5031 dfcb128 7cf5031 4a325e1 7cf5031 942e3f2 5386e26 942e3f2 7cf5031 4a325e1 942e3f2 dfcb128 4a325e1 7e184bf 7cf5031 64d4d94 7cf5031 942e3f2 64d4d94 97d9a2e 70be945 7cf5031 942e3f2 64d4d94 942e3f2 97d9a2e 942e3f2 5386e26 942e3f2 97d9a2e 942e3f2 97d9a2e 942e3f2 97d9a2e 64d4d94 36dcbf4 64d4d94 942e3f2 64d4d94 7cf5031 85c9819 bed229a 7cf5031 dfcb128 bed229a dfcb128 0f4e297 dfcb128 bed229a 0f4e297 dfcb128 0f4e297 dfcb128 0f4e297 dfcb128 0f4e297 dfcb128 0f4e297 dfcb128 0f4e297 7cf5031 5a47613 dfcb128 29a29d5 dfcb128 5a47613 dfcb128 5a47613 dfcb128 5a47613 dfcb128 5a47613 dfcb128 29a29d5 85c9819 dfcb128 85c9819 dfcb128 85c9819 4f3da4f dfcb128 4f3da4f dfcb128 158215f dfcb128 158215f dfcb128 4f3da4f 5386e26 dfcb128 5386e26 dfcb128 5386e26 dfcb128 5386e26 1160b9b dfcb128 1160b9b dfcb128 158215f dfcb128 158215f dfcb128 1160b9b 158215f dfcb128 158215f dfcb128 158215f dfcb128 158215f dfcb128 158215f dfcb128 158215f dfcb128 158215f dfcb128 158215f dfcb128 158215f dfcb128 158215f dfcb128 158215f dfcb128 158215f 36dcbf4 dfcb128 36dcbf4 dfcb128 36dcbf4 7cf5031 85c9819 7cf5031 bf8f953 5a47613 85c9819 4f3da4f 5386e26 1160b9b 158215f 36dcbf4 7cf5031 bed229a 7e184bf 7cf5031 7e184bf 7cf5031 bf8f953 5a47613 85c9819 4f3da4f 5386e26 1160b9b 158215f 36dcbf4 7e184bf bed229a 7cf5031 bed229a 7cf5031 7e184bf 7cf5031 bed229a 7cf5031 bed229a 7cf5031 bed229a 7cf5031 bf8f953 5a47613 85c9819 4f3da4f 5386e26 1160b9b 158215f 36dcbf4 7cf5031 bed229a 7cf5031 bf8f953 5a47613 85c9819 4f3da4f 5386e26 1160b9b 158215f 36dcbf4 bed229a eae50b4 7e184bf bef09db 7e184bf 7cf5031 4a325e1 97d9a2e bef09db 7cf5031 bef09db 7e184bf 4a2c0eb 7e184bf 4a2c0eb 7cf5031 4a325e1 4a2c0eb 7e184bf 4a325e1 bed229a 4a2c0eb 7e184bf bed229a 4a325e1 bed229a 7cf5031 158215f 7cf5031 4a325e1 7cf5031 bed229a 7cf5031 158215f 7cf5031 bed229a 7e184bf bef09db 32b8e9c bed229a bf8f953 |
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 |
import os
from dotenv import load_dotenv
from typing import TypedDict, Annotated, Dict, Any, Optional, Union, List
from pathlib import Path
from langgraph.graph.message import add_messages
from langchain_core.messages import AnyMessage, HumanMessage, AIMessage, SystemMessage, ToolMessage
from langgraph.graph import START, END, StateGraph
from langchain_openai import ChatOpenAI
import tempfile
import random
import json
import requests
from urllib.parse import quote, urlparse
from bs4 import BeautifulSoup
import html2text
import pandas as pd
from tabulate import tabulate
import base64
# Import all tool functions from tools.py
from tools import (
tools_config,
run_python_code,
scrape_webpage,
wikipedia_search,
tavily_search,
arxiv_search,
supabase_operation,
excel_to_text,
save_attachment_to_tempfile,
process_youtube_video,
transcribe_audio,
extract_python_code_from_complex_input,
process_image,
read_file,
process_online_document
)
load_dotenv()
# Remove the following functions from agent.py since they're now imported from tools.py:
# - run_python_code (lines ~28-175)
# - scrape_webpage (lines ~177-310)
# - wikipedia_search (lines ~345-405)
# - tavily_search (lines ~407-470)
# - arxiv_search (lines ~472-535)
# - supabase_operation (lines ~537-620)
# - excel_to_text (lines ~622-690)
# - save_attachment_to_tempfile (lines ~1680-1706)
# Also remove the tools_config definition (lines ~795-870) since it's imported from tools.py
# The rest of the file remains the same...
# System prompt to guide the model's behavior
#web_search: Search the google search engine when Tavily Search and Wikipedia Search do not return a result. Provide a specific search query.
#webpage_scrape: Scrape content from a specific webpage URL when Tavily Search and Wikipedia Search do not return a result. Provide a valid URL to extract information from a particular web page.
#Give preference to using Tavily Search and Wikipedia Search before using web_search or webpage_scrape. When Web_search does not return a result, use Tavily Search.
SYSTEM_PROMPT = """ You are a genuis deep reseach assistant called ScholarAI, made by Vividh Mahajan. Answer the following questions as best you can. If it is a basic question, answer it using your internal knowledge. If it is a complex question that requires facts, use the tools to answer it DO NOT rely on your internal knowledge unless the tools fail to provide a result:
For simple questions, you can use your internal knowledge and answer directly. If you do not understand the question, ask for clarification after trying to answer the question yourself.
The way you use the tools is by specifying a json blob. These are the only tools you can use:
Specifically, this json should have an `action` key (with the name of the tool to use) and an `action_input` key (with the input to the tool going here).
The only values that should be in the "action" field are:
python_code: Execute Python code. Use this tool to calculate math problems. make sure to use prints to be able to view the final result. args: {"code": {"type": "string"}}
wikipedia_search: Search Wikipedia for information about a specific topic. Optionally specify the number of results to return, args: {"query": {"type": "string"}, "num_results": {"type": "integer", "optional": true}}
tavily_search: Search the web using Tavily. Optionally specify search_depth as 'basic' or 'comprehensive'. DO NOT use Tavily Search for basic questions that you can answer using your internal knowledge especially if the question is not about factual information. DO NOT use Tavily Search for clarifiction. args: {"query": {"type": "string"}, "search_depth": {"type": "string", "optional": true}}
arxiv_search: Search ArXiv for publications,news and other resources. Optionally specify max_results to control the number of papers returned, args: {"query": {"type": "string"}, "max_results": {"type": "integer", "optional": true}}
webpage_scrape: Scrape a specific webpage, args: {"url": {"type": "string"}}
supabase_operation: Perform database operations, args: {"operation_type": {"type": "string"}, "table": {"type": "string"}, "data": {"type": "object", "optional": true}, "filters": {"type": "object", "optional": true}}
excel_to_text: Convert Excel to Markdown table with attachment, args: {"excel_path": {"type": "string"}, "file_content": {"type": "string"}, "sheet_name": {"type": "string", "optional": true}}
process_youtube_video: Process a YouTube video by extracting its transcript/captions and basic metadata by providing the video URL. Returns video metadata and transcript, args: {"url": {"type": "string"}, "summarize": {"type": "boolean", "optional": true}}
transcribe_audio: Transcribe audio files using OpenAI Whisper, args: {"audio_path": {"type": "string"}, "file_content": {"type": "string", "optional": true}, "language": {"type": "string", "optional": true}}
process_image: Process and analyze image files, args: {"image_path": {"type": "string"}, "image_url": {"type": "string", "optional": true}, "file_content": {"type": "string", "optional": true}, "analyze_content": {"type": "boolean", "optional": true}}
read_file: Read and display the contents of a text file, args: {"file_path": {"type": "string"}, "file_content": {"type": "string", "optional": true}, "line_start": {"type": "integer", "optional": true}, "line_end": {"type": "integer", "optional": true}}
If you get stuck, try using another tool. For example if you are unable to find relevant information from the tavily_search tool, try using the wikipedia_search tool and vice versa.
IMPORTANT: Make sure your JSON is properly formatted with double quotes around keys and string values.
Example use for tools:
```json
{
"action": "tavily_search",
"action_input": {"query": "What is the capital of France?", "search_depth": "basic"}
}
```
or
```json
{
"action": "process_youtube_video",
"action_input": {"url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ", "summarize": true}
}
```
or
```json
{
"action": "process_image",
"action_input": {"image_path": "example.jpg", "analyze_content": true}
}
```
ALWAYS follow this specific format for your responses. Your entire response will follow this pattern:
Question: [the user's question]
Thought: [your reasoning about what to do next, break it down into smaller steps and clearly state your thoughts]
Action:
```json
{
"action": "[tool_name]",
"action_input": {"[parameter_name]": "[parameter_value]"}
}
```
Observation: [the result from the tool will appear here]
Thought: [your reasoning after seeing the observation, break it down into smaller steps]
Action:
```json
{
"action": "[tool_name]",
"action_input": {"[parameter_name]": "[parameter_value]"}
}
```
Observation: [another tool result will appear here]
IMPORTANT: You MUST strictly follow the ReAct pattern (Reasoning, Action, Observation):
1. First reason about the problem in the "Thought" section
2. Then decide what action to take in the "Action" section (using the tools)
3. Wait for an observation from the tool
4. Based on the observation, continue with another thought
5. This cycle repeats until you have enough information to provide a final answer
NEVER fake or simulate tool output yourself. If you are unable to make progreess in a certain way, try a different tool or a different approach.
... (this Thought/Action/Observation cycle can repeat as needed) ...
Thought: I now know the final answer
Final Answer: Make sure to follow any formatting instructions given by the user. Do not give too long of an answer.
If you are unable arrive at a final answer, summarize the information you have gathered such as links and any relevent information and provide a final answer. Also state a reason for you not being able to answer the question.
Now begin! Reminder to ALWAYS use the exact characters `Final Answer:` when you provide a definitive answer. DO NOT USE TAVILY SEARCH FOR CLARIFICATION OR BASIC QUESTIONS THAT YOU CAN ANSWER USING YOUR INTERNAL KNOWLEDGE"""
#YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings. If you are asked for a number, don't use comma to write your number neither use units such as $ or percent sign unless specified otherwise. If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise. If you are asked for a comma separated list, apply the above rules depending of whether the element to be put in the list is a number or a string. For one word answers, start with the word with a capital letter.
# Generate the chat interface, including the tools
llm = ChatOpenAI(
model="gpt-4o-mini",
temperature=0.1
)
chat = llm
# Tools are defined but not bound to the LLM here
tools_config = [
# {
# "name": "web_search",
# "description": "Search the web for current information. Provide a specific search query in the format: {\"query\": \"your search query here\"}",
# "func": safe_web_search
# },
{
"name": "python_code",
"description": "Execute Python code. Provide the complete Python code as a string in the format: {\"code\": \"your python code here\"}",
"func": run_python_code
},
# {
# "name": "webpage_scrape",
# "description": "Scrape content from a specific webpage URL. Provide a valid URL in the format: {\"url\": \"https://example.com\"}",
# "func": scrape_webpage
# },
{
"name": "wikipedia_search",
"description": "Search Wikipedia for information about a specific topic. Provide a query in the format: {\"query\": \"your topic\", \"num_results\": 3}",
"func": wikipedia_search
},
{
"name": "tavily_search",
"description": "Search the web using Tavily for more comprehensive results. Provide a query in the format: {\"query\": \"your search query\", \"search_depth\": \"basic\"}",
"func": tavily_search
},
{
"name": "arxiv_search",
"description": "Search ArXiv for scientific papers. Provide a query in the format: {\"query\": \"your research topic\", \"max_results\": 5}",
"func": arxiv_search
},
{
"name": "supabase_operation",
"description": "Perform database operations on Supabase (insert, select, update, delete). Provide operation_type, table name, and optional data/filters. ",
"func": supabase_operation
},
{
"name": "excel_to_text",
"description": "Read an Excel file and return a Markdown table. You can provide either the path to an Excel file or use a file attachment. For attachments, provide a base64-encoded string of the file content and a filename.",
"func": excel_to_text
},
{
"name": "process_youtube_video",
"description": "Extract and analyze YouTube video content by providing the video URL. Returns video metadata and transcript.",
"func": process_youtube_video
},
{
"name": "transcribe_audio",
"description": "Transcribe audio files using OpenAI Whisper. You can provide either a file path or use a file attachment. For attachments, provide base64-encoded content. Optionally specify language for better accuracy.",
"func": transcribe_audio
}
]
# Instead of binding tools, we'll handle the JSON parsing ourselves
# chat_with_tools = chat.bind_tools([Tool(**tool) for tool in tools_config])
chat_with_tools = chat
# Generate the AgentState and Agent graph
class ActionInput(TypedDict, total=False):
query: Optional[str]
code: Optional[str]
url: Optional[str]
image_url: Optional[str]
num_results: Optional[int]
search_depth: Optional[str]
max_results: Optional[int]
class AgentState(TypedDict, total=False):
messages: Annotated[list[AnyMessage], add_messages]
current_tool: Optional[str]
action_input: Optional[ActionInput]
iteration_count: int # Added to track iterations
attachments: Dict[str, str] # Added to store file attachments (filename -> base64 content)
# tool_call_id: Optional[str] # Ensure this is present if used by your graph logic for tools
# Add prune_messages_for_llm function
def prune_messages_for_llm(
full_history: List[AnyMessage],
num_recent_to_keep: int = 6 # Keeps roughly 2-3 ReAct turns (Thought/Action, Observation)
) -> List[AnyMessage]:
"""
Prunes the message history for the LLM call.
This function expects a 'core' history (messages without the initial SystemMessage).
It keeps the first HumanMessage (original query) and the last `num_recent_to_keep` messages
from this core history, injecting a condensation note.
"""
if not full_history: # full_history here is actually core_history
return []
first_human_message: Optional[HumanMessage] = None
for msg in full_history: # Iterate over the provided core_history
if isinstance(msg, HumanMessage):
first_human_message = msg
break
# If history is too short or no initial human query found in core_history,
# return core_history as is. The calling function (assistant) will prepend SystemMessage.
# Threshold considers: first_human (1) + condensation_note (1) + num_recent_to_keep
if first_human_message is None or len(full_history) < (1 + 1 + num_recent_to_keep):
return full_history
# Pruning is needed for the core_history
recent_messages_from_core = full_history[-num_recent_to_keep:]
pruned_core_list: List[AnyMessage] = []
# Add the first human message
pruned_core_list.append(first_human_message)
# Add condensation note
pruned_core_list.append(
AIMessage(content="[System note: To manage context length, earlier parts of the conversation have been omitted. The original query and the most recent interactions are preserved.]")
)
# Add recent messages, ensuring not to duplicate the first_human_message if it's in the recent slice
for msg in recent_messages_from_core:
if msg is not first_human_message: # Check object identity
pruned_core_list.append(msg)
return pruned_core_list
def assistant(state: AgentState) -> Dict[str, Any]:
"""Assistant node that processes messages and decides on next action."""
from langchain_core.messages import AIMessage # Add import at the start of the function
print("\n=== Assistant Node ===")
full_current_history = state["messages"]
iteration_count = state.get("iteration_count", 0)
iteration_count += 1 # Increment for the current call
print(f"Iteration: {iteration_count}")
# Prepare messages for the LLM
system_msg = SystemMessage(content=SYSTEM_PROMPT)
# Core history excludes any SystemMessages found in the accumulated history.
# The pruning function operates on this core history.
core_history = [msg for msg in full_current_history if not isinstance(msg, SystemMessage)]
llm_input_core_messages: List[AnyMessage]
# Prune if it's time (e.g., after every 5th completed iteration, so check for current iteration 6, 11, etc.)
# Iteration 1-5: no pruning. Iteration 6: prune.
if iteration_count > 5 and (iteration_count - 1) % 5 == 0:
print(f"Pruning message history at iteration {iteration_count}")
llm_input_core_messages = prune_messages_for_llm(core_history, num_recent_to_keep=6)
else:
llm_input_core_messages = core_history
# Combine system message with the (potentially pruned) core messages
messages_for_llm = [system_msg] + llm_input_core_messages
# Get response from the assistant
try:
response = chat_with_tools.invoke(messages_for_llm, stop=["Observation:"])
# Check for empty response
if response is None or not hasattr(response, 'content') or not response.content or len(response.content.strip()) < 20:
print("Warning: Received empty or very short response from LLM")
# Create a new response object if it's None
if response is None:
response = AIMessage(content="")
print("Created new AIMessage object for None response")
# Check what kind of information we have in the last observation
last_observation = None
for msg in reversed(core_history):
if isinstance(msg, AIMessage) and "Observation:" in msg.content:
last_observation = msg.content
break
# Create an appropriate fallback response
if last_observation and "python_code" in state.get("current_tool", ""):
print("Creating fallback response for empty response after Python code execution")
fallback_content = (
"Thought: I've analyzed the results of the code execution. Based on the observations, "
"I need to determine the final answer.\n\n"
"Final Answer: "
)
# Look for common patterns in Python output that might indicate results
import re
result_match = re.search(r'(\[.+?\]|\{.+?\}|[A-Za-z,\s]+)$', last_observation)
if result_match:
fallback_content += result_match.group(1).strip()
else:
# Generic fallback based on the executed code
fallback_content += "Based on the code analysis, but the exact result was unclear from the output."
else:
# Generic fallback suggesting a reasonable next search
print("Creating generic fallback response for empty LLM response")
# Make sure we have a valid query
query = ""
if full_current_history and hasattr(full_current_history[0], 'content'):
query = full_current_history[0].content[:100] # Limit to first 100 chars
else:
query = "additional information for this question"
fallback_content = (
"Thought: I need more information to answer this question correctly. Let me search for additional details.\n\n"
"Action: \n```json\n"
"{\n"
' "action": "tavily_search",\n'
f' "action_input": {{"query": "{query}", "search_depth": "comprehensive"}}\n'
"}\n```"
)
# Use our fallback content instead of the empty response
response.content = fallback_content
print(f"Created fallback response: {fallback_content[:100]}...")
else:
content_preview = response.content[:300].replace('\n', ' ')
print(f"Response preview: {content_preview}...")
except Exception as e:
print(f"Error in LLM invocation: {str(e)}")
# Create a fallback response in case of LLM errors
response = AIMessage(content=(
"Thought: I encountered an error in processing. Let me try to proceed with what I know.\n\n"
"Final Answer: Unable to complete the task due to an error in processing."
))
# Extract the action JSON from the response text
action_json = extract_json_from_text(response.content)
print(f"Extracted action: {action_json.get('action') if action_json else 'None'}")
assistant_response_message = AIMessage(content=response.content)
state_update: Dict[str, Any] = {
"messages": [assistant_response_message],
"iteration_count": iteration_count
}
if action_json and "action" in action_json and "action_input" in action_json:
tool_name = action_json["action"]
tool_input = action_json["action_input"]
# Handle nested JSON issue - check if any value in action_input is a JSON string
if isinstance(tool_input, dict):
for key, value in tool_input.items():
if isinstance(value, str) and value.strip().startswith("{"):
try:
nested_json = json.loads(value)
if isinstance(nested_json, dict) and "action" in nested_json and "action_input" in nested_json:
# This is a nested structure, use the inner one
tool_name = nested_json["action"]
tool_input = nested_json["action_input"]
print(f"Unwrapped nested JSON. New tool: {tool_name}")
break
except json.JSONDecodeError:
continue
print(f"Using tool: {tool_name}")
tool_call_id = f"call_{random.randint(1000000, 9999999)}"
state_update["current_tool"] = tool_name
state_update["action_input"] = tool_input
else:
print("No tool action found or 'Final Answer' detected in response.")
state_update["current_tool"] = None
state_update["action_input"] = None
print("=== End Assistant Node ===\n")
return state_update
def extract_json_from_text(text: str) -> dict:
"""Extract JSON from text, handling markdown code blocks and other formats."""
try:
import re
import json
# Return empty if text is None or very short (less than 10 chars)
if not text or len(text.strip()) < 10:
print("Warning: Empty or very short text input to JSON extraction")
return None
print(f"Attempting to extract JSON from text: {text[:200]}...")
# First, clean up the text to handle specific patterns that might confuse parsing
text = text.replace('\\n', '\n').replace('\\"', '"')
# Case 1: "Final Answer:" detection - if present, return None to indicate we should end
if "Final Answer:" in text:
print("Detected 'Final Answer' - no tool action needed")
return None
# Case 2: Extract direct python dictionary representation without JSON formatting
if "action_input" in text and not '{"action"' in text and not '{"action_input"' in text:
# Try regex to extract a Python dict-like structure
action_match = re.search(r"action:\s*(\w+)", text, re.IGNORECASE)
input_match = re.search(r"action_input:\s*(\{.+?\})", text, re.DOTALL | re.IGNORECASE)
if action_match and input_match:
action = action_match.group(1).strip()
try:
action_input = eval(input_match.group(1)) # Be careful with eval!
if isinstance(action_input, dict):
return {"action": action, "action_input": action_input}
except:
pass
# Pattern 1: Look for "Action:" followed by a markdown code block
action_match = re.search(r"Action:\s*```(?:python|json)?\s*(.*?)```", text, re.DOTALL)
if action_match:
action_content = action_match.group(1).strip()
print(f"Found action content from markdown block: {action_content[:100]}...")
# Try to parse as JSON first
try:
parsed_json = json.loads(action_content)
if "action" in parsed_json and "action_input" in parsed_json:
return parsed_json
except json.JSONDecodeError:
# If it's Python code, create action structure
if "=" in action_content or "import" in action_content or "print" in action_content:
print("Detected Python code, formatting as action_input")
return {
"action": "python_code",
"action_input": {"code": action_content}
}
# Pattern 2: Look for regular markdown code blocks
code_blocks = re.findall(r"```(?:json|python)?(.+?)```", text, re.DOTALL)
for block in code_blocks:
block = block.strip()
print(f"Processing code block: {block[:100]}...")
# Try to parse as JSON
try:
parsed = json.loads(block)
if "action" in parsed and "action_input" in parsed:
print(f"Successfully parsed JSON block: {parsed}")
return parsed
except json.JSONDecodeError:
# If it's Python code, create action structure
if "=" in block or "import" in block or "print" in block or "def " in block:
print("Detected Python code in block, formatting as action_input")
return {
"action": "python_code",
"action_input": {"code": block}
}
# Pattern 3: Direct JSON object ({...}) in the text
json_matches = re.findall(r"\{[\s\S]*?\}", text)
for json_str in json_matches:
try:
parsed = json.loads(json_str)
if "action" in parsed and "action_input" in parsed:
print(f"Found valid JSON object: {parsed}")
return parsed
except json.JSONDecodeError:
continue
# Pattern 4: Look for patterns like 'action': 'tool_name', 'action_input': {...}
action_pattern = re.search(r"['\"](action)['\"]:\s*['\"](\w+)['\"]", text)
action_input_pattern = re.search(r"['\"](action_input)['\"]:\s*(\{.+\})", text, re.DOTALL)
if action_pattern and action_input_pattern:
action = action_pattern.group(2)
action_input_str = action_input_pattern.group(2)
try:
action_input = json.loads(action_input_str)
return {
"action": action,
"action_input": action_input
}
except json.JSONDecodeError:
pass
# Pattern 5: Look for simple text patterns like "I need to use tool X to search for Y"
tool_patterns = [
(r"(?:use|using|need to use|should use|will use)(?:\s+the)?\s+(\w+)(?:\s+tool)?\s+to\s+(?:search|find|look up|research)(?:\s+for)?\s+['\"](.*?)['\"]",
lambda m: {"action": m.group(1).lower(), "action_input": {"query": m.group(2)}}),
(r"(?:use|using|need to use|should use|will use)(?:\s+the)?\s+(\w+)(?:\s+tool)?\s+to\s+(?:search|find|look up|research)\s+(?:for\s+)?(.+?)(?=\.|$)",
lambda m: {"action": m.group(1).lower(), "action_input": {"query": m.group(2).strip()}}),
(r"(?:use|using|need to use|should use|will use)(?:\s+the)?\s+(\w+)(?:\s+tool)?\s+on\s+['\"](.*?)['\"]",
lambda m: {"action": m.group(1).lower(), "action_input": {"query": m.group(2)}})
]
for pattern, formatter in tool_patterns:
match = re.search(pattern, text, re.IGNORECASE)
if match:
try:
result = formatter(match)
# Map common words to actual tool names
tool_mapping = {
"tavily": "tavily_search",
"wikipedia": "wikipedia_search",
"arxiv": "arxiv_search",
"web": "tavily_search",
"python": "python_code",
"excel": "excel_to_text",
"youtube": "process_youtube_video",
"webpage": "webpage_scrape",
"scrape": "webpage_scrape",
"pdf": "process_online_document",
"document": "process_online_document",
"online": "process_online_document"
}
if result["action"].lower() in tool_mapping:
result["action"] = tool_mapping[result["action"].lower()]
print(f"Extracted tool action using pattern: {result}")
return result
except Exception as e:
print(f"Error formatting pattern match: {e}")
# Fallback: If we detect thinking about a specific topic, suggest a related tool
fallback_patterns = [
(r"(?:I need|I should|I will|let me|I can)(?:\s+(?:to))?\s+(?:search|find|look for)\s+(?:information|details|data)?\s+(?:about|on|regarding)?\s+(.+?)(?=\.|$)",
lambda m: {"action": "tavily_search", "action_input": {"query": m.group(1).strip()}}),
(r"(?:I will|I should|let me)(?:\s+now)?\s+(?:search|check)(?:\s+on)?\s+wikipedia\s+(?:for)?\s+(.+?)(?=\.|$)",
lambda m: {"action": "wikipedia_search", "action_input": {"query": m.group(1).strip()}}),
(r"(?:I need|I should|I will|let me)(?:\s+to)?\s+(?:execute|run|write|use)\s+(?:some|a)?\s+python\s+(?:code)?",
lambda m: {"action": "python_code", "action_input": {"code": "# Example code\nprint('Please specify Python code to execute')"}})
]
for pattern, formatter in fallback_patterns:
match = re.search(pattern, text, re.IGNORECASE)
if match:
try:
result = formatter(match)
print(f"Used fallback pattern to suggest tool: {result}")
return result
except Exception as e:
print(f"Error in fallback pattern: {e}")
print("Could not extract valid JSON from text using any pattern")
# Last resort fallback: If we've tried everything and failed, and the response is long
# enough to suggest the model is doing work but not formatting correctly, try to
# extract any query-like content and suggest a tavily search
if len(text) > 200 and "search" in text.lower():
# Extract a potential query - look for sentences with search-related terms
search_sentences = re.findall(r'[^.!?]*(?:search|find|look for|investigate|research)[^.!?]*[.!?]', text)
if search_sentences:
best_sentence = max(search_sentences, key=len)
# Clean up the query
query = re.sub(r'(?:I will|I should|I need to|I want to|Let me|I\'ll|I can)\s+(?:search|find|look for|investigate|research)(?:\s+for)?\s+', '', best_sentence)
query = query.strip('.!? \t\n')
if query and len(query) > 5:
print(f"Last resort fallback: suggesting Tavily search for: {query}")
return {"action": "tavily_search", "action_input": {"query": query}}
return None
except Exception as e:
print(f"Error extracting JSON: {e}")
return None
# Comment out the web_search_node function
# def web_search_node(state: AgentState) -> Dict[str, Any]:
# """Node that executes the web search tool."""
# print("Web Search Tool Called...\n\n")
#
# # Extract tool arguments
# action_input = state.get("action_input", {})
# print(f"Web search action_input: {action_input}")
#
# # Try different ways to extract the query
# query = ""
# if isinstance(action_input, dict):
# query = action_input.get("query", "")
# elif isinstance(action_input, str):
# query = action_input
#
# print(f"Searching for: '{query}'")
#
# # Call the search function with retry logic
# result = safe_web_search(query)
# print(f"Search result: {result}") # Print the full result for debugging
#
# # Check if we hit rate limits and add a helpful note
# if "rate limit" in result.lower() or "ratelimit" in result.lower():
# result += "\n\nNote: You can use your internal knowledge to provide a response since the search is rate limited."
#
# # Format the observation to continue the ReAct cycle
# # Don't include "Observation:" as the assistant is stopped at this token
# observation = result
#
# # Create a tool message with the result
# tool_message = AIMessage(
# content=f"Observation: {observation}"
# )
#
# # Print the observation that will be sent back to the assistant
# print("\n=== TOOL OBSERVATION ===")
# print(tool_message.content[:500] + "..." if len(tool_message.content) > 500 else tool_message.content)
# print("=== END OBSERVATION ===\n")
#
# # Return the updated state
# return {
# "messages": state["messages"] + [tool_message],
# "current_tool": None, # Reset the current tool
# "action_input": None # Clear the action input
# }
def python_code_node(state: AgentState) -> Dict[str, Any]:
"""Node that executes Python code."""
print("\n=== Python Code Node ===")
try:
# Extract tool arguments
action_input = state.get("action_input", {})
print(f"Input: {action_input.get('code', '')[:100]}...")
# Get the code string
code = ""
if isinstance(action_input, dict):
code = action_input.get("code", "")
elif isinstance(action_input, str):
code = action_input
print(f"Original code field (first 100 chars): {code[:100]}")
def extract_code_from_json(json_str):
"""Recursively extract code from nested JSON structures."""
try:
parsed = json.loads(json_str)
if isinstance(parsed, dict):
# Check for direct code field
if "code" in parsed:
return parsed["code"]
# Check for nested action_input structure
if "action_input" in parsed:
inner_input = parsed["action_input"]
if isinstance(inner_input, dict):
if "code" in inner_input:
return inner_input["code"]
# If inner_input is also JSON string, recurse
if isinstance(inner_input.get("code", ""), str) and inner_input["code"].strip().startswith("{"):
return extract_code_from_json(inner_input["code"])
return json_str
except:
return json_str
# Handle nested JSON structures
if isinstance(code, str) and code.strip().startswith("{"):
code = extract_code_from_json(code)
print("Extracted code from JSON structure")
print(f"Final code to execute: {code[:100]}...")
# Execute the code
try:
result = run_python_code(code)
print(f"Execution successful")
# Format the observation
tool_message = AIMessage(
content=f"Observation: {result.strip()}"
)
# Print the observation that will be sent back to the assistant
print("=== End Python Code Node ===\n")
# Return the updated state
return {
"messages": state["messages"] + [tool_message],
"current_tool": None, # Reset the current tool
"action_input": None # Clear the action input
}
except Exception as e:
error_message = f"Error executing Python code: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End Python Code Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
except Exception as e:
error_message = f"Error in Python code node: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End Python Code Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
def webpage_scrape_node(state: AgentState) -> Dict[str, Any]:
"""Node that scrapes content from a specific webpage URL."""
print("\n=== Webpage Scrape Node ===")
try:
# Extract tool arguments
action_input = state.get("action_input", {})
url = action_input.get("url", "") if isinstance(action_input, dict) else action_input
print(f"URL: {url}")
# Safety check - don't run with empty URL
if not url:
result = "Error: No URL provided. Please provide a valid URL to scrape."
else:
# Call the webpage scraping function
result = scrape_webpage(url)
# Format the observation
tool_message = AIMessage(
content=f"Observation: {result.strip()}"
)
print("=== End Webpage Scrape Node ===\n")
# Return the updated state
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
except Exception as e:
error_message = f"Error in webpage scrape node: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End Webpage Scrape Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
def wikipedia_search_node(state: AgentState) -> Dict[str, Any]:
"""Node that processes Wikipedia search requests."""
print("\n=== Wikipedia Search Node ===")
try:
# Extract tool arguments
action_input = state.get("action_input", {})
query = action_input.get("query", "") if isinstance(action_input, dict) else action_input
num_results = action_input.get("num_results", 3) if isinstance(action_input, dict) else 3
print(f"Query: {query} (max results: {num_results})")
# Safety check - don't run with empty query
if not query:
result = "Error: No search query provided. Please provide a valid query for Wikipedia search."
else:
# Call the Wikipedia search function
result = wikipedia_search(query, num_results)
# Format the observation
tool_message = AIMessage(
content=f"Observation: {result.strip()}"
)
print("=== End Wikipedia Search Node ===\n")
# Return the updated state
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
except Exception as e:
error_message = f"Error in Wikipedia search node: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End Wikipedia Search Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
def tavily_search_node(state: AgentState) -> Dict[str, Any]:
"""Node that processes Tavily search requests."""
print("\n=== Tavily Search Node ===")
try:
# Extract tool arguments
action_input = state.get("action_input", {})
query = action_input.get("query", "") if isinstance(action_input, dict) else action_input
search_depth = action_input.get("search_depth", "basic") if isinstance(action_input, dict) else "basic"
print(f"Query: {query} (depth: {search_depth})")
# Safety check - don't run with empty query
if not query:
result = "Error: No search query provided. Please provide a valid query for Tavily search."
else:
# Call the Tavily search function
result = tavily_search(query, search_depth)
# Format the observation
tool_message = AIMessage(
content=f"Observation: {result.strip()}"
)
print("=== End Tavily Search Node ===\n")
# Return the updated state
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
except Exception as e:
error_message = f"Error in Tavily search node: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End Tavily Search Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
def arxiv_search_node(state: AgentState) -> Dict[str, Any]:
"""Node that processes ArXiv search requests."""
print("\n=== ArXiv Search Node ===")
try:
# Extract tool arguments
action_input = state.get("action_input", {})
query = action_input.get("query", "") if isinstance(action_input, dict) else action_input
max_results = action_input.get("max_results", 5) if isinstance(action_input, dict) else 5
print(f"Query: {query} (max results: {max_results})")
# Safety check - don't run with empty query
if not query:
result = "Error: No search query provided. Please provide a valid query for ArXiv search."
else:
# Call the ArXiv search function
result = arxiv_search(query, max_results)
# Format the observation
tool_message = AIMessage(
content=f"Observation: {result.strip()}"
)
print("=== End ArXiv Search Node ===\n")
# Return the updated state
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
except Exception as e:
error_message = f"Error in ArXiv search node: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End ArXiv Search Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
def supabase_operation_node(state: AgentState) -> Dict[str, Any]:
"""Node that processes Supabase database operations."""
print("\n=== Supabase Operation Node ===")
try:
# Extract tool arguments
action_input = state.get("action_input", {})
operation_type = action_input.get("operation_type", "") if isinstance(action_input, dict) else ""
table = action_input.get("table", "") if isinstance(action_input, dict) else ""
print(f"Operation: {operation_type} on table {table}")
# Safety check
if not operation_type or not table:
result = "Error: Both operation_type and table are required. operation_type should be one of: insert, select, update, delete"
else:
# Call the Supabase operation function
result = supabase_operation(operation_type, table, action_input.get("data"), action_input.get("filters"))
# Format the observation
tool_message = AIMessage(
content=f"Observation: {result.strip()}"
)
print("=== End Supabase Operation Node ===\n")
# Return the updated state
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
except Exception as e:
error_message = f"Error in Supabase operation node: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End Supabase Operation Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
def excel_to_text_node(state: AgentState) -> Dict[str, Any]:
"""Node that processes Excel to Markdown table conversions."""
print("\n=== Excel to Text Node ===")
try:
# Extract tool arguments
action_input = state.get("action_input", {})
excel_path = action_input.get("excel_path", "") if isinstance(action_input, dict) else ""
sheet_name = action_input.get("sheet_name") if isinstance(action_input, dict) else None
print(f"File: {excel_path} (sheet: {sheet_name or 'default'})")
# Safety check
if not excel_path:
result = "Error: Excel file path is required"
else:
# Call the Excel to text function
result = excel_to_text(excel_path, sheet_name, action_input.get("file_content"))
# Format the observation
tool_message = AIMessage(
content=f"Observation: {result.strip()}"
)
print("=== End Excel to Text Node ===\n")
# Return the updated state
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
except Exception as e:
error_message = f"Error in Excel to text node: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End Excel to Text Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
def process_youtube_video_node(state: AgentState) -> Dict[str, Any]:
"""Node that processes YouTube videos."""
print("\n=== YouTube Video Processing Node ===")
try:
# Extract tool arguments
action_input = state.get("action_input", {})
url = action_input.get("url", "") if isinstance(action_input, dict) else action_input
summarize = action_input.get("summarize", True) if isinstance(action_input, dict) else True
print(f"URL: {url} (summarize: {summarize})")
# Safety check - don't run with empty URL
if not url:
result = "Error: No URL provided. Please provide a valid YouTube URL."
elif not url.startswith(("http://", "https://")) or not ("youtube.com" in url or "youtu.be" in url):
result = f"Error: Invalid YouTube URL format: {url}. Please provide a valid URL starting with http:// or https:// and containing youtube.com or youtu.be."
else:
# Call the YouTube processing function
try:
result = process_youtube_video(url, summarize)
except Exception as e:
result = f"Error processing YouTube video: {str(e)}\n\nThis could be due to:\n- The video is private or has been removed\n- Network connectivity issues\n- YouTube API changes\n- Rate limiting"
# Format the observation
tool_message = AIMessage(
content=f"Observation: {result.strip()}"
)
print("=== End YouTube Video Processing Node ===\n")
# Return the updated state
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
except Exception as e:
error_message = f"Error in YouTube video processing node: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End YouTube Video Processing Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
def transcribe_audio_node(state: AgentState) -> Dict[str, Any]:
"""Node that processes audio transcription requests."""
print("\n=== Audio Transcription Node ===")
try:
# Extract tool arguments
action_input = state.get("action_input", {})
audio_path = action_input.get("audio_path", "") if isinstance(action_input, dict) else ""
language = action_input.get("language") if isinstance(action_input, dict) else None
print(f"File: {audio_path} (language: {language or 'auto-detect'})")
# Safety check
if not audio_path:
result = "Error: Audio file path is required"
else:
# Call the transcribe audio function
result = transcribe_audio(audio_path, action_input.get("file_content"), language)
# Format the observation
tool_message = AIMessage(
content=f"Observation: {result.strip()}"
)
print("=== End Audio Transcription Node ===\n")
# Return the updated state
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
except Exception as e:
error_message = f"Error in audio transcription node: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End Audio Transcription Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
def process_image_node(state: AgentState) -> Dict[str, Any]:
"""Node that processes image analysis requests."""
print("\n=== Image Processing Node ===")
try:
# Extract tool arguments
action_input = state.get("action_input", {})
image_path = action_input.get("image_path", "") if isinstance(action_input, dict) else ""
image_url = action_input.get("image_url") if isinstance(action_input, dict) else None
analyze_content = action_input.get("analyze_content", True) if isinstance(action_input, dict) else True
print(f"Source: {image_url or image_path} (analyze: {analyze_content})")
# Safety check
if not image_path and not image_url:
result = "Error: Either image path or image URL is required"
else:
# Call the process image function
result = process_image(image_path, image_url, action_input.get("file_content"), analyze_content)
# Format the observation
tool_message = AIMessage(
content=f"Observation: {result.strip()}"
)
print("=== End Image Processing Node ===\n")
# Return the updated state
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
except Exception as e:
error_message = f"Error in image processing node: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End Image Processing Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
def read_file_node(state: AgentState) -> Dict[str, Any]:
"""Node that reads text file contents."""
print("\n=== File Reading Node ===")
try:
# Extract tool arguments
action_input = state.get("action_input", {})
file_path = action_input.get("file_path", "") if isinstance(action_input, dict) else ""
line_start = action_input.get("line_start") if isinstance(action_input, dict) else None
line_end = action_input.get("line_end") if isinstance(action_input, dict) else None
print(f"File: {file_path} (lines: {line_start}-{line_end if line_end else 'end'})")
# Safety check
if not file_path:
result = "Error: File path is required"
else:
# Call the read file function
result = read_file(file_path, action_input.get("file_content"), line_start, line_end)
# Format the observation
tool_message = AIMessage(
content=f"Observation: {result.strip()}"
)
print("=== End File Reading Node ===\n")
# Return the updated state
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
except Exception as e:
error_message = f"Error in file reading node: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End File Reading Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
def process_online_document_node(state: AgentState) -> Dict[str, Any]:
"""Node that processes online PDFs and images."""
print("\n=== Online Document Processing Node ===")
try:
# Extract tool arguments
action_input = state.get("action_input", {})
url = action_input.get("url", "") if isinstance(action_input, dict) else action_input
doc_type = action_input.get("doc_type", "auto") if isinstance(action_input, dict) else "auto"
print(f"URL: {url} (type: {doc_type})")
# Safety check - don't run with empty URL
if not url:
result = "Error: No URL provided. Please provide a valid URL to process."
elif not url.startswith(("http://", "https://")):
result = f"Error: Invalid URL format: {url}. Please provide a valid URL starting with http:// or https://."
else:
# Call the online document processing function
try:
result = process_online_document(url, doc_type)
except Exception as e:
result = f"Error processing online document: {str(e)}\n\nThis could be due to:\n- The document is not accessible\n- Network connectivity issues\n- Unsupported document type\n- Rate limiting"
# Format the observation
tool_message = AIMessage(
content=f"Observation: {result.strip()}"
)
print("=== End Online Document Processing Node ===\n")
# Return the updated state
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
except Exception as e:
error_message = f"Error in online document processing node: {str(e)}"
print(error_message)
tool_message = AIMessage(content=f"Observation: {error_message}")
print("=== End Online Document Processing Node ===\n")
return {
"messages": state["messages"] + [tool_message],
"current_tool": None,
"action_input": None
}
# Router function to direct to the correct tool
def router(state: AgentState) -> str:
"""Route to the appropriate tool based on the current_tool field."""
tool = state.get("current_tool")
action_input = state.get("action_input")
print(f"Routing to: {tool}")
print(f"Router received action_input: {action_input}")
if tool == "python_code":
return "python_code"
elif tool == "webpage_scrape":
return "webpage_scrape"
elif tool == "wikipedia_search":
return "wikipedia_search"
elif tool == "tavily_search":
return "tavily_search"
elif tool == "arxiv_search":
return "arxiv_search"
elif tool == "supabase_operation":
return "supabase_operation"
elif tool == "excel_to_text":
return "excel_to_text"
elif tool == "process_youtube_video":
return "process_youtube_video"
elif tool == "transcribe_audio":
return "transcribe_audio"
elif tool == "process_image":
return "process_image"
elif tool == "read_file":
return "read_file"
elif tool == "process_online_document":
return "process_online_document"
else:
return "end"
# Create the graph
def create_agent_graph() -> StateGraph:
"""Create the complete agent graph with individual tool nodes."""
builder = StateGraph(AgentState)
# Define nodes: these do the work
builder.add_node("assistant", assistant)
builder.add_node("python_code", python_code_node)
builder.add_node("webpage_scrape", webpage_scrape_node)
builder.add_node("wikipedia_search", wikipedia_search_node)
builder.add_node("tavily_search", tavily_search_node)
builder.add_node("arxiv_search", arxiv_search_node)
builder.add_node("supabase_operation", supabase_operation_node)
builder.add_node("excel_to_text", excel_to_text_node)
builder.add_node("process_youtube_video", process_youtube_video_node)
builder.add_node("transcribe_audio", transcribe_audio_node)
builder.add_node("process_image", process_image_node)
builder.add_node("read_file", read_file_node)
builder.add_node("process_online_document", process_online_document_node)
# Define edges: these determine how the control flow moves
builder.add_edge(START, "assistant")
# Debug the state passing
def debug_state(state):
print("\n=== DEBUG STATE ===")
print(f"State keys: {state.keys()}")
print(f"Current tool: {state.get('current_tool')}")
print(f"Action input: {state.get('action_input')}")
print("=== END DEBUG ===\n")
return state
# Add a checkpoint between nodes to track state
builder.add_node("debug", debug_state)
# Conditional edge from assistant to debug
builder.add_edge("assistant", "debug")
# Conditional edge from debug to tools or end
builder.add_conditional_edges(
"debug",
router,
{
"python_code": "python_code",
"webpage_scrape": "webpage_scrape",
"wikipedia_search": "wikipedia_search",
"tavily_search": "tavily_search",
"arxiv_search": "arxiv_search",
"supabase_operation": "supabase_operation",
"excel_to_text": "excel_to_text",
"process_youtube_video": "process_youtube_video",
"transcribe_audio": "transcribe_audio",
"process_image": "process_image",
"read_file": "read_file",
"process_online_document": "process_online_document",
"end": END
}
)
# Tools always go back to assistant
builder.add_edge("python_code", "assistant")
builder.add_edge("webpage_scrape", "assistant")
builder.add_edge("wikipedia_search", "assistant")
builder.add_edge("tavily_search", "assistant")
builder.add_edge("arxiv_search", "assistant")
builder.add_edge("supabase_operation", "assistant")
builder.add_edge("excel_to_text", "assistant")
builder.add_edge("process_youtube_video", "assistant")
builder.add_edge("transcribe_audio", "assistant")
builder.add_edge("process_image", "assistant")
builder.add_edge("read_file", "assistant")
builder.add_edge("process_online_document", "assistant")
# Compile the graph
return builder.compile()
# Main agent class that integrates with your existing app.py
class ScholarAI:
def __init__(self, max_iterations=35, temperature=0.1, max_tokens=2000, model="gpt-4o-mini", apify_api_token=None):
# Check for OpenAI API key
if not os.getenv("OPENAI_API_KEY"):
raise ValueError("OpenAI API key not found. Please set the OPENAI_API_KEY environment variable.")
try:
# Test the API key with a simple request
test_llm = ChatOpenAI(
model=model,
temperature=temperature,
max_tokens=max_tokens
)
test_llm.invoke("test") # This will fail if API key is invalid
except Exception as e:
error_msg = str(e).lower()
if "invalid_api_key" in error_msg or "incorrect_api_key" in error_msg:
raise ValueError("Invalid OpenAI API key. Please check your API key and try again.")
elif "rate_limit" in error_msg or "quota" in error_msg:
raise ValueError("OpenAI API rate limit exceeded or quota reached. Please try again later.")
else:
raise ValueError(f"Error initializing OpenAI client: {str(e)}")
self.graph = create_agent_graph()
self.tools = tools_config
self.max_iterations = max_iterations # Maximum iterations for the graph
# Update the global llm instance with the specified parameters
global llm
llm = ChatOpenAI(
model=model,
temperature=temperature,
max_tokens=max_tokens
)
# Set Apify API token if provided
if apify_api_token:
os.environ["APIFY_API_TOKEN"] = apify_api_token
print("Apify API token set successfully")
print(f"ScholarAI initialized with model={model}, temperature={temperature}, max_tokens={max_tokens}")
def __call__(self, question: str, attachments: dict = None) -> str:
"""
Process a question and return an answer.
Args:
question: The user's question text
attachments: Optional dictionary of attachments with keys as names and values as base64-encoded content
"""
# Process attachments if provided
attachment_info = ""
if attachments and isinstance(attachments, dict) and len(attachments) > 0:
attachment_names = list(attachments.keys())
attachment_info = f"\n\nI've attached the following files: {', '.join(attachment_names)}. "
# Add different instructions based on detected file types
excel_files = [name for name in attachment_names if name.endswith(('.xlsx', '.xls'))]
if excel_files:
attachment_info += f"Use the excel_to_text tool with the file_content parameter to process the Excel files."
# Initialize the state with the question and attachment info
question_with_attachments = question + attachment_info if attachment_info else question
initial_state = {
"messages": [HumanMessage(content=f"Question: {question_with_attachments}")],
"current_tool": None,
"action_input": None,
"iteration_count": 0, # Initialize iteration_count
"attachments": attachments or {} # Store attachments in the state
}
# Run the graph
print(f"Starting graph execution with question: {question}")
if attachments:
print(f"Included attachments: {list(attachments.keys())}")
try:
# Set a reasonable recursion limit based on max_iterations
result = self.graph.invoke(initial_state, {"recursion_limit": self.max_iterations})
# Print the final state for debugging
print(f"Final state keys: {result.keys()}")
print(f"Final message count: {len(result['messages'])}")
# Extract the final message and return just the final answer
final_message = result["messages"][-1].content
print("Final message: ", final_message)
# Extract just the final answer part
if "Final Answer:" in final_message:
final_answer = final_message.split("Final Answer:", 1)[1].strip()
return final_answer
return final_message
except Exception as e:
print(f"Error processing question: {str(e)}")
# Otherwise return the error
return f"I encountered an error while processing your question: {str(e)}"
# Example usage:
if __name__ == "__main__":
agent = ScholarAI(max_iterations=25)
response = agent("""The attached Excel file contains the sales of menu items for a local fast-food chain. What were the total sales that the chain made from food (not including drinks)? Express your answer in USD with two decimal places. TEMPP\excel.xlsx """)
print("\nFinal Response:")
print(response)
|