File size: 247,741 Bytes
e1f1c52
 
 
 
 
 
 
996fbf4
 
 
 
 
 
 
 
 
e1f1c52
 
 
 
 
 
 
996fbf4
 
 
 
 
 
 
 
 
 
e1f1c52
 
 
996fbf4
 
 
e1f1c52
 
996fbf4
e1f1c52
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
 
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
 
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
 
 
 
996fbf4
 
e1f1c52
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
 
 
 
996fbf4
 
e1f1c52
 
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
996fbf4
e1f1c52
 
996fbf4
e1f1c52
 
 
 
 
996fbf4
 
 
 
 
 
 
 
 
e1f1c52
 
 
 
996fbf4
 
e1f1c52
 
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
996fbf4
e1f1c52
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
 
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
 
 
 
996fbf4
 
e1f1c52
 
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
996fbf4
e1f1c52
 
996fbf4
e1f1c52
996fbf4
e1f1c52
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
996fbf4
 
 
e1f1c52
996fbf4
 
e1f1c52
996fbf4
 
 
 
 
e1f1c52
996fbf4
 
 
 
e1f1c52
996fbf4
e1f1c52
 
 
 
996fbf4
 
e1f1c52
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
 
 
996fbf4
e1f1c52
 
 
 
996fbf4
 
e1f1c52
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
 
 
 
996fbf4
 
e1f1c52
 
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
996fbf4
e1f1c52
 
996fbf4
e1f1c52
 
 
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
996fbf4
 
e1f1c52
 
 
 
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
 
 
 
 
996fbf4
e1f1c52
 
996fbf4
e1f1c52
 
 
 
 
996fbf4
 
 
e1f1c52
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
996fbf4
 
 
e1f1c52
996fbf4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f1c52
 
996fbf4
 
e1f1c52
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "9dd69c4b",
   "metadata": {},
   "source": [
    "# Gaia agent"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d2d7cdff",
   "metadata": {},
   "source": [
    "## Import libraries and define global variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "700c86ca",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/cora/Documents/agents/gaia-agent/.venv/lib/python3.12/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    }
   ],
   "source": [
    "import requests\n",
    "from dotenv import load_dotenv\n",
    "from IPython.display import Image, display, Markdown\n",
    "from typing import Literal\n",
    "from langchain.tools import tool\n",
    "from langgraph.graph import StateGraph, START, END\n",
    "from langgraph.graph import MessagesState\n",
    "from langgraph.prebuilt import ToolNode\n",
    "from langchain_openai import ChatOpenAI\n",
    "from langchain_core.rate_limiters import InMemoryRateLimiter\n",
    "from langchain_core.messages import SystemMessage, HumanMessage, AIMessage, ToolMessage\n",
    "from langgraph.types import Command\n",
    "from langchain_tavily import TavilySearch\n",
    "import arxiv\n",
    "from markdownify import markdownify as md\n",
    "import wikipedia\n",
    "from huggingface_hub import hf_hub_download, InferenceClient\n",
    "import pandas as pd\n",
    "import re\n",
    "from pathlib import Path\n",
    "from bs4 import BeautifulSoup\n",
    "import os\n",
    "import math\n",
    "from pydantic import BaseModel, Field\n",
    "from youtube_transcript_api import YouTubeTranscriptApi\n",
    "import subprocess\n",
    "import base64\n",
    "from PyPDF2 import PdfReader\n",
    "from tempfile import NamedTemporaryFile\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "ef8bb3b8",
   "metadata": {},
   "outputs": [],
   "source": [
    "load_dotenv()\n",
    "\n",
    "wikipedia.set_user_agent(\"GaiaAgent (alexinhocora@gmail.com)\")\n",
    "\n",
    "tavily_search = TavilySearch(max_results=5, topic=\"general\")\n",
    "\n",
    "arxiv_client = arxiv.Client()\n",
    "\n",
    "ytt_api = YouTubeTranscriptApi()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "61c93bfb",
   "metadata": {},
   "outputs": [],
   "source": [
    "API_URL = \"https://agents-course-unit4-scoring.hf.space\"\n",
    "\n",
    "questions_url = f\"{API_URL}/questions\"\n",
    "file_url = f\"{API_URL}/files\" + \"/{task_id}\"\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "06b6c832",
   "metadata": {},
   "source": [
    "## Define Tools"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "209f3acd",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def calculator(expression: str) -> str:\n",
    "    \"\"\"Evaluate a mathematical expression.\n",
    "\n",
    "    Args:\n",
    "        expression: A string containing the mathematical expression to evaluate.\n",
    "        examples:\n",
    "            \"2 + 2 * (3 - 1)\"\n",
    "            \"sqrt(16) + 5\"\n",
    "            \"sin(pi / 2) * 10\"\n",
    "    \"\"\"\n",
    "    try:\n",
    "        # Source - https://stackoverflow.com/q/3513292\n",
    "        # Posted by flybywire, modified by community. See post 'Timeline' for change history\n",
    "        # Retrieved 2026-07-12, License - CC BY-SA 3.0\n",
    "        safe_names = {\n",
    "            name: getattr(math, name) for name in dir(math) if not name.startswith(\"_\")\n",
    "        }\n",
    "\n",
    "        safe_names.update(\n",
    "            {\n",
    "                \"abs\": abs,\n",
    "                \"round\": round,\n",
    "                \"min\": min,\n",
    "                \"max\": max,\n",
    "                \"int\": int,\n",
    "                \"float\": float,\n",
    "                \"complex\": complex,\n",
    "            }\n",
    "        )\n",
    "\n",
    "        result = eval(\n",
    "            expression,\n",
    "            {\"__builtins__\": {}},\n",
    "            safe_names,\n",
    "        )\n",
    "\n",
    "        return str(result)\n",
    "\n",
    "    except Exception as e:\n",
    "        return f\"Error: {str(e)}\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "b989bcba",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def wikipedia_search(query: str) -> str:\n",
    "    \"\"\"Retrieve a summary of the top 3 Wikipedia pages for a given search query.\n",
    "\n",
    "    Args:\n",
    "        query: Search key term to look up in wikipedia.\n",
    "    \"\"\"\n",
    "    documents = []\n",
    "    suggestions = wikipedia.search(query, results=3)\n",
    "    for page in suggestions:\n",
    "        try:\n",
    "            page = wikipedia.page(page, auto_suggest=False)\n",
    "            sections = page.sections\n",
    "            if not sections:\n",
    "                # Sometime fails to get the proper sections, backup plan is to parse the html and get the sections from the table of contents\n",
    "                soup = BeautifulSoup(page.html(), \"html.parser\")\n",
    "                sections = [\n",
    "                    a.attrs.get(\"href\", \"\").strip(\"#\")\n",
    "                    for a in soup.select(\"#toc li > a\")\n",
    "                ]\n",
    "            documents.append(f\"\"\"# {page.title}\n",
    "**URL:** {page.url}\\\\\n",
    "**Sections:** {\",\".join(sections)}\n",
    "\n",
    "{page.summary.strip()}\"\"\")\n",
    "\n",
    "        except Exception as e:\n",
    "            print(f\"Error retrieving page for {page}: {e}\")\n",
    "            continue\n",
    "\n",
    "    formatted_docs = \"\\n\\n---\\n\\n\".join(documents)\n",
    "\n",
    "    return formatted_docs\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "4a7e2f5b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "# Mercedes Sosa\n",
       "**URL:** https://en.wikipedia.org/wiki/Mercedes_Sosa\\\n",
       "**Sections:** Life,Awards,Death,Tributes,Discography,Studio_albums,EPs,Live_albums,Compilation_albums,Filmography,References,Further_reading,External_links\n",
       "\n",
       "Haydée Marta Mercedes Sosa Girón (9 July 1935 – 4 October 2009) was an Argentine singer who was popular throughout Latin America and many countries outside the region. With her roots in Argentine folk music, Sosa became one of the preeminent exponents of El nuevo cancionero. She gave voice to songs written by many Latin American songwriters. She was hailed as the \"voice of the voiceless ones\", and often called \"the conscience of Latin America\".\n",
       "Sosa performed in venues such as the Lincoln Center in New York City, the Théâtre Mogador in Paris, the Sistine Chapel in Vatican City, as well as sold-out shows in New York's Carnegie Hall and the Roman Colosseum during her final decade of life. Her career spanned four decades and she was the recipient of six Latin Grammy awards (2000, 2003, 2004, 2006, 2009, 2011), including a Latin Grammy Lifetime Achievement Award in 2004 and two posthumous Latin Grammy Award for Best Folk Album in 2009 and 2011. She won the Premio Gardel in 2000, the main musical award in Argentina. She served as an ambassador for UNICEF.\n",
       "\n",
       "---\n",
       "\n",
       "# Cantora, un Viaje Íntimo\n",
       "**URL:** https://en.wikipedia.org/wiki/Cantora,_un_Viaje_%C3%8Dntimo\\\n",
       "**Sections:** Background,Content,Critical_reception,Year-end_rankings,Track_listing,Charts,References\n",
       "\n",
       "Cantora, un Viaje Íntimo  (English: Cantora, An Intimate Journey) is a double album by Argentine singer Mercedes Sosa, released on 2009 through Sony Music Argentina. The album features Cantora 1 and Cantora 2, the project is Sosa's final album before her death on October 4, 2009.\n",
       "At the 10th Annual Latin Grammy Awards, Cantora 1 was nominated for Album of the Year and won Best Folk Album and Best Recording Package, the latter award went to Alejandro Ros, the art director of the album. Additionally, Sosa won two out of five nominations for the albums at the Gardel Awards 2010, the double album was nominated for Album of the Year and Production of the Year and won Best DVD while both Cantora 1 and Cantora 2 were nominated for Best Female Folk Album, with the former winning the category.\n",
       "The double album was a commercial success, being certified platinum by the CAPIF selling more than 200,000 copies in Argentina, Cantora 1 was also certified platinum selling 40,000 copies while Cantora 2 was certified gold selling 20,000 copies. The album also peaked at numbers 22 and 8 at the Top Latin Albums and Latin Pop Albums charts in United States, respectively, being Sosa's only appearances on both charts.\n",
       "At documentary film titled Mercedes Sosa, Cantora un viaje íntimo was released on 2009, it was directed by Rodrigo Vila and features the recording process of the album as well as testimonies from the different guest artists that appeared on the project.\n",
       "\n",
       "---\n",
       "\n",
       "# Joan Baez discography\n",
       "**URL:** https://en.wikipedia.org/wiki/Joan_Baez_discography\\\n",
       "**Sections:** Studio_albums,1959,1960s,1970s,1980–present,Live_albums,Soundtrack_albums,Compilations,Singles,EPs,Video_albums,Contributions,References,External_links\n",
       "\n",
       "This is a discography for American folk singer and songwriter Joan Baez."
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "query = \"Mercedes Sosa discography\"\n",
    "\n",
    "Markdown(\n",
    "    wikipedia_search.invoke(\n",
    "        {\n",
    "            \"query\": query,\n",
    "        }\n",
    "    )\n",
    ")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "2e25c03e",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def wikipedia_section(\n",
    "    title: str, section: str | None = None, max_chars: int | None = None\n",
    ") -> str:\n",
    "    \"\"\"Retrieve current Wikipedia content.\n",
    "    You must provide the exact title of the page and a valid section name to retrieve.\n",
    "    Always use this tool instead of `fetch_webpage` for Wikipedia pages.\n",
    "    This tool retrieves the current article revision. It must not be used as\n",
    "    historical evidence for questions containing \"as of\" or a past date.\n",
    "\n",
    "    Example usage:\n",
    "        # Retrieve the \"Discography\" section of the \"Mercedes Sosa\" Wikipedia page\n",
    "        wikipedia_section.invoke({\n",
    "            \"title\": \"Mercedes Sosa\",\n",
    "            \"section\": \"Discography\"\n",
    "        })\n",
    "        # Retrive the entire content of the \"Mercedes Sosa\" Wikipedia page\n",
    "        wikipedia_section.invoke({\n",
    "            \"title\": \"Mercedes Sosa\",\n",
    "        })\n",
    "        # Return the first 500 characters of the Python (programming language) Wikipedia page\n",
    "        wikipedia_section.invoke({\n",
    "            \"title\": \"Python (programming language)\",\n",
    "            \"max_chars\": 500\n",
    "        })\n",
    "\n",
    "    Args:\n",
    "        title: The exact title of the wikipedia page to retrieve. For example, \"Mercedes Sosa\" or \"Python (programming language)\".\n",
    "        section: The exact section title to retrieve from the page. For example, \"Discography\" or \"Early life\". Use None to retrieve the beginning of the entire article.\n",
    "        max_chars: The maximum number of characters to retrieve from the section content. None to retrieves from the start of the section to the end of the content.\n",
    "    \"\"\"\n",
    "    # page.section fails to retrieve the section content when the section includes tables or other complex HTML elments\n",
    "    # So we will retrieve the page HTML and parse the content\n",
    "    page = wikipedia.page(title, auto_suggest=False)\n",
    "    soup = BeautifulSoup(page.html(), \"html.parser\")\n",
    "\n",
    "    # Clean the HTML content and convert it to Markdown\n",
    "    for tag in [\n",
    "        \"script\",\n",
    "        \"style\",\n",
    "        \"noscript\",\n",
    "        \"nav\",\n",
    "        \"footer\",\n",
    "        \"header\",\n",
    "        \"svg\",\n",
    "        \"img\",\n",
    "        \"figure\",\n",
    "        \"#toc\",\n",
    "        \"span.mw-editsection\",\n",
    "    ]:\n",
    "        for t in soup.select(tag):\n",
    "            t.decompose()\n",
    "\n",
    "        page_md = md(\n",
    "            soup.prettify(),\n",
    "            heading_style=\"ATX\",\n",
    "            bullets=\"-\",\n",
    "            strip=[\"a\", \"img\"],\n",
    "        )\n",
    "\n",
    "    # Avoid llm stress with unnecesary token consume data\n",
    "    if (i := page_md.find(\"## References\")) > 0:\n",
    "        page_md = page_md[:i]\n",
    "\n",
    "    if section is None:\n",
    "        if max_chars is not None:\n",
    "            content = page_md[:max_chars]\n",
    "        else:\n",
    "            content = page_md\n",
    "\n",
    "        return f\"\"\"# {title}\n",
    "**URL:** {page.url}\n",
    "\n",
    "{content.strip()}\"\"\"\n",
    "\n",
    "    # Look for the section in the markdown content\n",
    "    if (\n",
    "        m := re.search(r\"#{2,4} \" + re.escape(section), page_md, re.IGNORECASE)\n",
    "    ) is None:\n",
    "        return f\"Section '{section}' not found in the page '{title}', you must provide a valid section name.\"\n",
    "\n",
    "    # Returns the max_chars characters of the section content\n",
    "    if max_chars is not None:\n",
    "        content = page_md[m.start() : m.end() + max_chars]\n",
    "    else:\n",
    "        content = page_md[m.start() :]\n",
    "\n",
    "    return f\"\"\"# {title}\n",
    "**URL:** {page.url}\n",
    "**Section:** {section}\n",
    "\n",
    "{content.strip()}\"\"\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "741679f3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "# Mercedes Sosa\n",
       "**URL:** https://en.wikipedia.org/wiki/Mercedes_Sosa\n",
       "**Section:** Discography\n",
       "\n",
       "## Discography\n",
       "\n",
       "Sosa recorded forty albums.\n",
       "[\n",
       "3\n",
       "]\n",
       "[\n",
       "8\n",
       "]\n",
       "\n",
       "### Studio albums\n",
       "\n",
       "| Year | Album details |\n",
       "| --- | --- |\n",
       "| 1962 | *La Voz de la Zafra*  - Label:   RCA |\n",
       "| 1965 | *Canciones Con Fundamento*  - Label: El Grillo |\n",
       "| 1966 | *Hermano*  - Label:   Philips |\n",
       "| 1966 | *Yo No Canto Por Cantar*  - Label: Philips |\n",
       "| 1967 | *Para Cantarle a Mi Gente*  - Label: Philips |\n",
       "| 1968 | *Con Sabor a Mercedes Sosa*  - Label: Philips |\n",
       "| 1969 | *Mujeres Argentinas*  - Label: Philips |\n",
       "| 1970 | *El Grito de la Tierra*  - Label: Philips |\n",
       "| 1970 | *Navidad con Mercedes Sosa*  - Label: Philips |\n",
       "| 1971 | *Homenaje a Violeta Parra*  - Label: Philips |\n",
       "| 1972 | *Hasta la Victoria*  - Label: Philips |\n",
       "| 1972 | *Cantata Sudamericana*  - Label: Philips |\n",
       "| 1973 | *Traigo un Pueblo en Mi Voz*  - Label: Philips |\n",
       "| 1975 | *A Que Florezca Mi Pueblo*  - Label: Philips |\n",
       "| 1976 | *En Dirección Del Viento*  - Label: Philips |\n",
       "| 1977 | *Mercedes Sosa Interpreta a Atahualpa Yupanqui*  - Label: Philips |\n",
       "| 1979 | *Serenata Para la Tierra de Uno*  - Label: Philips |\n",
       "| 1981 | *A Quien Doy / Cuando Me Acuerdo de Mi País*  - Label: Philips |\n",
       "| 1982 | *Como Un Pájaro Libre*  - Label: Philips |\n",
       "| 1983 | *Mercedes Sosa*  - Label: Philips |\n",
       "| 1984 | *¿Será Posible el Sur?*  - Label: Philips |\n",
       "| 1985 | *Vengo a Ofrecer Mi Corazón*  - Label: Philips |\n",
       "| 1986 | *Mercedes Sosa '86*  - Label: Philips |\n",
       "| 1987 | *Mercedes Sosa '87*  - Label: Philips |\n",
       "| 1993 | *Sino*  - Label: Philips/   Polygram |\n",
       "| 1994 | *Gestos de Amor*  - Label:   Polydor |\n",
       "| 1996 | *Escondido en Mi País*  - Label: Polydor |\n",
       "| 1997 | *Alta Fidelidad*  (w/ Charly García )  - Label: Mercury |\n",
       "| 1998 | *Al Despertar*  - Label:   Mercury |\n",
       "| 1999 | *Misa Criolla*  - Label: Mercury |\n",
       "| 2005 | *Corazón Libre*  - Label: Edge |\n",
       "| 2009 | *Cantora 1*  (w/various artists)  - Label: RCA |\n",
       "| 2009 | *Cantora 2*  (w/various artists)  - Label: RCA |\n",
       "| 2011 | *Censurada*  - Label: Philips |\n",
       "| 2015 | *Lucerito*  - Label: RCA |\n",
       "\n",
       "### EPs\n",
       "\n",
       "| Year | EP details |\n",
       "| --- | --- |\n",
       "| 1975 | *Niño de Mañana*  - Label: Philips |\n",
       "\n",
       "### Live albums\n",
       "\n",
       "| Year | Album details |\n",
       "| --- | --- |\n",
       "| 1973 | *Si Se Calla El Cantor*  (with Gloria Martin)  - Label: Philips |\n",
       "| 1980 | *Gravado Ao Vivo No Brasil*  - Label: Philips |\n",
       "| 1982 | *Mercedes Sosa en Argentina*  - Label:   Phonogram   /Philips |\n",
       "| 1985 | *Corazón Americano*  (with Milton Nascimento & León Gieco )  - Label: Philips |\n",
       "| 1989 | *Live in Europe*  - Label: Tropical Music/Polygram Argentina |\n",
       "| 1991 | *De Mí*  - Label: Philips |\n",
       "| 2002 | *Acústico en Vivo*  - Label: Sony Music Argentina |\n",
       "| 2003 | *Argentina Quiere Cantar*  (with Víctor Heredia & León Gieco )  - Label:   Odeon   /   EMI |\n",
       "| 2010 | *Deja la Vida Volar (En Gira)*  - Label: RCA |\n",
       "| 2014 | *Angel*  - Label: Universal Music |\n",
       "| 2024 | *En vivo en el Gran Rex 2006*  - Label: INAMU Discos |\n",
       "| *Mercedes Sosa en Nueva York, 1974*  - Label: Sony Music Argentina |\n",
       "\n",
       "### Compilation albums\n",
       "\n",
       "| Year | Album details |\n",
       "| --- | --- |\n",
       "| 1975 | *Disco de Oro*  - Label: Philips |\n",
       "| 1983 | *Recital*  - Label: Philips |\n",
       "| 1988 | *Amigos Míos*  - Label: Philips |\n",
       "| 1993 | *30 Años*  - Label: Polygram Argentina |\n",
       "| 1995 | *Oro*  - Label: Polygram |\n",
       "| 1997 | *The Best of Mercedes Sosa*  - Label: Mercury |\n",
       "| 2013 | *Siempre en Ti*  - Label: Universal Music |\n",
       "\n",
       "## Filmography\n",
       "\n",
       "- *Güemes, la tierra en armas*\n",
       "  (1971)\n",
       "- *Argentinísima*\n",
       "  (1972)\n",
       "- *Esta es mi Argentina*\n",
       "  (1974)\n",
       "- *Mercedes Sosa, como un pájaro libre*\n",
       "  (1983)\n",
       "- *Será possible el sur: Mercedes Sosa*\n",
       "  (1985)\n",
       "- *Historias de Argentina en vivo*\n",
       "  (2001)"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "title = \"Mercedes Sosa\"\n",
    "section = \"Discography\"\n",
    "\n",
    "Markdown(\n",
    "    wikipedia_section.invoke(\n",
    "        {\n",
    "            \"title\": title,\n",
    "            \"section\": section,\n",
    "        }\n",
    "    )\n",
    ")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "cf1be700",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def arxiv_search(query: str) -> str:\n",
    "    \"\"\"Retrieve the top 5 relevant arxiv papers for a given search query.\n",
    "    Use the exact paper title whenever it is known.\n",
    "    Return the canonical title, authors, arXiv ID,\n",
    "    abstract URL, and direct PDF URL.\n",
    "\n",
    "    Args:\n",
    "        query: Search key term to look up in arxiv.\n",
    "    \"\"\"\n",
    "    search = arxiv.Search(\n",
    "        query=query, max_results=5, sort_by=arxiv.SortCriterion.Relevance\n",
    "    )\n",
    "\n",
    "    documents = [\n",
    "        f\"\"\"# {r.title}\n",
    "\n",
    "**URL:** {r.entry_id}\\\\\n",
    "**PDF:** {r.pdf_url}\\\\\n",
    "**Authors:** {\",\".join(a.name for a in r.authors)}\\\\\n",
    "**Published:** {r.published.date()}\n",
    "\n",
    "{r.summary.strip()}\"\"\"\n",
    "        for r in arxiv_client.results(search)\n",
    "    ]\n",
    "\n",
    "    formatted_docs = \"\\n\\n---\\n\\n\".join(documents)\n",
    "    return formatted_docs\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "3ee570d9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "# Tierkreis: A Dataflow Framework for Hybrid Quantum-Classical Computing\n",
       "\n",
       "**URL:** http://arxiv.org/abs/2211.02350v1\\\n",
       "**PDF:** https://arxiv.org/pdf/2211.02350v1\\\n",
       "**Authors:** Seyon Sivarajah,Lukas Heidemann,Alan Lawrence,Ross Duncan\\\n",
       "**Published:** 2022-11-04\n",
       "\n",
       "We present Tierkreis, a higher-order dataflow graph program representation and runtime designed for compositional, quantum-classical hybrid algorithms. The design of the system is motivated by the remote nature of quantum computers, the need for hybrid algorithms to involve cloud and distributed computing, and the long-running nature of these algorithms. The graph-based representation reflects how designers reason about and visualise algorithms, and allows automatic parallelism and asynchronicity. A strong, static type system and higher-order semantics allow for high expressivity and compositionality in the program. The flexible runtime protocol enables third-party developers to add functionality using any language or environment. With Tierkreis, quantum software developers can easily build, visualise, verify, test, and debug complex hybrid workflows, and immediately deploy them to the cloud or a custom distributed environment.\n",
       "\n",
       "---\n",
       "\n",
       "# Quantum Computing: Vision and Challenges\n",
       "\n",
       "**URL:** http://arxiv.org/abs/2403.02240v5\\\n",
       "**PDF:** https://arxiv.org/pdf/2403.02240v5\\\n",
       "**Authors:** Sukhpal Singh Gill,Oktay Cetinkaya,Stefano Marrone,Daniel Claudino,David Haunschild,Leon Schlote,Huaming Wu,Carlo Ottaviani,Xiaoyuan Liu,Sree Pragna Machupalli,Kamalpreet Kaur,Priyansh Arora,Ji Liu,Ahmed Farouk,Houbing Herbert Song,Steve Uhlig,Kotagiri Ramamohanarao\\\n",
       "**Published:** 2024-03-04\n",
       "\n",
       "The recent development of quantum computing, which uses entanglement, superposition, and other quantum fundamental concepts, can provide substantial processing advantages over traditional computing. These quantum features help solve many complex problems that cannot be solved otherwise with conventional computing methods. These problems include modeling quantum mechanics, logistics, chemical-based advances, drug design, statistical science, sustainable energy, banking, reliable communication, and quantum chemical engineering. The last few years have witnessed remarkable progress in quantum software and algorithm creation and quantum hardware research, which has significantly advanced the prospect of realizing quantum computers. It would be helpful to have comprehensive literature research on this area to grasp the current status and find outstanding problems that require considerable attention from the research community working in the quantum computing industry. To better understand quantum computing, this paper examines the foundations and vision based on current research in this area. We discuss cutting-edge developments in quantum computer hardware advancement and subsequent advances in quantum cryptography, quantum software, and high-scalability quantum computers. Many potential challenges and exciting new trends for quantum technology research and development are highlighted in this paper for a broader debate.\n",
       "\n",
       "---\n",
       "\n",
       "# Entangling Disciplines: Causality, Entropy and Time-Travel Paradoxes on a Quantum Computer\n",
       "\n",
       "**URL:** http://arxiv.org/abs/2506.15909v1\\\n",
       "**PDF:** https://arxiv.org/pdf/2506.15909v1\\\n",
       "**Authors:** Maria Violaris\\\n",
       "**Published:** 2025-06-18\n",
       "\n",
       "Merging disciplines has led to incredible learnings and breakthroughs throughout history, including the discovery of quantum computing: a cross between computation and quantum physics. In this paper, I will discuss how we can cross quantum computing with topics in fundamental physics. This leads to fruitful, interactive learning opportunities that fuse deep open physics problems with key insights about quantum information science. By outlining quantum circuit experiments that can be run on current and near-term quantum computers, I demonstrate how to help learners engage with principles in special relativity, general relativity and thermodynamics. In turn, these connections can advance their understanding of quantum computing. Learners can further explore the quantum computing activities in this paper via the Quantum Paradoxes content series of videos, blogs and code tutorials that I created with IBM Quantum.\n",
       "\n",
       "---\n",
       "\n",
       "# Unifying communication paradigms in measurement-based delegated quantum computing\n",
       "\n",
       "**URL:** http://arxiv.org/abs/2506.21988v2\\\n",
       "**PDF:** https://arxiv.org/pdf/2506.21988v2\\\n",
       "**Authors:** Fabian Wiesner,Jens Eisert,Anna Pappa\\\n",
       "**Published:** 2025-06-27\n",
       "\n",
       "Delegated quantum computing (DQC) allows clients with low quantum capabilities to outsource computations to a server hosting a quantum computer. This process is often envisioned within the measurement-based quantum computing framework, as it naturally facilitates blindness of inputs and computation. Hence, the overall process of setting up and conducting the computation encompasses a sequence of three stages: preparing the qubits, entangling the qubits to obtain the resource state, and measuring the qubits to run the computation. There are two primary approaches to distributing these stages between the client and the server that impose different constraints on cryptographic techniques and experimental implementations. In the prepare-and-send setting, the client prepares the qubits and sends them to the server, while in the receive-and-measure setting, the client receives the qubits from the server and measures them. Although these settings have been extensively studied independently, their interrelation and whether setting-dependent theoretical constraints are inevitable remain unclear. By implementing the key components of most DQC protocols in the respective missing setting, we provide a method to build prospective protocols in both settings simultaneously and to translate existing protocols from one setting into the other.\n",
       "\n",
       "---\n",
       "\n",
       "# Arbitrary Ground State Observables from Quantum Computed Moments\n",
       "\n",
       "**URL:** http://arxiv.org/abs/2312.06975v1\\\n",
       "**PDF:** https://arxiv.org/pdf/2312.06975v1\\\n",
       "**Authors:** Harish J. Vallury,Lloyd C. L. Hollenberg\\\n",
       "**Published:** 2023-12-12\n",
       "\n",
       "The determination of ground state properties of quantum systems is a fundamental problem in physics and chemistry, and is considered a key application of quantum computers. A common approach is to prepare a trial ground state on the quantum computer and measure observables such as energy, but this is often limited by hardware constraints that prevent an accurate description of the target ground state. The quantum computed moments (QCM) method has proven to be remarkably useful in estimating the ground state energy of a system by computing Hamiltonian moments with respect to a suboptimal or noisy trial state. In this paper, we extend the QCM method to estimate arbitrary ground state observables of quantum systems. We present preliminary results of using QCM to determine the ground state magnetisation and spin-spin correlations of the Heisenberg model in its various forms. Our findings validate the well-established advantage of QCM over existing methods in handling suboptimal trial states and noise, extend its applicability to the estimation of more general ground state properties, and demonstrate its practical potential for solving a wide range of problems on near-term quantum hardware."
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Markdown(arxiv_search.invoke({\"query\": \"quantum computing\"}))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "87a7a328",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def web_search(query: str) -> str:\n",
    "    \"\"\"Retrieve the top relevant web search results for a given query.\n",
    "\n",
    "    Args:\n",
    "        query: Search key term to look for\n",
    "    \"\"\"\n",
    "\n",
    "    data = tavily_search.invoke({\"query\": query})\n",
    "    search_docs = data.get(\"results\", [])\n",
    "\n",
    "    formatted_docs = \"\\n\\n---\\n\\n\".join(\n",
    "        [\n",
    "            f\"\"\"# {doc[\"title\"]}\n",
    "**URL:** {doc[\"url\"]}\\\\\n",
    "**SCORE:** {doc[\"score\"]:.2f} (Relevance score from 0 to 1, higher is better)\n",
    "\n",
    "{doc[\"content\"].strip()}\"\"\"\n",
    "            for doc in search_docs\n",
    "        ]\n",
    "    )\n",
    "\n",
    "    return formatted_docs\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "eddc2552",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "# What is LangChain?\n",
       "**URL:** https://aws.amazon.com/what-is/langchain\\\n",
       "**SCORE:** 0.94 (Relevance score from 0 to 1, higher is better)\n",
       "\n",
       "LangChain is an open source framework for building applications based on large language models (LLMs). LLMs are large deep-learning models pre-trained on large amounts of data that can generate responses to user queries—for example, answering questions or creating images from text-based prompts. LangChain provides tools and abstractions to improve the customization, accuracy, and relevancy of the information the models generate. For example, developers can use LangChain components to build new prompt chains or customize existing templates. To do that, machine learning engineers must integrate the LLM with the organization’s internal data sources and apply *prompt engineering*—a practice where a data scientist refines inputs to a generative model with a specific structure and context. For example, you can use LangChain to build applications that read data from stored internal documents and summarize them into conversational responses. LangChain provides AI developers with tools to connect language models with external data sources. In the LangChain framework, a link accepts input from the user and passes it to the LangChain libraries for processing.\n",
       "\n",
       "---\n",
       "\n",
       "# What Is LangChain? | IBM\n",
       "**URL:** https://www.ibm.com/think/topics/langchain\\\n",
       "**SCORE:** 0.92 (Relevance score from 0 to 1, higher is better)\n",
       "\n",
       "LangChain is an open source orchestration framework for application development using large language models (LLMs). Available in both Python- and Javascript-based libraries, LangChain’s tools and APIs simplify the process of building LLM-driven applications like chatbots and AI agents. LangChain serves as a generic interface for nearly any LLM, providing a centralized development environment to build LLM applications and integrate them with external data sources and software workflows. This modular environment also allows for programs that use multiple LLMs: for example, an application that uses one LLM to interpret user queries and another LLM to author a response. ### Integrations with LLMs. LLMs are not standalone applications: they are pre-trained statistical models that must be paired with an application (and, in some cases, specific data sources) in order to meet their purpose. While it’s the GPT model that interprets the user’s input and composes a natural language response, it’s the application that (among other things) provides an interface for the user to type and read and a UX design that governs the chatbot experience.\n",
       "\n",
       "---\n",
       "\n",
       "# langchain-ai/langchain: The agent engineering platform.\n",
       "**URL:** https://github.com/langchain-ai/langchain\\\n",
       "**SCORE:** 0.92 (Relevance score from 0 to 1, higher is better)\n",
       "\n",
       "LangChain is a framework for building agents and LLM-powered applications. It helps you chain together interoperable components and third-party integrations\n",
       "\n",
       "---\n",
       "\n",
       "# What Is LangChain? Examples and definition\n",
       "**URL:** https://cloud.google.com/use-cases/langchain\\\n",
       "**SCORE:** 0.91 (Relevance score from 0 to 1, higher is better)\n",
       "\n",
       "LangChain is an open-source orchestration framework that simplifies building applications with large language models (LLMs). It provides tools and components to connect LLMs with various data sources, enabling the creation of complex, multi-step workflows. Available as libraries in Python and JavaScript, LangChain helps developers enhance LLM capabilities beyond text generation by linking them to external data and computation. This helps facilitate the development of advanced AI applications like intelligent chatbots, sophisticated question-answering systems, and automated data analysis tools. Build AI-powered apps on Vertex AI (now Gemini Enterprise Agent Platform) with LangChain. # LangChain and AI. Moreover, LangChain provides a consistent interface for interacting with LLMs, abstracting away the complexities of API management. LangChain works by \"chaining\" together different components to create a cohesive workflow for LLM-powered applications. LangChain provides a suite of features designed to facilitate the development of LLM-powered applications. Fully managed Google Cloud service enabling developers to deploy, manage, and scale AI agents in production.\n",
       "\n",
       "---\n",
       "\n",
       "# Introduction to LangChain\n",
       "**URL:** https://www.geeksforgeeks.org/artificial-intelligence/introduction-to-langchain\\\n",
       "**SCORE:** 0.89 (Relevance score from 0 to 1, higher is better)\n",
       "\n",
       "# Introduction to LangChain. LangChain is an open-source framework that simplifies building applications using large language models. It helps developers connect LLMs with external data, tools and workflows and is available in both Python and JavaScript. * ****Chains:**** Define a sequence of steps where each step can use an LLM, process data or call tools. > For example, a user might ask, **“What’s the weather like today?”** This query serves as the input to the LangChain pipeline. Similarity Search:**** This vector is compared with vectors stored in a database to find the most relevant matches based on meaning. Generating a Response:**** The retrieved context is passed to a language model, which processes it and generates a meaningful answer. > For example, if the query is about the weather, the LLM might generate a response like, **“Today’s weather is sunny with a high of 75°F.”**. In this example, the API key is added directly in the code and passed while initializing the model."
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Markdown(web_search.invoke({\"query\": \"Langchain framework\"}))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "f600c236",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def fetch_webpage(url: str, max_chars: int | None = None) -> str:\n",
    "    \"\"\"Fetch a webpage content.\n",
    "    DON'T USE THIS TOOL to try to fetch Wikipedia or arxiv content, use the dedicated tools for those sources `wikipedia_section` and `arxiv_search`.\n",
    "\n",
    "\n",
    "    Args:\n",
    "        url: Webpage URL to fetch\n",
    "        max_chars: Maximum number of characters to return from the webpage content. By default, returns the entire content.\n",
    "    \"\"\"\n",
    "    try:\n",
    "        response = requests.get(url)\n",
    "        response.raise_for_status()\n",
    "        soup = BeautifulSoup(response.text, \"html.parser\")\n",
    "        for tag in [\n",
    "            \"script\",\n",
    "            \"style\",\n",
    "            \"noscript\",\n",
    "            \"nav\",\n",
    "            \"link\",\n",
    "            \"footer\",\n",
    "            \"header\",\n",
    "            \"iframe\",\n",
    "            \"svg\",\n",
    "            \"img\",\n",
    "            \"figure\",\n",
    "        ]:\n",
    "            for t in soup.select(tag):\n",
    "                t.decompose()\n",
    "\n",
    "        page_md = md(\n",
    "            soup.prettify(),\n",
    "            heading_style=\"ATX\",\n",
    "            bullets=\"-\",\n",
    "            strip=[\"a\", \"img\"],\n",
    "        )\n",
    "\n",
    "        if max_chars is not None:\n",
    "            return page_md[:max_chars]\n",
    "        return page_md\n",
    "\n",
    "    except requests.exceptions.RequestException as e:\n",
    "        return f\"Failed to fetch the webpage. Error: {str(e)}.\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "ffc2fa56",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/markdown": [
       "What Is LangChain? | IBM\n",
       "\n",
       "\n",
       "\n",
       "Get hands-on experience with IBM tech\n",
       "\n",
       "Join one of the largest technical IBM community gatherings\n",
       "\n",
       "# What is LangChain?\n",
       "\n",
       "Authors\n",
       "\n",
       "Dave Bergmann\n",
       "\n",
       "Senior Staff Writer, AI Models\n",
       "\n",
       "IBM Think\n",
       "\n",
       "Cole Stryker\n",
       "\n",
       "Staff Editor, AI Models\n",
       "\n",
       "IBM Think\n",
       "\n",
       "## LangChain overview\n",
       "\n",
       "LangChain is an open source orchestration framework for application development using\n",
       "large language models (LLMs)\n",
       ". Available in both Python- and Javascript-based libraries, LangChain’s tools and\n",
       "APIs\n",
       "simplify the process of building LLM-driven applications like\n",
       "chatbots\n",
       "and\n",
       "AI agents\n",
       ".\n",
       "\n",
       "LangChain serves as a generic interface for nearly any LLM, providing a centralized development environment to build LLM applications and integrate them with external data sources and software workflows. LangChain’s module-based approach allows developers and data scientists to dynamically compare different prompts and even different\n",
       "foundation models\n",
       "with minimal need to rewrite code. This modular environment also allows for programs that use multiple LLMs: for example, an application that uses one LLM to interpret user queries and another LLM to author a response.\n",
       "\n",
       "Launched by Harrison Chase in October 2022, LangChain enjoyed a meteoric rise to prominence: as of June 2023, it was the single fastest-growing\n",
       "open source\n",
       "project on Github.\n",
       "1\n",
       "Coinciding with the momentous launch of OpenAI’s\n",
       "ChatGPT\n",
       "the following month, LangChain has played a significant role in making\n",
       "generative AI\n",
       "(genAI) more accessible to enthusiasts and startups in the wake of its widespread popularity. Advancements in accessibility for agentic AI are currently enabling a revolution in\n",
       "automation\n",
       ".\n",
       "\n",
       "LangChain can facilitate most use cases for LLMs and\n",
       "natural language processing (NLP)\n",
       ", like chatbots,\n",
       "intelligent search\n",
       ",\n",
       "question-answering\n",
       ",\n",
       "summarization\n",
       "services or even\n",
       "AI agents\n",
       "capable of\n",
       "robotic process automation\n",
       ".\n",
       "\n",
       "### Integrations with LLMs\n",
       "\n",
       "LLMs are not standalone applications: they are pre-trained statistical models that must be paired with an application (and, in some cases, specific data sources) in order to meet their purpose.\n",
       "\n",
       "For example, Chat-GPT is not an LLM: it is a chatbot application that, depending on the version you’ve chosen, uses the GPT-3.5 or GPT-4 language model. While it’s the GPT model that interprets the user’s input and composes a natural language response, it’s the application that (among other things) provides an interface for the user to type and read and a UX design that governs the chatbot experience. Even at the enterprise level, Chat-GPT is not the only application using the GPT model: Microsoft uses GPT-4 to power Bing Chat.\n",
       "\n",
       "Furthermore, though foundation models (like those powering LLMs) are pre-trained on massive\n",
       "datasets\n",
       ", they are not omniscient. If a particular task requires access to specific contextual information, like internal documentation or domain expertise, LLMs must be connected to those external data sources. Even if you simply want your model to reflect real-time awareness of current events, it requires external information: a model’s internal data is only up-to-date through the time period during which it was pre-trained.\n",
       "\n",
       "Likewise, if a given generative AI task requires access to external software workflows—for example, if you wanted your virtual agent to\n",
       "integrate with Slack\n",
       "—then you will need a way to integrate the LLM with the\n",
       "API\n",
       "for that software.\n",
       "\n",
       "While these integrations can generally be achieved with fully manual code,\n",
       "orchestration\n",
       "frameworks such as LangChain and the\n",
       "IBM watsonx\n",
       "portfolio of artificial intelligence products greatly simplify the process. They also make it much easier to experiment with different LLMs to compare results, as different models can be swapped in and out with minimal changes to code.\n",
       "\n",
       "## How does LangChain work?\n",
       "\n",
       "At LangChain’s core is a development environment that streamlines the programming of LLM applications through the use of\n",
       "*abstraction*\n",
       ": the simplification of code by representing one or more complex processes as a named component that encapsulates all of its constituent steps.\n",
       "\n",
       "Abstractions are a common element of everyday life and language. For example, “\n",
       "*π*\n",
       "” allows us to represent the ratio of the length of a circle’s circumference to that of its diameter without having to write out its infinite digits. Similarly, a thermostat allows us to control the temperature in our home without needing to understand the complex circuitry this entails—we only need to know how different thermostat settings translate to different temperatures.\n",
       "\n",
       "LangChain is essentially a library of abstractions for Python and Javascript, representing common steps and concepts necessary to work with language models. These modular components—like functions and object classes—serve as the building blocks of generative AI programs. They can be “\n",
       "*chained*\n",
       "” together to create applications, minimizing the amount of code and fine understanding required to execute complex NLP tasks. Though LangChain’s abstracted approach may limit the extent to which an expert programmer can finely customize an application, it empowers specialists and newcomers alike to quickly experiment and prototype.\n",
       "\n",
       "### Importing language models\n",
       "\n",
       "Nearly any LLM can be used in LangChain. Importing language models into LangChain is easy, provided you have an API key. The\n",
       "*LLM*\n",
       "class is designed to provide a standard interface for all models.\n",
       "\n",
       "Most LLM providers will require you to create an account in order to receive an API key. Some of these APIs—particularly those for proprietary closed-source models, like those offered by OpenAI or Anthropic—may have associated costs.\n",
       "\n",
       "Many open source models, like Meta AI’s LLaMa, Deepseek's Deepseek-LLM, IBM's\n",
       "Granite\n",
       "and Google’s Flan-T5, can be accessed through\n",
       "\n",
       "Hugging Face\n",
       ".\n",
       "IBM watsonx\n",
       ", through its\n",
       "partnership with Hugging Face\n",
       ", also offers a curated suite of open source models. Creating an account with either service will allow you to generate an API key for any of the models offered by that provider.\n",
       "\n",
       "LangChain is not limited to out-of-the-box foundation models: the\n",
       "\n",
       "*CustomLLM*\n",
       "class\n",
       "allows for custom LLM wrappers. Likewise, you can use the\n",
       "IBM watsonx APIs and Python SDK\n",
       ", which includes a LangChain integration, to build applications in LangChain with models that you’ve already trained or fine-tuned for your specific needs using the\n",
       "*WatsonxLLM*\n",
       "class (and that model’s specific\n",
       "project ID\n",
       ").\n",
       "\n",
       "### Prompt templates\n",
       "\n",
       "Prompts are the instructions given to an LLM. The “art” of composing prompts that effectively provide the context necessary for the LLM to interpret input and structure output in the way most useful to you is often called\n",
       "prompt engineering\n",
       ".\n",
       "\n",
       "The\n",
       "*PromptTemplate*\n",
       "class in LangChain formalizes the composition of prompts without the need to manually hard code context and queries. Important elements of a prompt are likewise entered as formal classes, like\n",
       "*input\\_variables*\n",
       ". A prompt template can thus contain and reproduce context, instructions (like “do not use technical terms”), a set of examples to guide its responses (in what is called “few-shot prompting”), a specified output format or a standardized question to be answered. You can save and name an effectively structured prompt template and easily reuse it as needed.\n",
       "\n",
       "Though these elements can all be manually coded,\n",
       "*PromptTemplate*\n",
       "modules empower smooth integration with other LangChain features, like the eponymous\n",
       "*chains.*\n",
       "\n",
       "### Chains\n",
       "\n",
       "As its name implies,\n",
       "*chains*\n",
       "are the core of LangChain’s workflows. They combine LLMs with other components, creating applications by executing a sequence of functions.\n",
       "\n",
       "The most basic chain is\n",
       "LLMChain\n",
       "\n",
       ". It simply calls a model and prompt template for that model. For example, imagine you saved a prompt as “ExamplePrompt” and wanted to run it against Flan-T5. You can import LLMChain from langchain.chains, then define\n",
       "\n",
       "chain\\_example = LLMChain(llm = flan-t5, prompt = ExamplePrompt)\n",
       "\n",
       ". To run the chain for a given input, you simply call\n",
       "\n",
       "chain\\_example.run(“input”).\n",
       "\n",
       "To use the output of one function as the input for the next function, you can use\n",
       "*SimpleSequentialChain*\n",
       ". Each function could utilize different prompts, different tools, different parameters or even different models, depending on your specific needs.\n",
       "\n",
       "### Indexes\n",
       "\n",
       "To achieve certain tasks, LLMs will need access to specific external data sources not included in its training dataset, such as internal documents, emails or datasets. LangChain collectively refers to such external documentation as “indexes”.\n",
       "\n",
       "#### Document loaders\n",
       "\n",
       "LangChain offers a wide variety of\n",
       "\n",
       "document loaders\n",
       "for third party applications. This allows for easy importation of data from sources like file storage services (like Dropbox, Google Drive and Microsoft OneDrive), web content (like YouTube, PubMed or specific URLs), collaboration tools (like Airtable, Trello, Figma and Notion), databases (like Pandas, MongoDB and Microsoft), among many others.\n",
       "\n",
       "#### Vector databases\n",
       "\n",
       "Unlike “traditional” structured databases,\n",
       "vector databases\n",
       "represent data points by converting them into\n",
       "*vector embeddings*\n",
       ": numerical representations in the form of vectors with a fixed number of dimensions, often clustering related data points using\n",
       "unsupervised learning methods\n",
       ". This enables low latency queries, even for massive datasets, which greatly increases efficiency.\n",
       "Vector embeddings\n",
       "also store each vector’s metadata, further enhancing search possibilities.\n",
       "\n",
       "LangChain provides integrations for over 25 different embedding methods, as well as for over 50 different vector stores (both cloud-hosted and local).\n",
       "\n",
       "#### Text splitters\n",
       "\n",
       "To increase speed and reduce computational demands, it’s often wise to split large text documents into smaller pieces. LangChain’s\n",
       "*TextSplitters*\n",
       "split text up into small, semantically meaningful chunks that can then be combined using methods and parameters of your choosing.\n",
       "\n",
       "#### Retrieval\n",
       "\n",
       "Once external sources of knowledge have been connected, the model must be able to quickly retrieve and integrate relevant information as needed. Like watsonx, LangChain offers\n",
       "retrieval augmented generation (RAG)\n",
       ": its\n",
       "*retriever*\n",
       "modules accept a string query as an input and return a list of\n",
       "*Document*\n",
       "’s as output.\n",
       "\n",
       "With LangChain, we can also\n",
       "build agentic RAG systems\n",
       ". In traditional RAG applications, the LLM is provided with a vector database to reference when forming its responses. In contrast, agentic AI applications are not restricted to only data retrieval.\n",
       "Agenic RAG\n",
       "can also encompass tools for tasks such as solving mathematical calculations, writing emails, performing data analysis and more.\n",
       "\n",
       "### Memory\n",
       "\n",
       "LLMs, by default, do not have any long-term memory of previous interactions (unless that chat history is used as input for a query). LangChain solves this problem with simple utilities for adding memory to a system, with options ranging from retaining the entirety of all conversations to retaining a summarization of the conversation thus far to retaining the\n",
       "*n*\n",
       "most recent exchanges.\n",
       "\n",
       "### Tools\n",
       "\n",
       "Despite their heralded power and versatility, LLMs have important limitations: namely, a lack of up-to-date information, a lack of domain-specific expertise and a general difficulty with math.\n",
       "\n",
       "LangChain\n",
       "*tools*\n",
       "are a set of functions that empower LangChain agents to interact with real-world information in order to expand or improve the services it can provide. Examples of prominent pre-built LangChain tools include:\n",
       "\n",
       "- **Wolfram Alpha:**\n",
       "  provides access to powerful computational and data visualization functions, enabling sophisticated mathematical capabilities.\n",
       "- **Google Search:**\n",
       "  provides access to Google Search, equipping applications and agents with real-time information.\n",
       "- **OpenWeatherMap:**\n",
       "  fetches weather information.\n",
       "- **Wikipedia**\n",
       "  : provides efficient access to information from Wikipedia articles.\n",
       "\n",
       "## The latest tech news, backed by expert insights\n",
       "\n",
       "Stay up to date on the most important—and intriguing—industry trends on AI, automation, data and beyond with the Think Newsletter, delivered twice weekly.\n",
       "\n",
       "## Thank you!\n",
       "\n",
       "You are subscribed.\n",
       "\n",
       "## LangChain agents\n",
       "\n",
       "We can build an\n",
       "agent\n",
       "with the LangChain framework to give an LLM the ability to make decisions, use tools and complete complex tasks step-by-step, rather than just generating a single text response. Unlike a simple prompt-response interaction with just an LLM, an agent powered by LangChain can\n",
       "think\n",
       ",\n",
       "plan\n",
       ", execute a sequence of actions,\n",
       "learn\n",
       "and adapt.\n",
       "\n",
       "LangChain provides a streamlined user experience with a ready-made, extensible framework for creating AI agents, so there’s no need to build new tool selection logic, reasoning loops (such as for\n",
       "ReAct agents\n",
       "), observation/action tracking or prompt orchestration and formatting.\n",
       "\n",
       "The specific LangChain packages, classes and methods vary depending on the AI platform you intend to use. Some key components of the\n",
       "*WatsonxLLM*\n",
       "class that allow for communication with watsonx.ai models using LangChain include:\n",
       "\n",
       "- ***langchain\\_ibm***\n",
       "  : The package responsible for the LangChain IBM integration. It is necessary to install this package to use any of the following classes and methods.\n",
       "- ***ibm\\_watsonx\\_ai***\n",
       "  : The library that allows connection to watsonx.ai services like IBM Cloud and IBM Cloud Pak for Data.\n",
       "- ***APIClient***\n",
       "  : The main class of the\n",
       "  *ibm\\_watsonx\\_ai*\n",
       "  library that manages the API service resources. The parameters include the API credentials and endpoint.\n",
       "- ***WatsonxLLM***\n",
       "  : The wrapper for IBM watsonx.ai foundation models. This wrapper provides chain integration and is necessary to import. The parameters include the model ID, watsonx.ai API key, URL endpoint, project ID as well as any LLM parameters.\n",
       "- ***ModelInference***\n",
       "  : The class that instantiates the model interface. The parameters include the model ID, watsonx.ai credentials, project ID, model parameters and more. Once instantiated, the model can then be passed into the class.\n",
       "- ***invoke***\n",
       "  : The method that calls the model directly with a single prompt of string type.\n",
       "- ***generate***\n",
       "  : The method that calls the model with multiple prompts of string type in a list.\n",
       "\n",
       "Another LangChain class for building AI agents with the integration of tool calling and chaining with watsonx.ai models is\n",
       "*ChatWatsonx*\n",
       ". This class, which is leveraged in many of our\n",
       "tutorials\n",
       ", uses the\n",
       "*bind\\_tools*\n",
       "method to pass a list of tools to the LLM upon each iteration. These can include both custom and pre-built tools. To retrieve the AI agent response, the\n",
       "*invoke*\n",
       "method can be used. Once the agent is invoked, the\n",
       "*tool\\_calls*\n",
       "attribute of the response displays the name, arguments, id and type of each tool call made, if any.\n",
       "\n",
       "### LangGraph\n",
       "\n",
       "LangGraph\n",
       ", created by LangChain, is an open source AI agent framework that supports multi-agent orchestration and enables developers to build\n",
       "agentic workflows\n",
       "where different agents interact, specialize and collaborate.\n",
       "\n",
       "At its core, LangGraph uses the power of graph-based architectures to model and manage the intricate relationships between various components of an\n",
       "AI agent workflow\n",
       ". Combined with the human-in-the-loop monitoring mechanism and a set of API and tool integrations, LangGraph provides users with a versatile platform for developing AI solutions and workflows including\n",
       "chatbots\n",
       "\n",
       ", state graphs and\n",
       "\n",
       "other agent-based systems\n",
       ".\n",
       "\n",
       "With the\n",
       "*langchain-mcp-adapters*\n",
       "library, LangGraph agents can also use tools defined on model context protocol (MCP) servers. The\n",
       "*mcp*\n",
       "library allows users to build custom MCP servers as well. Essentially, MCP enables a secure connection between an AI system, such as an AI agent, and external tools. Thus, various LLMs can connect to the same tools and data sources given the standard MCP.\n",
       "\n",
       "## LangSmith\n",
       "\n",
       "Released in the fall of 2023, LangSmith aims to bridge the gap between the accessible prototyping capabilities that brought LangChain to prominence and building production-quality LLM applications.\n",
       "\n",
       "LangSmith provides tools to monitor, evaluate and\n",
       "debug\n",
       "applications, including the ability to automatically trace all model calls to spot errors and test performance under different model configurations. The use of LangSmith is not limited to applications built using the LangChain ecosystem. The\n",
       "evaluation\n",
       "of agent performance is done using LLM-as-a-judge evaluators. This\n",
       "observability\n",
       "and these key metrics aim to optimize more robust, cost-efficient applications.\n",
       "\n",
       "## Getting started with LangChain\n",
       "\n",
       "LangChain is open source and free to use: source code is\n",
       "\n",
       "available for download on Github\n",
       ".\n",
       "\n",
       "LangChain can also be installed on Python with a simple pip command:\n",
       "*pip install langchain*\n",
       ". To install all LangChain dependencies (rather than only those you find necessary), you can run the command\n",
       "*pip install langchain[all]*\n",
       ".\n",
       "\n",
       "Many step-by-step tutorials are provided by IBM including\n",
       "LangChain tool calling\n",
       ",\n",
       "agentic RAG\n",
       ",\n",
       "LLM agent orchestration\n",
       ",\n",
       "agentic chunking\n",
       "and\n",
       "more\n",
       ".\n",
       "\n",
       "## LangChain use cases\n",
       "\n",
       "AI Applications made with LangChain provide great utility for a variety of use cases, from straightforward question-answering and text generation tasks to more complex solutions that use an LLM as a “reasoning engine.”\n",
       "\n",
       "### Chatbots\n",
       "\n",
       "Chatbots are among the most intuitive uses of LLMs. LangChain can be used to provide proper context for the specific use of a chatbot, and to integrate chatbots into existing communication channels and workflows with their own APIs.\n",
       "\n",
       "### Summarization\n",
       "\n",
       "Language models can be tasked with summarizing many types of text, from breaking down complex academic articles and transcripts to providing a digest of incoming emails.\n",
       "\n",
       "### Question answering\n",
       "\n",
       "Using specific documents or specialized knowledge bases (like Wolfram, arXiv or PubMed), LLMs can retrieve relevant information from storage and articulate helpful answers). If fine-tuned or properly prompted, some LLMs can answer many questions even without external information.\n",
       "\n",
       "### Data augmentation\n",
       "\n",
       "LLMs can be used to generate\n",
       "synthetic data\n",
       "for use in\n",
       "machine learning\n",
       ". For example, an LLM can be trained to generate additional data samples that closely resemble the data points in a training dataset.\n",
       "\n",
       "### Virtual agents\n",
       "\n",
       "Integrated with the right workflows, LangChain’s Agent modules can use an LLM to autonomously determine next steps and take action using robotic process automation (RPA).\n",
       "\n",
       "Link copied"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "url = \"https://www.ibm.com/think/topics/langchain\"\n",
    "\n",
    "Markdown(fetch_webpage.invoke({\"url\": url}))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "6474b647",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def read_excel_file(file_path: str, sheet_name: str = \"Sheet0\") -> str:\n",
    "    \"\"\"Read an Excel file and return its content as a string.\n",
    "\n",
    "    Args:\n",
    "        file_path: Path to the Excel file.\n",
    "        sheet_name: Name of the sheet to read from the Excel file.\n",
    "    \"\"\"\n",
    "    try:\n",
    "        excel_file = pd.ExcelFile(file_path)\n",
    "        sheet_name = (\n",
    "            sheet_name\n",
    "            if sheet_name in excel_file.sheet_names\n",
    "            else excel_file.sheet_names[0]\n",
    "        )\n",
    "        df = pd.read_excel(excel_file, sheet_name=sheet_name)\n",
    "        return df.to_string(index=False)\n",
    "\n",
    "    except Exception as e:\n",
    "        return f\"Failed to read the Excel file. Error: {str(e)}\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "216efb0c",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def get_youtube_transcript(video_url: str) -> str:\n",
    "    \"\"\"Retrieve the transcription of a YouTube video.\n",
    "\n",
    "    Args:\n",
    "        video_url: The URL of the YouTube video.\n",
    "    \"\"\"\n",
    "    try:\n",
    "        # Extract the video ID from the URL\n",
    "        video_id = re.search(\n",
    "            r\"https?://www.youtube.com/watch\\?v=(?P<video_id>[a-zA-Z0-9_-]+)\", video_url\n",
    "        )\n",
    "        if not video_id:\n",
    "            return \"Invalid YouTube URL. Please provide a valid URL.\"\n",
    "        video_id = video_id.group(\"video_id\")\n",
    "\n",
    "        transcript = ytt_api.fetch(video_id)\n",
    "        return f\"# YouTube video transcript:\\n{'=' * 50}\\n\" + \"\\n\".join(\n",
    "            [\n",
    "                f\" - [{f.start:05.2f}:{(f.start + f.duration):05.2f}] {f.text}\"\n",
    "                for f in transcript\n",
    "            ]\n",
    "        )\n",
    "\n",
    "    except Exception as e:\n",
    "        return f\"An error occurred: {e}\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "2f0c0a86",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'An error occurred: \\nCould not retrieve a transcript for the video https://www.youtube.com/watch?v=L1vXCYZAYYM! This is most likely caused by:\\n\\nSubtitles are disabled for this video\\n\\nIf you are sure that the described cause is not responsible for this error and that a transcript should be retrievable, please create an issue at https://github.com/jdepoix/youtube-transcript-api/issues. Please add which version of youtube_transcript_api you are using and provide the information needed to replicate the error. Also make sure that there are no open issues which already describe your problem!'"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fetch = get_youtube_transcript.invoke(\n",
    "    {\"video_url\": \"https://www.youtube.com/watch?v=L1vXCYZAYYM\"}\n",
    ")\n",
    "\n",
    "fetch\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "a42cefb6",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def execute_python_file(file_path: str) -> str:\n",
    "    \"\"\"Execute a Python file and return the output.\n",
    "\n",
    "    Args:\n",
    "        file_path: The path to the Python file to execute.\n",
    "    \"\"\"\n",
    "\n",
    "    try:\n",
    "        result = subprocess.run([\"python\", file_path], check=True, capture_output=True)\n",
    "        return result.stdout.decode()\n",
    "    except subprocess.CalledProcessError as e:\n",
    "        return f\"Error executing subprocess: {e}\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "dd1e2a9b",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def execute_python_code(code: str) -> str:\n",
    "    \"\"\"Execute a Python code snippet and return the output.\n",
    "\n",
    "    Args:\n",
    "        code: The Python code to execute.\n",
    "    \"\"\"\n",
    "    try:\n",
    "        result = subprocess.run([\"python\", \"-c\", code], check=True, capture_output=True)\n",
    "        return result.stdout.decode()\n",
    "    except subprocess.CalledProcessError as e:\n",
    "        return f\"Error executing subprocess: {e}\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "b8bca52c",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def transcript_audio_file(file_path: str) -> str:\n",
    "    \"\"\"Transcribe an audio file using OpenAI's Whisper model.\n",
    "\n",
    "    Args:\n",
    "        file_path: The path to the audio file to transcribe.\n",
    "    \"\"\"\n",
    "    try:\n",
    "        client = InferenceClient(\n",
    "            provider=\"hf-inference\",\n",
    "            api_key=os.environ[\"HF_TOKEN\"],\n",
    "        )\n",
    "\n",
    "        output = client.automatic_speech_recognition(\n",
    "            file_path, model=\"openai/whisper-large-v3-turbo\"\n",
    "        )\n",
    "        return output.text\n",
    "\n",
    "    except Exception as e:\n",
    "        return str(e)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "4961e12a",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def describe_image_file(file_path: str) -> str:\n",
    "    \"\"\"Describe an image file using OpenAI's CLIP model.\n",
    "\n",
    "    Args:\n",
    "        file_path: The path to the image file to describe.\n",
    "    \"\"\"\n",
    "    try:\n",
    "        extension = Path(file_path).suffix.lower()\n",
    "\n",
    "        mime_types = {\n",
    "            \".jpg\": \"image/jpeg\",\n",
    "            \".jpeg\": \"image/jpeg\",\n",
    "            \".png\": \"image/png\",\n",
    "            \".gif\": \"image/gif\",\n",
    "            \".bmp\": \"image/bmp\",\n",
    "            \".tiff\": \"image/tiff\",\n",
    "        }\n",
    "\n",
    "        if extension not in mime_types:\n",
    "            return f\"Unsupported file type: {extension}. Supported types are: {', '.join(mime_types.keys())}\"\n",
    "\n",
    "        mime_type = mime_types[extension]\n",
    "        with open(file_path, \"rb\") as f:\n",
    "            image_bytes = f.read()\n",
    "            image_base64 = base64.b64encode(image_bytes).decode(\"utf-8\")\n",
    "\n",
    "        llm = ChatOpenAI(model_name=\"gpt-4o\", temperature=0)\n",
    "\n",
    "        message = {\n",
    "            \"role\": \"user\",\n",
    "            \"content\": [\n",
    "                {\n",
    "                    \"type\": \"text\",\n",
    "                    \"text\": \"\"\"Describe the content of this image:\n",
    "                            - In case of an image of chess, describe the position of the pieces and the state of the game.\n",
    "                            - In case there is text in the image, transcribe it and describe it.\n",
    "                            - In case of a graph, describe the axes, the data and the trends.\n",
    "                            - In case of a map, describe the location and the features.\n",
    "                            - In case of a diagram, describe the components and their relationships.\n",
    "                            - In case of a photo, describe the scene, the objects and the people.\"\"\",\n",
    "                },\n",
    "                {\n",
    "                    \"type\": \"image\",\n",
    "                    \"base64\": f\"{image_base64}\",\n",
    "                    \"mime_type\": mime_type,\n",
    "                },\n",
    "            ],\n",
    "        }\n",
    "\n",
    "        return llm.invoke([message])\n",
    "\n",
    "    except Exception as e:\n",
    "        return str(e)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "85e7463a",
   "metadata": {},
   "outputs": [],
   "source": [
    "@tool\n",
    "def download_and_read_pdf(url: str) -> str:\n",
    "    \"\"\"Download and extract text from a PDF.\n",
    "\n",
    "    Use this tool to inspect the complete contents of academic papers,\n",
    "    especially acknowledgments, funding statements, grant numbers,\n",
    "    award numbers, references, and appendices.\n",
    "\n",
    "    The URL must be a verified PDF URL, not a guessed URL.\n",
    "\n",
    "    Args:\n",
    "        url: The URL of the PDF file to download.\n",
    "    \"\"\"\n",
    "    try:\n",
    "        response = requests.get(url)\n",
    "        response.raise_for_status()\n",
    "\n",
    "        with NamedTemporaryFile(suffix=\".pdf\") as fp:\n",
    "            fp.write(response.content)\n",
    "            text_content = \"\"\n",
    "            pdf_reader = PdfReader(fp)\n",
    "            for page in pdf_reader.pages:\n",
    "                text_content += page.extract_text() + \"\\n\"\n",
    "\n",
    "        pdf_resume = f\"\"\"# PDF Content from {url}\n",
    "{text_content[:6000]}\n",
    "...\n",
    "{text_content[-6000:]}\n",
    "\"\"\"\n",
    "        return pdf_resume\n",
    "\n",
    "    except Exception as e:\n",
    "        return f\"Failed to download or read the PDF. Error: {str(e)}\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "14934f80",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# PDF Content from https://arxiv.org/pdf/2305.10403.pdf\n",
      "PaLM 2 Technical Report\n",
      "Google *\n",
      "Abstract\n",
      "We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities\n",
      "and is more compute-efficient than its predecessor PaLM. PaLM 2 is a Transformer-based model trained using a\n",
      "mixture of objectives. Through extensive evaluations on English and multilingual language, and reasoning tasks,\n",
      "we demonstrate that PaLM 2 has significantly improved quality on downstream tasks across different model sizes,\n",
      "while simultaneously exhibiting faster and more efficient inference compared to PaLM. This improved efficiency\n",
      "enables broader deployment while also allowing the model to respond faster, for a more natural pace of interaction.\n",
      "PaLM 2 demonstrates robust reasoning capabilities exemplified by large improvements over PaLM on BIG-Bench\n",
      "and other reasoning tasks. PaLM 2 exhibits stable performance on a suite of responsible AI evaluations, and enables\n",
      "inference-time control over toxicity without additional overhead or impact on other capabilities. Overall, PaLM 2\n",
      "achieves state-of-the-art performance across a diverse set of tasks and capabilities.\n",
      "When discussing the PaLM 2 family, it is important to distinguish between pre-trained models (of various sizes),\n",
      "fine-tuned variants of these models, and the user-facing products that use these models. In particular, user-facing\n",
      "products typically include additional pre- and post-processing steps. Additionally, the underlying models may evolve\n",
      "over time. Therefore, one should not expect the performance of user-facing products to exactly match the results\n",
      "reported in this report.\n",
      "*See Section 7 for a list of authors. Please send correspondence to palm2-technical-report@google.com . NOTE: This paper refers to the\n",
      "version of PaLM 2 we announced in May 2023. Our research and development of PaLM 2 continues, including version updates for PaLM 2 as it is\n",
      "implemented into products.\n",
      "1arXiv:2305.10403v3  [cs.CL]  13 Sep 2023\n",
      "Contents\n",
      "1 Introduction 3\n",
      "2 Scaling law experiments 7\n",
      "2.1 Scaling laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7\n",
      "2.2 Downstream metric evaluations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8\n",
      "3 Training dataset 9\n",
      "4 Evaluation 9\n",
      "4.1 Language proficiency exams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10\n",
      "4.2 Classification and question answering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10\n",
      "4.3 Reasoning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13\n",
      "4.4 Coding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15\n",
      "4.5 Translation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16\n",
      "4.6 Natural language generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19\n",
      "4.7 Memorization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21\n",
      "5 Responsible usage 23\n",
      "5.1 Inference-time control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24\n",
      "5.2 Recommendations for developers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26\n",
      "6 Conclusion 26\n",
      "7 Authorship, attribution, and acknowledgements 28\n",
      "A Detailed results 42\n",
      "A.1 Scaling laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42\n",
      "A.2 Instruction tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43\n",
      "A.3 Multilingual commonsense reasoning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43\n",
      "A.4 Coding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45\n",
      "A.5 Natural language generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45\n",
      "B Examples of model capabilities 45\n",
      "B.1 Multilinguality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45\n",
      "B.2 Creative generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50\n",
      "B.3 Coding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50\n",
      "C Language proficiency exams 62\n",
      "D Responsible AI 63\n",
      "D.1 Dataset analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63\n",
      "D.2 Evaluation approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66\n",
      "D.3 Dialog uses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67\n",
      "D.4 Classification uses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71\n",
      "D.5 Translation uses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73\n",
      "D.6 Question answering uses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77\n",
      "D.7 Language modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80\n",
      "D.8 Measurement quality rubrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82\n",
      "D.9 CrowdWorksheets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85\n",
      "D.10 Model Card . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92\n",
      "2\n",
      "1 Introduction\n",
      "Language modeling has long been an important research area since Shannon (1951) estimated the information in\n",
      "language with next word prediction. Modeling began with n-gram based approaches (Kneser & Ney, 1995) but rapidly\n",
      "advanced with LSTMs (Hochreiter & Schmidhuber, 1997; Graves, 2014). Later work showed that language modelling\n",
      "also led to language understanding (Dai & Le, 2015). With increased scale and the Transformer architecture (Vaswani\n",
      "et al., 2017), large language models (LLMs) have shown strong performance in language understanding and generation\n",
      "capabilities over the last few years, leading to breakthrough performance in reasoning, math, science, and language tasks\n",
      "(Howard & Ruder, 2\n",
      "...\n",
      "the platform offer to facilitate communication with annotators? How did this\n",
      "channel of communication influence the annotation process and/or resulting annotations? Annotators could request\n",
      "clarifications from the task creators, and discuss sample annotations that were found to be annotated incorrectly on\n",
      "inspection.\n",
      "How much were annotators compensated? Were any particular pay standards used to determine compensation? If so,\n",
      "please describe. The supplier was paid an average of $42 per hour, with different languages incurring different costs.\n",
      "Google does not have visibility into actual annotator pay, as the supplier sets the rate according to annotator location\n",
      "and experience.\n",
      "Dataset Analysis and Evaluation\n",
      "How was annotation and annotator quality defined, and how did you assess it in your dataset? Standard vendor\n",
      "guidelines for quality checks were used to monitor annotation quality, with an additional linguist to resolve annotator\n",
      "disagreements. In addition, the research team reviewed ratings for languages with outlier aggregate results.\n",
      "Was any analysis of annotation disagreement patterns conducted? If so, what were the high-level findings? No systematic\n",
      "analysis of disagreement patterns were done, except to resolve high level misunderstandings of the task. In those cases,\n",
      "the task guidelines were updated and the same was communicated to the annotators.\n",
      "90\n",
      "Was any analysis of potential sources of disagreement conducted? No\n",
      "How do the individual annotator responses relate to the final labels released in the dataset? N/A\n",
      "Dataset Release and Maintenance Is there reason to believe the annotation judgments in this dataset may lose\n",
      "validity over time? If so, are there plans to update the dataset? Perceptions of toxic language will likely change over\n",
      "time along with changing language or terminology and broader social views of acceptable language use. No current\n",
      "plans to update the dataset.\n",
      "Are there any conditions or definitions that, if changed, could impact the utility of the dataset? Over time societal views\n",
      "may deviate from those captured in the annotations collected. For example, annotations reflect social views at the time\n",
      "of data collection and will not capture shifts in attitudes regarding acceptable language.\n",
      "Are there any plans to track, impose limitations on, or otherwise influence how the dataset is used? If so, how? No\n",
      "Were annotators informed about how the data is externalized? If changes to the dataset are made, will they be informed?\n",
      "No\n",
      "Is there a process by which annotators can later choose to withdraw their data from the dataset? Please detail. No\n",
      "91\n",
      "D.10 Model Card\n",
      "We present the PaLM 2 model card Mitchell et al. (2019) as a starting point for navigating this paper.\n",
      "Table 31: PaLM 2 Model Card\n",
      "Model Summary\n",
      "Model Architecture PaLM-2 is a new state-of-the-art language model. We have small, medium,\n",
      "and large variants that use stacked layers based on the Transformer archi-\n",
      "tecture, with varying parameters depending on model size. Further details\n",
      "of model size and architecture are withheld from external publication.\n",
      "Input(s) The model takes text as input.\n",
      "Output(s) The model generates text as output.\n",
      "Usage\n",
      "Application PaLM 2 is designed for accelerating research on language models, for\n",
      "use as a building block in features within Google products, and as a\n",
      "building block for select experimental applications such as Bard and Magi.\n",
      "Services and products built on top of PaLM 2 are also being made available\n",
      "to external developers via MakerSuite and APIs, with additional process\n",
      "and technical safeguards related to safety policies.\n",
      "We recommend reviewing guides and tools for responsible development,\n",
      "and that all downstream developers consider the potential for harms and\n",
      "bias in the specific context of their application Shelby et al. (2023), particu-\n",
      "larly since changes in decoding strategy and prompts can have a significant\n",
      "impact on generated responses. While we note the effectiveness of dialog-\n",
      "prompting in reducing toxic generation, we stress that these results may not\n",
      "generalize to other forms of representational harm, or to other prompting\n",
      "methods or use contexts. We encourage application developers to consider\n",
      "recent work demonstrating the promise of new methods for more precisely\n",
      "measuring application-specific harms (Mozes et al., 2023).\n",
      "Known Caveats PaLM 2 should not be made available as part of a general-purpose service\n",
      "or product, or used within a specific downstream application without a\n",
      "prior assessment and mitigation of the safety and fairness concerns specific\n",
      "to the downstream use.\n",
      "System Type\n",
      "System Description PaLM 2 is a language model.\n",
      "Upstream Dependencies Section 3 and Appendix D.1\n",
      "Downstream Dependencies See Section D.2 for discussion on anticipated priority uses.\n",
      "Implementation Frameworks\n",
      "Hardware & Software Hardware: TPU v4 (Jouppi et al., 2020).\n",
      "Software: Pathways (Barham et al., 2022), Paxml (aka Pax) (Pax, 2022),\n",
      "Saxml (aka Sax) (Sax, 2022), JAX (Bradbury et al., 2018), GSPMD (Xu\n",
      "et al., 2021b).\n",
      "Compute Requirements Not reported.\n",
      "Model Characteristics\n",
      "92\n",
      "Model Initialization The model is trained from a random initialization.\n",
      "Model Status This is a static model trained on an offline dataset.\n",
      "Model Stats Not reported.\n",
      "Data Overview\n",
      "Training Dataset See Section 3 and Appendix D.1\n",
      "Evaluation Dataset See Section 4 and Appendix D for details.\n",
      "Fine-tuning Dataset See Appendix A.2.\n",
      "Evaluation Results\n",
      "Benchmark Information See Section 4 and Appendix D.2.\n",
      "Evaluation Results See Section 4 and Appendix D.\n",
      "Model Usage & Limitations\n",
      "Sensitive Use PaLM 2 is capable of open-ended text generation. This model should not\n",
      "be used to cause harm.\n",
      "Known Limitations PaLM 2 is not testing in this paper in settings outside of research that\n",
      "can affect performance, and it should not be used for downstream appli-\n",
      "cations without further analysis of potential harm or bias in the proposed\n",
      "downstream application.\n",
      "Ethical Considerations & Risks See Chowdhery et al. (2022) and Section 5.\n",
      "93\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(download_and_read_pdf.invoke({\"url\": \"https://arxiv.org/pdf/2305.10403.pdf\"}))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "94065b98",
   "metadata": {},
   "source": [
    "## Build graph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "19529684",
   "metadata": {},
   "outputs": [],
   "source": [
    "class MessagesStateWithRetry(MessagesState):\n",
    "    \"\"\"Extended state that includes retry counter to prevent infinite loops.\"\"\"\n",
    "\n",
    "    retry_count: int = 0\n",
    "    max_retries: int = 3  # Maximum number of review retry attempts\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "bdd5256e",
   "metadata": {},
   "outputs": [],
   "source": [
    "tools = [\n",
    "    calculator,\n",
    "    wikipedia_search,\n",
    "    wikipedia_section,\n",
    "    arxiv_search,\n",
    "    web_search,\n",
    "    fetch_webpage,\n",
    "    read_excel_file,\n",
    "    get_youtube_transcript,\n",
    "    execute_python_file,\n",
    "    execute_python_code,\n",
    "    transcript_audio_file,\n",
    "    describe_image_file,\n",
    "    download_and_read_pdf,\n",
    "]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "df4afab0",
   "metadata": {},
   "outputs": [],
   "source": [
    "SYSTEM_PROMPT = \"\"\"\n",
    "You are a GAIA benchmark problem-solving assistant specializing in multi-step reasoning tasks.\n",
    "\n",
    "CRITICAL FIRST STEPS:\n",
    "\n",
    "1. **File Attachment Check**: If the question mentions or includes an attached file,\n",
    "   IMMEDIATELY inspect it using the appropriate tool BEFORE any other actions.\n",
    "   - Excel/CSV: Use read_excel_file or appropriate reader\n",
    "   - Image: Describe what you need to analyze\n",
    "   - Always verify the file path and extension\n",
    "\n",
    "2. **Question Analysis**: Identify:\n",
    "   - Required precision (decimal places, units, format)\n",
    "   - Temporal constraints (\"as of DATE\", \"current\", \"latest\")\n",
    "   - Inclusion/exclusion criteria (be explicit about what counts)\n",
    "   - Expected answer type (number, string, list, date)\n",
    "\n",
    "TOOL SELECTION STRATEGY:\n",
    "\n",
    "- **Wikipedia tasks**:\n",
    "  - Use `wikipedia_search` to discover relevant pages and sections\n",
    "  - Use `wikipedia_section` to retrieve detailed content\n",
    "  - NEVER use for historical \"as of [past date]\" questions - Wikipedia shows current content\n",
    "\n",
    "- **Sports questions** (player stats, rosters, uniform numbers, team info):\n",
    "  - Be specific: \"Taishō Tamai uniform number Hokkaido Nippon-Ham Fighters July 2023\"\n",
    "  - The tool automatically fetches and parses pages - look for patterns like \"#18\" or \"No. 18\"\n",
    "  - Try these strategies:\n",
    "    * Try `web_search` with different query phrasing\n",
    "    * Use `fetch_webpage` on specific URLs from sports_statistics results\n",
    "    * Look for official team roster pages or player profile pages\n",
    "  - Be persistent: try 2-3 different searches before giving up\n",
    "  - For player bios without numbers, `wikipedia_search` may work\n",
    "\n",
    "- **Calculations**:\n",
    "  - ALWAYS use `calculator` for arithmetic - never calculate manually\n",
    "  - Break complex calculations into verified steps\n",
    "  - State your calculation expression clearly before executing\n",
    "\n",
    "- **Academic Paper Research**:\n",
    "    - NEVER guess or invent URLs, DOI values, arXiv IDs, paper titles,\n",
    "        award numbers, grant numbers, dates, or author names.\n",
    "        1. Use `fetch_webpage` to inspect the webpage.\n",
    "        2. Extract the exact paper title and, when available, its actual URL.\n",
    "        3. If the URL is unavailable, call `arxiv_search` using the exact\n",
    "            paper title in quotation marks.\n",
    "        4. Verify that the title and authors match before continuing.\n",
    "        5. Obtain the canonical PDF URL from the search result.\n",
    "        6. Use `download_and_read_pdf` to inspect the paper.\n",
    "        7. For funding questions, search the complete PDF for:\n",
    "            - acknowledgments\n",
    "            - acknowledgement\n",
    "            - supported\n",
    "            - funding\n",
    "            - grant\n",
    "            - award\n",
    "            - NASA\n",
    "            - the author's full name and initials\n",
    "        8. Answer only when the requested value appears in the paper itself.\n",
    "    - NEVER construct an arXiv URL from a guessed identifier.\n",
    "    - NEVER use `fetch_webpage` to retrieve arXiv content.\n",
    "    - Use `arxiv_search` for arXiv discovery and\n",
    "    `download_and_read_pdf` for the paper content.\n",
    "    - A failed URL is not evidence that the paper cannot be found.\n",
    "    Retry using the exact title, authors, and distinctive title phrases.\n",
    "\n",
    "\n",
    "- **Web Research**:\n",
    "  - Use `web_search` for current events, recent data, or when Wikipedia is insufficient\n",
    "  - Use `arxiv_search` for academic papers only\n",
    "  - Use `download_and_read_pdf` for PDF content retrieval\n",
    "  - Use `fetch_webpage` for specific URLs (but NOT for Wikipedia or arXiv)\n",
    "\n",
    "- **YouTube Videos**:\n",
    "    - Use `get_youtube_transcript` to retrieve transcripts\n",
    "    - If transciption is incomplete, try to look information about the video on the web with `web_search` or `fetch_webpage`.\n",
    "\n",
    "- **LibreText**:\n",
    "    - Use `web_search` to find relevant LibreText query the official `https://chem.libretexts.org` + the title of the section, author names, license, compilation date, and distinctive phrases.\n",
    "    - When found the correct website, use `fetch_webpage` to retrieve the content of the LibreText page, and then parse it for relevant information.\n",
    "\n",
    "- Execute Python code to solve problems or analyze data when necessary. Use `execute_python_code` for snippets and `execute_python_file` for full scripts.\n",
    "\n",
    "VERIFICATION PROCESS:\n",
    "\n",
    "1. After retrieving data, explicitly state what you found\n",
    "2. Before calculating, list all values you'll use\n",
    "3. Verify your calculation includes all required categories and excludes forbidden ones\n",
    "4. Check your answer matches the requested format and precision\n",
    "5. You can check for a list of vegetables on the wiki page \"List of vegetables\" to verify if a vegetable is included in the list.\n",
    "6. If you are unsure about a food or drink item, you can verify it on the wiki page like \"Ice cream\".\n",
    "\n",
    "HANDLING AMBIGUITY:\n",
    "\n",
    "- If data is unclear, state your interpretation explicitly\n",
    "- If multiple interpretations exist, mention them\n",
    "- If data is missing, try alternative sources before giving up\n",
    "\n",
    "MANDATORY RELIABILITY RULES:\n",
    "\n",
    "- Never guess URLs, arXiv IDs, DOI values, award numbers, dates,\n",
    "  uniform numbers, names, or other exact identifiers.\n",
    "\n",
    "- For finite operation tables, commutativity, associativity, or exhaustive\n",
    "  comparisons, MUST use `execute_python_code`. The row is the first operand\n",
    "  and the column is the second operand. Check every relevant pair, collect\n",
    "  all qualifying elements, remove duplicates, and sort as requested.\n",
    "\n",
    "- For academic-paper questions:\n",
    "  1. Locate the referenced webpage with `web_search`.\n",
    "  2. Inspect it with `fetch_webpage`.\n",
    "  3. Extract the exact paper title or verified link.\n",
    "  4. Search the exact title with `arxiv_search`; never guess an arXiv ID.\n",
    "  5. Verify title and authors.\n",
    "  6. Use `download_and_read_pdf` for acknowledgments, grants, funding,\n",
    "     awards, or facts found inside the paper.\n",
    "\n",
    "- For LibreTexts questions, search the official domain using exact section\n",
    "  titles, author names, license, compilation date, and distinctive phrases.\n",
    "  Open the exact dated material and inspect the requested section. Do not\n",
    "  answer only from search snippets.\n",
    "\n",
    "- For historical sports questions, treat the requested date as mandatory.\n",
    "  Verify the player's number on a dated roster. For “number before and\n",
    "  after”, find N-1 and N+1, verify both players have the requested position,\n",
    "  and format their surnames exactly as requested.\n",
    "\n",
    "- If a result is irrelevant or incomplete, refine the search at least once\n",
    "  using exact titles, phrases, authors, dates, or official sources.\n",
    "\n",
    "FINAL ANSWER FORMAT:\n",
    "\n",
    "Finish with exactly: FINAL ANSWER: [answer]\n",
    "\n",
    "CRITICAL FORMAT RULES:\n",
    "\n",
    "1. **Numbers - NEVER include symbols or unit text**:\n",
    "   - No thousands separators (not 89,706 but 89706)\n",
    "   - No currency symbols (not $89706 but 89706)\n",
    "   - No unit abbreviations (not 89706 USD but just 89706.00)\n",
    "   - No percentage signs (not 50% but just 50)\n",
    "\n",
    "2. **Understanding unit phrases**:\n",
    "   - \"Express in USD\" = use USD as the measurement scale (DO NOT append \"USD\")\n",
    "   - \"Express in dollars\" = same as above (DO NOT append \"dollars\")\n",
    "   - \"in kilometers\" = use km scale (DO NOT append \"km\")\n",
    "   - \"as a percentage\" = calculate percentage value (DO NOT append \"%\")\n",
    "\n",
    "3. **When TO append unit text** (RARE):\n",
    "   - ONLY when explicitly asked: \"include the text USD after the number\"\n",
    "   - ONLY when explicitly asked: \"append the currency code\"\n",
    "   - Otherwise, NEVER append unit text\n",
    "\n",
    "4. **Decimal precision**:\n",
    "   - Match EXACTLY the requested decimal places\n",
    "   - \"two decimal places\" → 89706.00 (not 89706 or 89706.0)\n",
    "   - If no precision specified, use reasonable precision for the context\n",
    "\n",
    "5. **String answers**:\n",
    "   - Use minimal words, be direct\n",
    "   - No extra explanation after FINAL ANSWER\n",
    "\n",
    "6. **List answers**:\n",
    "   - Comma-separated values\n",
    "   - No extra commentary\n",
    "\n",
    "CORRECT EXAMPLES:\n",
    "- \"Express your answer in USD with two decimal places\" → FINAL ANSWER: 89706.00\n",
    "- \"Give the answer in kilometers\" → FINAL ANSWER: 42.5\n",
    "- \"What percentage?\" → FINAL ANSWER: 23.4\n",
    "- \"Include USD after the number\" → FINAL ANSWER: 89706.00 USD\n",
    "\n",
    "INCORRECT EXAMPLES (DO NOT DO THIS):\n",
    "- FINAL ANSWER: 89706.00 USD  ← WRONG if question says \"in USD\" (should be 89706.00)\n",
    "- FINAL ANSWER: $89706.00  ← WRONG (never use $ symbol)\n",
    "- FINAL ANSWER: 89,706.00  ← WRONG (no thousands separators)\n",
    "- FINAL ANSWER: 23.4%  ← WRONG if question says \"as a percentage\" (should be 23.4)\n",
    "\"\"\"\n",
    "\n",
    "FINAL_REVIEW_SYSTEM_PROMPT = \"\"\"\n",
    "You are a final-answer verification specialist for GAIA benchmark tasks.\n",
    "\n",
    "You receive:\n",
    "- <original_request>: The user's question\n",
    "- <current_attempt>: Tool calls and results from this attempt\n",
    "- <candidate_answer>: The proposed final answer\n",
    "\n",
    "Your task: Verify the answer is correct based ONLY on the current attempt's evidence.\n",
    "\n",
    "IMPORTANT: Be lenient and pragmatic. Only flag critical errors that make the answer wrong.\n",
    "Minor format issues are acceptable if the core answer value is correct.\n",
    "\n",
    "VALIDATION PRIORITIES (in order):\n",
    "\n",
    "1. **Answer Value Correctness** (MOST IMPORTANT):\n",
    "   - Does the numeric/text value match the tool evidence?\n",
    "   - Is the calculation result correct according to the calculator output?\n",
    "   - If the value is right, proceed to next checks\n",
    "\n",
    "2. **Data Selection**:\n",
    "   - Were the required data points used?\n",
    "   - Only flag if CRITICAL data was missed or wrong data was included\n",
    "   - Don't flag if alternative valid approaches were used\n",
    "\n",
    "3. **Tool Appropriateness**:\n",
    "   - Only flag if the WRONG tool was used (e.g., Wikipedia for historical data)\n",
    "   - Trust calculator tool results - don't second-guess the math\n",
    "\n",
    "4. **Format Compliance** (CHECK CAREFULLY):\n",
    "   - NO currency symbols ($, €, £)\n",
    "   - NO unit text UNLESS explicitly requested (\"include the text USD\")\n",
    "   - \"Express in USD\" means the value in USD scale, NOT appending \"USD\" text\n",
    "   - NO thousands separators\n",
    "   - Correct decimal precision as requested\n",
    "   - BUT: if answer value is correct and only has minor format issues, be lenient\n",
    "\n",
    "WHEN TO ACCEPT AN ANSWER:\n",
    "\n",
    "Accept if ANY of these are true:\n",
    "- The answer value is correct and format is reasonable (even if not perfect)\n",
    "- The answer is mathematically correct per the calculator tool\n",
    "- Minor format issues exist but the core answer is right and interpretable\n",
    "\n",
    "WHEN TO REJECT AN ANSWER:\n",
    "\n",
    "Only reject if:\n",
    "- The numeric/text value is factually wrong\n",
    "- Critical required data was not used in the calculation\n",
    "- The wrong calculation was performed\n",
    "- Major format violation (e.g., has \"USD\" text when question said \"in USD\")\n",
    "- Wrong decimal precision (e.g., 0 decimals when 2 were requested)\n",
    "\n",
    "DO NOT REJECT FOR:\n",
    "- Minor decimal precision differences if reasonable\n",
    "- Extra/missing trailing zeros (unless explicitly required)\n",
    "- Different but equally valid interpretation of ambiguous questions\n",
    "\n",
    "COMMON FORMAT ERRORS TO CATCH:\n",
    "- \"Express in USD\" → answer has \"USD\" appended (REJECT - should be just the number)\n",
    "- \"in kilometers\" → answer has \"km\" appended (REJECT - should be just the number)\n",
    "- Answer has $ symbol (REJECT - never allowed)\n",
    "- Answer has commas (REJECT - no thousands separators)\n",
    "\n",
    "TRUST THE TOOLS:\n",
    "- If calculator was used, trust its output - don't recalculate manually\n",
    "- If the right data went into calculator, and result was transcribed correctly, accept it\n",
    "\n",
    "OUTPUT REQUIREMENTS:\n",
    "\n",
    "Return structured ReviewResult with:\n",
    "- `data_selection_correct`: Were the right data points used? (Be lenient)\n",
    "- `answer_correct`: Is the core value correct per the evidence? (Most important)\n",
    "- `format_correct`: Is formatting correct per rules? (Check unit text carefully)\n",
    "- `errors`: ONLY list critical errors that make the answer wrong\n",
    "- `feedback`: If rejecting, provide specific correction (don't reject for trivial issues)\n",
    "\n",
    "When in doubt about VALUE correctness, ACCEPT. But enforce format rules for units/symbols.\n",
    "\"\"\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "50435ee7",
   "metadata": {},
   "outputs": [],
   "source": [
    "class ReviewResult(BaseModel):\n",
    "    data_selection_correct: bool = Field(\n",
    "        description=(\n",
    "            \"Whether all required rows, columns, categories, and values \"\n",
    "            \"were included and excluded correctly.\"\n",
    "        )\n",
    "    )\n",
    "\n",
    "    answer_correct: bool = Field(\n",
    "        description=(\n",
    "            \"Whether the candidate answer has the correct value according \"\n",
    "            \"to the raw tool evidence.\"\n",
    "        )\n",
    "    )\n",
    "\n",
    "    format_correct: bool = Field(\n",
    "        description=(\"Whether the candidate follows the requested final-answer format.\")\n",
    "    )\n",
    "\n",
    "    errors: list[str] = Field(\n",
    "        description=(\n",
    "            \"All material errors. Do not mention missing explanations unless \"\n",
    "            \"the original user explicitly requested an explanation.\"\n",
    "        )\n",
    "    )\n",
    "\n",
    "    feedback: str = Field(\n",
    "        description=(\n",
    "            \"Actionable correction instructions for the assistant. \"\n",
    "            \"Empty when everything is correct.\"\n",
    "        )\n",
    "    )\n",
    "\n",
    "\n",
    "rate_limiter = InMemoryRateLimiter(\n",
    "    requests_per_second=1 / 30,\n",
    "    check_every_n_seconds=0.2,\n",
    "    max_bucket_size=1,\n",
    ")\n",
    "\n",
    "llm = ChatOpenAI(\n",
    "    model=\"gpt-4o\",\n",
    "    temperature=0,\n",
    "    rate_limiter=rate_limiter,\n",
    "    max_retries=5,\n",
    "    timeout=120,\n",
    ")\n",
    "\n",
    "llm_with_tools = llm.bind_tools(tools, parallel_tool_calls=False)\n",
    "\n",
    "review_model = llm.with_structured_output(ReviewResult)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "a24c5ad5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<langgraph.graph.state.StateGraph at 0x7fd363200410>"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def assistant(state: MessagesStateWithRetry):\n",
    "    return {\n",
    "        \"messages\": [\n",
    "            llm_with_tools.invoke([SystemMessage(SYSTEM_PROMPT)] + state[\"messages\"])\n",
    "        ]\n",
    "    }\n",
    "\n",
    "\n",
    "def build_review_input(state: MessagesStateWithRetry) -> str:\n",
    "    messages = state[\"messages\"]\n",
    "\n",
    "    original_request = messages[0].content\n",
    "\n",
    "    for candidate_index in range(len(messages) - 1, -1, -1):\n",
    "        if (\n",
    "            isinstance(messages[candidate_index], AIMessage)\n",
    "            and messages[candidate_index].content\n",
    "            and not messages[candidate_index].tool_calls\n",
    "        ):\n",
    "            break\n",
    "\n",
    "    candidate_answer = messages[candidate_index].content\n",
    "\n",
    "    # Last feedback index\n",
    "    last_review_index = next(\n",
    "        (\n",
    "            index\n",
    "            for index in range(candidate_index - 1, -1, -1)\n",
    "            if (\n",
    "                isinstance(messages[index], HumanMessage)\n",
    "                and str(messages[index].content).startswith(\n",
    "                    \"FINAL ANSWER REVIEW FAILED\"\n",
    "                )\n",
    "            )\n",
    "        ),\n",
    "        -1,\n",
    "    )\n",
    "\n",
    "    attempt_start = last_review_index + 1\n",
    "\n",
    "    current_attempt = []\n",
    "\n",
    "    for message in messages[attempt_start:candidate_index]:\n",
    "        if isinstance(message, AIMessage):\n",
    "            for t in message.tool_calls or []:\n",
    "                current_attempt.append(\n",
    "                    f\"<tool_call>\\nname: {t['name']}\\narguments: {t.get('args', {})}\\n</tool_call>\"\n",
    "                )\n",
    "\n",
    "        elif isinstance(message, ToolMessage):\n",
    "            current_attempt.append(\n",
    "                f\"<tool_result>\\n\"\n",
    "                f\"name: {message.name}\\n\"\n",
    "                f\"content:\\n{message.content}\\n\"\n",
    "                f\"</tool_result>\"\n",
    "            )\n",
    "\n",
    "    return f\"\"\"\n",
    "<original_request>\n",
    "{original_request}\n",
    "</original_request>\n",
    "\n",
    "<current_attempt>\n",
    "{\"\\n\\n---\\n\\n\".join(current_attempt)}\n",
    "</current_attempt>\n",
    "\n",
    "<candidate_answer>\n",
    "{candidate_answer}\n",
    "</candidate_answer>\n",
    "\"\"\".strip()\n",
    "\n",
    "\n",
    "def review_passed(review: ReviewResult) -> bool:\n",
    "    \"\"\"Check if review passed with lenient criteria.\n",
    "\n",
    "    Accept if:\n",
    "    - Answer is correct (most important)\n",
    "    - No critical errors\n",
    "    - Format is reasonable (even if not perfect)\n",
    "    \"\"\"\n",
    "    # Most important: is the answer value correct?\n",
    "    if not review.answer_correct:\n",
    "        return False\n",
    "\n",
    "    # If answer is correct but has minor format issues, still accept\n",
    "    if review.format_correct and review.data_selection_correct:\n",
    "        return True\n",
    "\n",
    "    # If answer is correct but format/data selection slightly off,\n",
    "    # only reject if there are actual critical errors listed\n",
    "    if not review.errors:\n",
    "        # No errors listed means issues are minor, accept it\n",
    "        return True\n",
    "\n",
    "    # Check if errors are critical\n",
    "    critical_keywords = [\n",
    "        \"wrong value\",\n",
    "        \"incorrect calculation\",\n",
    "        \"missing required\",\n",
    "        \"excluded incorrectly\",\n",
    "    ]\n",
    "    has_critical_error = any(\n",
    "        any(keyword in error.lower() for keyword in critical_keywords)\n",
    "        for error in review.errors\n",
    "    )\n",
    "\n",
    "    # If no critical errors and answer is correct, accept\n",
    "    return not has_critical_error\n",
    "\n",
    "\n",
    "def final_review(\n",
    "    state: MessagesStateWithRetry,\n",
    ") -> Command[Literal[\"assistant\", \"__end__\"]]:\n",
    "    review = review_model.invoke(\n",
    "        [\n",
    "            SystemMessage(content=FINAL_REVIEW_SYSTEM_PROMPT),\n",
    "            HumanMessage(build_review_input(state)),\n",
    "        ]\n",
    "    )\n",
    "\n",
    "    if review_passed(review):\n",
    "        # Last assistant message is the final answer, so we can end the conversation\n",
    "        return Command(goto=END)\n",
    "\n",
    "    # Check if we've exceeded max retries\n",
    "    current_retry = state.get(\"retry_count\", 0)\n",
    "    max_retries = state.get(\"max_retries\", 3)\n",
    "\n",
    "    if current_retry >= max_retries:\n",
    "        print(\n",
    "            f\"Max retries ({max_retries}) reached. Accepting current answer despite errors.\"\n",
    "        )\n",
    "        # End the conversation with the current answer even if it has errors\n",
    "        return Command(goto=END)\n",
    "\n",
    "    errors = \"\\n\".join(f\"- {error}\" for error in review.errors)\n",
    "\n",
    "    reviewer_feedback = HumanMessage(\n",
    "        content=(\n",
    "            f\"FINAL ANSWER REVIEW FAILED (Attempt {current_retry + 1}/{max_retries}).\\n\\n\"\n",
    "            f\"Errors found:\\n{errors}\\n\\n\"\n",
    "            f\"Correction instructions:\\n{review.feedback}\\n\\n\"\n",
    "            \"Review the raw source data and only the current attempt. \"\n",
    "            \"Include all required categories, exclude all disallowed \"\n",
    "            \"categories, recalculate when necessary, and finish with the \"\n",
    "            \"required FINAL ANSWER format.\"\n",
    "        )\n",
    "    )\n",
    "\n",
    "    return Command(\n",
    "        update={\n",
    "            \"messages\": [reviewer_feedback],\n",
    "            \"retry_count\": current_retry + 1,\n",
    "        },\n",
    "        goto=\"assistant\",\n",
    "    )\n",
    "\n",
    "\n",
    "def route_after_assistant(\n",
    "    state: MessagesStateWithRetry,\n",
    ") -> Literal[\"tools\", \"final_review\"]:\n",
    "    last_message = state[\"messages\"][-1]\n",
    "\n",
    "    if isinstance(last_message, AIMessage) and last_message.tool_calls:\n",
    "        return \"tools\"\n",
    "\n",
    "    return \"final_review\"\n",
    "\n",
    "\n",
    "builder = StateGraph(MessagesStateWithRetry)\n",
    "\n",
    "# Nodes\n",
    "builder.add_node(\"assistant\", assistant)\n",
    "builder.add_node(\"tools\", ToolNode(tools))\n",
    "builder.add_node(\"final_review\", final_review)\n",
    "\n",
    "\n",
    "# Edges\n",
    "builder.add_edge(START, \"assistant\")\n",
    "builder.add_conditional_edges(\"assistant\", route_after_assistant)\n",
    "builder.add_edge(\"tools\", \"assistant\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e0bbedf6",
   "metadata": {},
   "source": [
    "## Basic agent"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "3de05c44",
   "metadata": {},
   "outputs": [],
   "source": [
    "class BasicAgent:\n",
    "    def __init__(self, max_retries: int = 3):\n",
    "        # Compile\n",
    "        self.graph = builder.compile()\n",
    "        self.max_retries = max_retries\n",
    "\n",
    "    def __call__(self, question: str) -> str:\n",
    "        messages = self.graph.invoke(\n",
    "            {\"messages\": question, \"retry_count\": 0, \"max_retries\": self.max_retries}\n",
    "        )\n",
    "        m = re.search(\n",
    "            r\".*FINAL ANSWER: (?P<answer>.*)\", messages[\"messages\"][-1].content\n",
    "        )\n",
    "        if m is None:\n",
    "            return {\"messages\": messages[\"messages\"], \"answer\": \"Unresolved\"}\n",
    "\n",
    "        final_answer = m.group(\"answer\")\n",
    "        return {\"messages\": messages[\"messages\"], \"answer\": final_answer}\n",
    "\n",
    "    def draw_mermaid(self):\n",
    "        return self.graph.get_graph(xray=True).draw_mermaid_png()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "12b94c36",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "agent = BasicAgent()\n",
    "\n",
    "display(Image(agent.draw_mermaid()))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "94bd1df2",
   "metadata": {},
   "source": [
    "## Tests questions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "2362b398",
   "metadata": {},
   "outputs": [],
   "source": [
    "try:\n",
    "    response = requests.get(questions_url)\n",
    "    response.raise_for_status()\n",
    "    questions = response.json()\n",
    "except requests.exceptions.HTTPError as e:\n",
    "    print(f\"Server responded with status {e.response.status_code}.\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "d80dfaa1",
   "metadata": {},
   "outputs": [],
   "source": [
    "def question_query(question: dict) -> str:\n",
    "    task_id = question.get(\"task_id\")\n",
    "    question_query = question.get(\"question\")\n",
    "    if fname := question.get(\"file_name\"):\n",
    "        attached = hf_hub_download(\n",
    "            repo_id=\"gaia-benchmark/GAIA\",\n",
    "            filename=f\"2023/validation/{fname}\",\n",
    "            repo_type=\"dataset\",\n",
    "            token=os.environ.get(\"HF_TOKEN\"),\n",
    "        )\n",
    "        extension = Path(attached).suffix\n",
    "        question_query += f\"\"\"\\n\\nAttached file extensions: {extension}\n",
    "\n",
    "Attached file path: {attached}\"\"\"\n",
    "    return task_id, HumanMessage(question_query)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "d7b9861b",
   "metadata": {},
   "outputs": [],
   "source": [
    "def pretty_print_response(task_id: str, response: dict) -> dict:\n",
    "    nodes = []\n",
    "    for m in response[\"messages\"]:\n",
    "        node = \"\"\n",
    "        if isinstance(m, HumanMessage):\n",
    "            node += f\"HumanMessage({m.content[:20]}...)\"\n",
    "        if isinstance(m, AIMessage):\n",
    "            node += \"AIMessage\"\n",
    "            if m.tool_calls:\n",
    "                node += \"(\"\n",
    "                for t in m.tool_calls:\n",
    "                    node += f\"{t['type']}({t['name']}({t.get('args', {})})),\"\n",
    "                node += \")\"\n",
    "            else:\n",
    "                node += f\"({m.content[:20]}...)\"\n",
    "        if isinstance(m, ToolMessage):\n",
    "            node += f\"ToolMessage({m.name}({m.content[:20]}...))\"\n",
    "        nodes.append(node)\n",
    "\n",
    "    return {\n",
    "        \"task_id\": task_id,\n",
    "        \"submitted_answer\": response[\"answer\"],\n",
    "        \"reasoning_trace\": \" -> \".join(n for n in nodes),\n",
    "    }\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "203b0e47",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0] {'task_id': '8e867cd7-cff9-4e6c-867a-ff5ddc2550be', 'submitted_answer': '3', 'reasoning_trace': \"HumanMessage(How many studio albu...) -> AIMessage(tool_call(wikipedia_section({'title': 'Mercedes Sosa', 'section': 'Discography'})),) -> ToolMessage(wikipedia_section(# Mercedes Sosa\\n**UR...)) -> AIMessage(Between 2000 and 200...)\"}\n",
      "[1] {'task_id': 'a1e91b78-d3d8-4675-bb8d-62741b4b68a6', 'submitted_answer': '3', 'reasoning_trace': \"HumanMessage(In the video https:/...) -> AIMessage(tool_call(get_youtube_transcript({'video_url': 'https://www.youtube.com/watch?v=L1vXCYZAYYM'})),) -> ToolMessage(get_youtube_transcript(An error occurred: \\n...)) -> AIMessage(tool_call(web_search({'query': 'highest number of bird species on camera simultaneously in video L1vXCYZAYYM'})),) -> ToolMessage(web_search(# gaia-benchmark/GAI...)) -> AIMessage(The highest number o...)\"}\n",
      "[2] {'task_id': '2d83110e-a098-4ebb-9987-066c06fa42d0', 'submitted_answer': 'right', 'reasoning_trace': 'HumanMessage(.rewsna eht sa \"tfel...) -> AIMessage(If you understand th...)'}\n",
      "[3] {'task_id': 'cca530fc-4052-43b2-b130-b30968d8aa44', 'submitted_answer': 'Qd1+', 'reasoning_trace': 'HumanMessage(Review the chess pos...) -> AIMessage(tool_call(describe_image_file({\\'file_path\\': \\'/home/cora/.cache/huggingface/hub/datasets--gaia-benchmark--GAIA/snapshots/682dd723ee1e1697e00360edccf2366dc8418dd9/2023/validation/cca530fc-4052-43b2-b130-b30968d8aa44.png\\'})),) -> ToolMessage(describe_image_file(content=\"The image s...)) -> AIMessage(Based on the describ...)'}\n",
      "[4] {'task_id': '4fc2f1ae-8625-45b5-ab34-ad4433bc21f8', 'submitted_answer': 'FunkMonk', 'reasoning_trace': \"HumanMessage(Who nominated the on...) -> AIMessage(tool_call(wikipedia_search({'query': 'Featured Article dinosaur November 2016'})),) -> ToolMessage(wikipedia_search(# Dinosaur\\n**URL:** ...)) -> AIMessage(tool_call(wikipedia_section({'title': 'Wikipedia:Featured_articles', 'section': 'Promotions'})),) -> ToolMessage(wikipedia_section(Section 'Promotions'...)) -> AIMessage(tool_call(wikipedia_section({'title': 'Wikipedia:Featured_articles', 'section': None})),) -> ToolMessage(wikipedia_section(# Wikipedia:Featured...)) -> AIMessage(tool_call(wikipedia_section({'title': 'Wikipedia:Featured_articles', 'section': 'Biology'})),) -> ToolMessage(wikipedia_section(# Wikipedia:Featured...)) -> AIMessage(tool_call(wikipedia_section({'title': 'Dromaeosauroides', 'section': None})),) -> ToolMessage(wikipedia_section(# Dromaeosauroides\\n*...)) -> AIMessage(The only Featured Ar...)\"}\n",
      "[5] {'task_id': '6f37996b-2ac7-44b0-8e68-6d28256631b4', 'submitted_answer': 'a, b, c, d, e', 'reasoning_trace': 'HumanMessage(Given this table def...) -> AIMessage(tool_call(execute_python_code({\\'code\\': \"# Define the operation table as a dictionary\\\\noperation_table = {\\\\n    (\\'a\\', \\'a\\'): \\'a\\', (\\'a\\', \\'b\\'): \\'b\\', (\\'a\\', \\'c\\'): \\'c\\', (\\'a\\', \\'d\\'): \\'b\\', (\\'a\\', \\'e\\'): \\'d\\',\\\\n    (\\'b\\', \\'a\\'): \\'b\\', (\\'b\\', \\'b\\'): \\'c\\', (\\'b\\', \\'c\\'): \\'a\\', (\\'b\\', \\'d\\'): \\'e\\', (\\'b\\', \\'e\\'): \\'c\\',\\\\n    (\\'c\\', \\'a\\'): \\'c\\', (\\'c\\', \\'b\\'): \\'a\\', (\\'c\\', \\'c\\'): \\'b\\', (\\'c\\', \\'d\\'): \\'b\\', (\\'c\\', \\'e\\'): \\'a\\',\\\\n    (\\'d\\', \\'a\\'): \\'b\\', (\\'d\\', \\'b\\'): \\'e\\', (\\'d\\', \\'c\\'): \\'b\\', (\\'d\\', \\'d\\'): \\'e\\', (\\'d\\', \\'e\\'): \\'d\\',\\\\n    (\\'e\\', \\'a\\'): \\'d\\', (\\'e\\', \\'b\\'): \\'b\\', (\\'e\\', \\'c\\'): \\'a\\', (\\'e\\', \\'d\\'): \\'d\\', (\\'e\\', \\'e\\'): \\'c\\'\\\\n}\\\\n\\\\n# Define the set S\\\\nS = {\\'a\\', \\'b\\', \\'c\\', \\'d\\', \\'e\\'}\\\\n\\\\n# Find counter-examples for commutativity\\\\ncounter_examples = set()\\\\n\\\\nfor x in S:\\\\n    for y in S:\\\\n        if operation_table[(x, y)] != operation_table[(y, x)]:\\\\n            counter_examples.add(x)\\\\n            counter_examples.add(y)\\\\n\\\\n# Return the sorted list of elements involved in counter-examples\\\\nsorted(counter_examples)\"})),) -> ToolMessage(execute_python_code(...)) -> AIMessage(The subset of \\\\( S \\\\...)'}\n",
      "[6] {'task_id': '9d191bce-651d-4746-be2d-7ef8ecadb9c2', 'submitted_answer': 'extremely', 'reasoning_trace': \"HumanMessage(Examine the video at...) -> AIMessage(tool_call(get_youtube_transcript({'video_url': 'https://www.youtube.com/watch?v=1htKBjuUWec'})),) -> ToolMessage(get_youtube_transcript(# YouTube video tran...)) -> AIMessage(In response to the q...)\"}\n",
      "[7] {'task_id': 'cabe07ed-9eca-40ea-8ead-410ef5e83f91', 'submitted_answer': 'Unresolved', 'reasoning_trace': \"HumanMessage(What is the surname ...) -> AIMessage(tool_call(web_search({'query': 'LibreText Introductory Chemistry 1.E Exercises Marisa Alviar-Agnew Henry Agnew CK-12 license'})),) -> ToolMessage(web_search(# LibreTexts - Wikip...)) -> AIMessage(tool_call(web_search({'query': 'LibreText Introductory Chemistry 1.E Exercises equine veterinarian surname'})),) -> ToolMessage(web_search(# LibreTexts\\n**URL:*...)) -> AIMessage(tool_call(web_search({'query': 'LibreText Introductory Chemistry 1.E Exercises equine veterinarian'})),) -> ToolMessage(web_search(# 1.E: Introduction ...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://chem.libretexts.org/Bookshelves/General_Chemistry/Map%3A_Chemistry_and_Chemical_Reactivity_(Kotz_et_al.)/01%3A_Introduction_to_Chemistry/1.E_Exercises', 'max_chars': 3000})),) -> ToolMessage(fetch_webpage(1.E Exercises - Chem...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://chem.libretexts.org/Courses/can/CHEM_410%3A_Chemistry_for_Health_Science/01%3A_Introduction_to_Chemistry/1.E%3A_Introduction_to_Chemistry_(Exercises)', 'max_chars': 3000})),) -> ToolMessage(fetch_webpage(1.E: Introduction to...)) -> AIMessage(I was unable to find...)\"}\n",
      "[8] {'task_id': '3cef3a44-215e-4aed-8e3b-b1e3f08063b7', 'submitted_answer': 'Bell pepper, Broccoli, Celery, Green beans, Lettuce, Sweet potatoes, Zucchini', 'reasoning_trace': \"HumanMessage(I'm making a grocery...) -> AIMessage(tool_call(wikipedia_section({'title': 'List of vegetables'})),) -> ToolMessage(wikipedia_section(# List of vegetables...)) -> AIMessage(Based on the list of...)\"}\n",
      "[9] {'task_id': '99c9cc74-fdc8-46c6-8f8d-3ce2d3bfeea3', 'submitted_answer': 'Unresolved', 'reasoning_trace': \"HumanMessage(Hi, I'm making a pie...) -> AIMessage(tool_call(transcript_audio_file({'file_path': '/home/cora/.cache/huggingface/hub/datasets--gaia-benchmark--GAIA/snapshots/682dd723ee1e1697e00360edccf2366dc8418dd9/2023/validation/99c9cc74-fdc8-46c6-8f8d-3ce2d3bfeea3.mp3'})),) -> ToolMessage(transcript_audio_file( In a saucepan, comb...)) -> AIMessage(The ingredients for ...)\"}\n",
      "[10] {'task_id': '305ac316-eef6-4446-960a-92d80d542f82', 'submitted_answer': 'Wojciech', 'reasoning_trace': \"HumanMessage(Who did the actor wh...) -> AIMessage(tool_call(wikipedia_search({'query': 'Polish-language version of Everybody Loves Raymond cast'})),) -> ToolMessage(wikipedia_search(# Everybody Loves Ra...)) -> AIMessage(tool_call(wikipedia_section({'title': 'Everybody Loves Raymond', 'section': 'Foreign remakes'})),) -> ToolMessage(wikipedia_section(# Everybody Loves Ra...)) -> AIMessage(tool_call(web_search({'query': 'Wszyscy kochają Romana cast'})),) -> ToolMessage(web_search(# Wszyscy kochają Ro...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://en.wikipedia.org/wiki/Wszyscy_kochaj%C4%85_Romana'})),) -> ToolMessage(fetch_webpage(Failed to fetch the ...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://www.filmweb.pl/serial/Wszyscy+kochaj%C4%85+Romana-2011-624596/cast/actors'})),) -> ToolMessage(fetch_webpage(Wszyscy kochają Roma...)) -> AIMessage(tool_call(web_search({'query': 'Magda M. cast'})),) -> ToolMessage(web_search(# Watch Magda M. (20...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://en.wikipedia.org/wiki/Magda_M.'})),) -> ToolMessage(fetch_webpage(Failed to fetch the ...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://www.tvmaze.com/shows/5863/magda-m/cast'})),) -> ToolMessage(fetch_webpage(Magda M. - Cast | TV...)) -> AIMessage(The actor who played...)\"}\n",
      "[11] {'task_id': 'f918266a-b3e0-4914-865d-4faa564f1aef', 'submitted_answer': '0', 'reasoning_trace': \"HumanMessage(What is the final nu...) -> AIMessage(tool_call(execute_python_file({'file_path': '/home/cora/.cache/huggingface/hub/datasets--gaia-benchmark--GAIA/snapshots/682dd723ee1e1697e00360edccf2366dc8418dd9/2023/validation/f918266a-b3e0-4914-865d-4faa564f1aef.py'})),) -> ToolMessage(execute_python_file(Working...\\nPlease wa...)) -> AIMessage(The final numeric ou...)\"}\n",
      "[12] {'task_id': '3f57289b-8c60-48be-bd80-01f8099ca449', 'submitted_answer': '519', 'reasoning_trace': \"HumanMessage(How many at bats did...) -> AIMessage(tool_call(web_search({'query': 'Reggie Jackson 1977 season stats'})),) -> ToolMessage(web_search(# Reggie Jackson 197...)) -> AIMessage(tool_call(web_search({'query': 'Reggie Jackson 1977 regular season stats'})),) -> ToolMessage(web_search(# Reggie Jackson's 1...)) -> AIMessage(tool_call(web_search({'query': 'Reggie Jackson 1977 regular season at bats'})),) -> ToolMessage(web_search(# On October 18, 197...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://www.baseball-almanac.com/players/hittinglogs.php?p=jacksre01&y=1977'})),) -> ToolMessage(fetch_webpage(Failed to fetch the ...)) -> AIMessage(tool_call(web_search({'query': 'Reggie Jackson 1977 regular season at bats Baseball Reference'})),) -> ToolMessage(web_search(# 1977 Major League ...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://www.baseball-reference.com/players/j/jacksre01.shtml'})),) -> ToolMessage(fetch_webpage(Failed to fetch the ...)) -> AIMessage(I am unable to acces...) -> HumanMessage(FINAL ANSWER REVIEW ...) -> AIMessage(tool_call(web_search({'query': 'Reggie Jackson 1977 regular season at bats'})),) -> ToolMessage(web_search(# On October 18, 197...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://www.baseball-reference.com/players/j/jacksre01.shtml'})),) -> ToolMessage(fetch_webpage(Failed to fetch the ...)) -> AIMessage(I am unable to acces...) -> HumanMessage(FINAL ANSWER REVIEW ...) -> AIMessage(tool_call(web_search({'query': 'New York Yankees player with most walks 1977'})),) -> ToolMessage(web_search(# 1977 Yankees Playe...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://www.statmuse.com/mlb/ask/which-yankee-had-the-most-walks-in-the-1977-regular-season-and-how-many-at-bats-did-he-have'})),) -> ToolMessage(fetch_webpage(Which Yankee Had The...)) -> AIMessage(Roy White had the mo...)\"}\n",
      "[13] {'task_id': '1f975693-876d-457b-a649-393859e79bf3', 'submitted_answer': '132, 133, 134, 197, 245', 'reasoning_trace': \"HumanMessage(Hi, I was out sick f...) -> AIMessage(tool_call(transcript_audio_file({'file_path': '/home/cora/.cache/huggingface/hub/datasets--gaia-benchmark--GAIA/snapshots/682dd723ee1e1697e00360edccf2366dc8418dd9/2023/validation/1f975693-876d-457b-a649-393859e79bf3.mp3'})),) -> ToolMessage(transcript_audio_file( Before you all go, ...)) -> AIMessage(The page numbers you...)\"}\n",
      "[14] {'task_id': '840bfca7-4f7b-481a-8794-c560c340185d', 'submitted_answer': '80GSFC21M0002', 'reasoning_trace': \"HumanMessage(On June 6, 2023, an ...) -> AIMessage(tool_call(web_search({'query': 'Carolyn Collins Petersen article Universe Today June 6 2023'})),) -> ToolMessage(web_search(# There Are Hundreds...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://www.universetoday.com/articles/there-are-hundreds-of-mysterious-filaments-at-the-center-of-the-milky-way', 'max_chars': 3000})),) -> ToolMessage(fetch_webpage(There Are Hundreds o...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://www.universetoday.com/articles/there-are-hundreds-of-mysterious-filaments-at-the-center-of-the-milky-way', 'max_chars': 6000})),) -> ToolMessage(fetch_webpage(There Are Hundreds o...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://www.universetoday.com/articles/there-are-hundreds-of-mysterious-filaments-at-the-center-of-the-milky-way', 'max_chars': 9000})),) -> ToolMessage(fetch_webpage(There Are Hundreds o...)) -> AIMessage(tool_call(arxiv_search({'query': 'The Population of the Galactic Center Filaments: Position Angle Distribution Reveals a Degree-scale Collimated Outflow from Sgr A* along the Galactic Plane'})),) -> ToolMessage(arxiv_search(# Observation of the...)) -> AIMessage(tool_call(web_search({'query': 'The Population of the Galactic Center Filaments: Position Angle Distribution Reveals a Degree-scale Collimated Outflow from Sgr A* along the Galactic Plane PDF'})),) -> ToolMessage(web_search(# The Population of ...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://ui.adsabs.harvard.edu/abs/2023ApJ...949L..31Y/abstract', 'max_chars': 3000})),) -> ToolMessage(fetch_webpage(...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://researchers.mq.edu.au/en/publications/the-population-of-the-galactic-center-filaments-position-angle-di', 'max_chars': 3000})),) -> ToolMessage(fetch_webpage(The Population of th...)) -> AIMessage(tool_call(download_and_read_pdf({'url': 'https://doi.org/10.3847/2041-8213/acd54b'})),) -> ToolMessage(download_and_read_pdf(Failed to download o...)) -> AIMessage(tool_call(web_search({'query': 'The Population of the Galactic Center Filaments: Position Angle Distribution Reveals a Degree-scale Collimated Outflow from Sgr A* along the Galactic Plane PDF download'})),) -> ToolMessage(web_search(# The Population of ...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://www.semanticscholar.org/paper/The-population-of-Galactic-centre-filaments-III%3A-Yusef-Zadeh-Arendt/fd66ce38e439783ac3726b9778c7a0ed0ba81110', 'max_chars': 3000})),) -> ToolMessage(fetch_webpage(...)) -> AIMessage(tool_call(web_search({'query': 'The Population of the Galactic Center Filaments: Position Angle Distribution Reveals a Degree-scale Collimated Outflow from Sgr A* along the Galactic Plane PDF link'})),) -> ToolMessage(web_search(# The Population of ...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://ora.ox.ac.uk/objects/uuid:c702c7cb-af8d-417a-aef9-05d623db9388', 'max_chars': 3000})),) -> ToolMessage(fetch_webpage(The Population of th...)) -> AIMessage(tool_call(arxiv_search({'query': 'The Population of the Galactic Center Filaments: Position Angle Distribution Reveals a Degree-scale Collimated Outflow from Sgr A* along the Galactic Plane'})),) -> ToolMessage(arxiv_search(# Observation of the...)) -> AIMessage(tool_call(web_search({'query': 'The Population of the Galactic Center Filaments: Position Angle Distribution Reveals a Degree-scale Collimated Outflow from Sgr A* along the Galactic Plane PDF download link'})),) -> ToolMessage(web_search(# The Population of ...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://x.com/Ritholtz/status/1665561329250316290', 'max_chars': 3000})),) -> ToolMessage(fetch_webpage(Barry Ritholtz on X:...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://iopscience.iop.org/article/10.3847/2041-8213/acd54b', 'max_chars': 3000})),) -> ToolMessage(fetch_webpage(Radware Bot Manager ...)) -> AIMessage(I was unable to acce...) -> HumanMessage(FINAL ANSWER REVIEW ...) -> AIMessage(tool_call(web_search({'query': 'The Population of the Galactic Center Filaments: Position Angle Distribution Reveals a Degree-scale Collimated Outflow from Sgr A* along the Galactic Plane acknowledgments NASA award number R. G. Arendt'})),) -> ToolMessage(web_search(# THE Definition & M...)) -> AIMessage(tool_call(web_search({'query': 'NASA award number R. G. Arendt support The Population of the Galactic Center Filaments paper'})),) -> ToolMessage(web_search(# Statistical proper...)) -> AIMessage(The work performed b...)\"}\n",
      "[15] {'task_id': 'bda648d7-d618-4883-88f4-3466eabd860e', 'submitted_answer': 'Saint Petersburg', 'reasoning_trace': \"HumanMessage(Where were the Vietn...) -> AIMessage(tool_call(web_search({'query': 'Nedoshivina 2010 paper Kuznetzov Vietnamese specimens deposition'})),) -> ToolMessage(web_search(# Tortricidae) in Ja...)) -> AIMessage(The Vietnamese speci...)\"}\n",
      "[16] {'task_id': 'cf106601-ab4f-4af9-b045-5295fe67b37d', 'submitted_answer': 'CUB', 'reasoning_trace': \"HumanMessage(What country had the...) -> AIMessage(tool_call(wikipedia_section({'title': '1928 Summer Olympics', 'section': 'Participating nations'})),) -> ToolMessage(wikipedia_section(# 1928 Summer Olympi...)) -> AIMessage(The countries with t...)\"}\n",
      "[17] {'task_id': 'a0c07678-e491-4bbc-8f0b-07405144218f', 'submitted_answer': 'Yamasaki, Uehara', 'reasoning_trace': \"HumanMessage(Who are the pitchers...) -> AIMessage(tool_call(web_search({'query': 'Taishō Tamai uniform number Hokkaido Nippon-Ham Fighters July 2023'})),) -> ToolMessage(web_search(# [Official] Taisho ...)) -> AIMessage(tool_call(fetch_webpage({'url': 'https://pacificleague.com/en/player/517064', 'max_chars': 5000})),) -> ToolMessage(fetch_webpage(Information for user...)) -> AIMessage(tool_call(wikipedia_section({'title': 'Taishō Tamai'})),) -> ToolMessage(wikipedia_section(# Taishō Tamai\\n**URL...)) -> AIMessage(Taishō Tamai's unifo...)\"}\n",
      "[18] {'task_id': '7bd855d8-463d-4ed5-93ca-5fe35145f733', 'submitted_answer': '90798.00', 'reasoning_trace': 'HumanMessage(The attached Excel f...) -> AIMessage(tool_call(read_excel_file({\\'file_path\\': \\'/home/cora/.cache/huggingface/hub/datasets--gaia-benchmark--GAIA/snapshots/682dd723ee1e1697e00360edccf2366dc8418dd9/2023/validation/7bd855d8-463d-4ed5-93ca-5fe35145f733.xlsx\\'})),) -> ToolMessage(read_excel_file(   Location  Burgers...)) -> AIMessage(tool_call(execute_python_code({\\'code\\': \"import pandas as pd\\\\n\\\\n# Data from the Excel file\\\\nsales_data = {\\\\n    \\'Location\\': [\\'Pinebrook\\', \\'Wharvton\\', \\'Sagrada\\', \\'Algrimand\\', \\'Marztep\\', \\'San Cecelia\\', \\'Pimento\\', \\'Tinseles\\', \\'Rosdale\\'],\\\\n    \\'Burgers\\': [1594, 1983, 2019, 1958, 2015, 2011, 2017, 1967, 2007],\\\\n    \\'Hot Dogs\\': [1999, 2008, 2022, 1971, 2016, 2010, 1999, 1969, 2009],\\\\n    \\'Salads\\': [2002, 2014, 2022, 1982, 2018, 2012, 2001, 1982, 2021],\\\\n    \\'Fries\\': [2005, 2015, 2023, 1989, 2019, 2013, 2003, 1994, 1989],\\\\n    \\'Ice Cream\\': [1977, 2017, 2021, 1998, 2021, 2015, 1969, 2005, 2005],\\\\n    \\'Soda\\': [1980, 2018, 2019, 2009, 2022, 2016, 2967, 2006, 2011]\\\\n}\\\\n\\\\n# Convert to DataFrame\\\\ndf = pd.DataFrame(sales_data)\\\\n\\\\n# Calculate total sales from food items\\\\nfood_items = [\\'Burgers\\', \\'Hot Dogs\\', \\'Salads\\', \\'Fries\\', \\'Ice Cream\\']\\\\ntotal_food_sales = df[food_items].sum().sum()\\\\ntotal_food_sales\"})),) -> ToolMessage(execute_python_code(...)) -> AIMessage(The total sales from...) -> HumanMessage(FINAL ANSWER REVIEW ...) -> AIMessage(tool_call(execute_python_code({\\'code\\': \\'# Recalculate the total sales from food items\\\\n# Ensure no inclusion of drinks and correct calculation\\\\n\\\\n# Calculate total sales from food items again\\\\ncorrect_total_food_sales = df[food_items].sum().sum()\\\\ncorrect_total_food_sales\\'})),) -> ToolMessage(execute_python_code(Error executing subp...)) -> AIMessage(tool_call(execute_python_code({\\'code\\': \"# Re-import pandas and recalculate the total sales from food items\\\\nimport pandas as pd\\\\n\\\\n# Data from the Excel file\\\\nsales_data = {\\\\n    \\'Location\\': [\\'Pinebrook\\', \\'Wharvton\\', \\'Sagrada\\', \\'Algrimand\\', \\'Marztep\\', \\'San Cecelia\\', \\'Pimento\\', \\'Tinseles\\', \\'Rosdale\\'],\\\\n    \\'Burgers\\': [1594, 1983, 2019, 1958, 2015, 2011, 2017, 1967, 2007],\\\\n    \\'Hot Dogs\\': [1999, 2008, 2022, 1971, 2016, 2010, 1999, 1969, 2009],\\\\n    \\'Salads\\': [2002, 2014, 2022, 1982, 2018, 2012, 2001, 1982, 2021],\\\\n    \\'Fries\\': [2005, 2015, 2023, 1989, 2019, 2013, 2003, 1994, 1989],\\\\n    \\'Ice Cream\\': [1977, 2017, 2021, 1998, 2021, 2015, 1969, 2005, 2005],\\\\n    \\'Soda\\': [1980, 2018, 2019, 2009, 2022, 2016, 2967, 2006, 2011]\\\\n}\\\\n\\\\n# Convert to DataFrame\\\\ndf = pd.DataFrame(sales_data)\\\\n\\\\n# Calculate total sales from food items\\\\nfood_items = [\\'Burgers\\', \\'Hot Dogs\\', \\'Salads\\', \\'Fries\\', \\'Ice Cream\\']\\\\ntotal_food_sales = df[food_items].sum().sum()\\\\ntotal_food_sales\"})),) -> ToolMessage(execute_python_code(...)) -> AIMessage(The recalculated tot...)'}\n",
      "[19] {'task_id': '5a0c1adf-205e-4841-a666-7c3ef95def9d', 'submitted_answer': 'Claus', 'reasoning_trace': \"HumanMessage(What is the first na...) -> AIMessage(tool_call(wikipedia_search({'query': 'Malko Competition winners'})),) -> ToolMessage(wikipedia_search(# Malko Competition\\n...)) -> AIMessage(tool_call(wikipedia_section({'title': 'Malko Competition', 'section': 'Recipients'})),) -> ToolMessage(wikipedia_section(# Malko Competition\\n...)) -> AIMessage(The only Malko Compe...)\"}\n"
     ]
    }
   ],
   "source": [
    "gaia_reponses = []\n",
    "\n",
    "for i, question in enumerate(questions):\n",
    "    task_id, q_query = question_query(question)\n",
    "    response = agent(q_query)\n",
    "    result = pretty_print_response(task_id, response)\n",
    "    gaia_reponses.append(result)\n",
    "    print(f\"[{i}] {result}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c36f2b10",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "================================\u001b[1m Human Message \u001b[0m=================================\n",
      "\n",
      "On June 6, 2023, an article by Carolyn Collins Petersen was published in Universe Today. This article mentions a team that produced a paper about their observations, linked at the bottom of the article. Find this paper. Under what NASA award number was the work performed by R. G. Arendt supported by?\n",
      "==================================\u001b[1m Ai Message \u001b[0m==================================\n",
      "Tool Calls:\n",
      "  web_search (call_0vOiLAagqh95F6UIgFlWF5L7)\n",
      " Call ID: call_0vOiLAagqh95F6UIgFlWF5L7\n",
      "  Args:\n",
      "    query: Carolyn Collins Petersen article Universe Today June 6 2023\n",
      "=================================\u001b[1m Tool Message \u001b[0m=================================\n",
      "Name: web_search\n",
      "\n",
      "# 17566 Search Results for \"Universe Today June 6 2023 Carolyn Collins Petersen article\"\n",
      "**URL:** https://www.nasa.gov?search=Universe+Today+June+6+2023+Carolyn+Collins+Petersen+article\\\n",
      "**SCORE:** 0.89 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "Title: 17566 Search Results for \"Universe Today June 6 2023 Carolyn Collins Petersen article\"\n",
      "### NASA Releases Artemis II Moon Mission Launch Countdown. ### NASA’s Artemis II Moon Mission Daily Agenda. ### Track NASA’s Artemis II Mission in Real Time. ### About NASA. NASA Releases Artemis II Moon Mission Launch Countdown. ### NASA Releases Artemis II Moon Mission Launch Countdown. article  6 days ago Preparing for Artemis II: Training for a Mission Around the Moon. article  2 months ago To Protect Artemis II Astronauts, NASA Experts Keep Eyes on Sun. article  5 days ago NASA Releases Artemis II Moon Mission Launch Countdown. ### NASA Releases Artemis II Moon Mission Launch Countdown. article  3 hours ago NASA Selects Intuitive Machines to Deliver Artemis Science, Tech to Moon. ### NASA Selects Intuitive Machines to Deliver Artemis Science, Tech to Moon. article  1 week ago NASA-JAXA’s XRISM Telescope Clocks Hot Wind of Galaxy M82. article  1 week ago NASA’s Hubble Revisits Crab Nebula to Track 25 Years of Expansion. article  2 days ago NASA Selects Intuitive Machines to Deliver Artemis Science, Tech to Moon. ### NASA Selects Intuitive Machines to Deliver Artemis Science, Tech to Moon. article  1 week ago NASA’s X-59 Experimental Supersonic Aircraft Makes Second Flight. article  2 weeks ago NASA Simulations Improve Artemis II Launch Environment. ### NASA Simulations Improve Artemis II Launch Environment. article  5 days ago NASA Tech and Science Bound for Low Earth Orbit on Commercial Launch. NASA’s Artemis II Moon Mission Daily Agenda. ### NASA’s Artemis II Moon Mission Daily Agenda. article  3 weeks ago Preparing for Artemis II: Training for a Mission Around the Moon. article  2 months ago Track NASA’s Artemis II Mission in Real Time. ### Track NASA’s Artemis II Mission in Real Time. article  7 days ago Agenda diaria de la misión a la Luna de Artemis II de la NASA. article  3 weeks ago La NASA refuerza Artemis: añade una misión y perfecciona su arquitectura general. *https://science.nasa.gov/universe/what-can-we-learn-from-the-universes-baby-picture*. *https://www.nasa.gov/missions/chandra/nasa-telescopes-spot-surprisingly-mature-cluster-in-early-universe*. *https://www.nasa.gov/universe/new-video-shows-james-webb-space-telescopes-instruments-removed-from-super-cold-chamber*. **article 6** days ago · NASA Simulations Improve ... **article 6** days ago · NASA's SpaceX Crew-11 Wraps Up ... **article6** days ago · NASA, Oxford Discover Warmer ... *https://science.nasa.gov/missions/hubble/hubble-reaches-the-undiscovered-country-of-primeval-galaxies*. **article 6** days ago · **6** Things to Know From NASA About New US, European Sea Satellite.\n",
      "\n",
      "---\n",
      "\n",
      "# Universe Today Carolyn Collins Petersen June 6 2023 Article\n",
      "**URL:** https://www.instagram.com/popular/universe-today-carolyn-collins-petersen-june-6-2023-article\\\n",
      "**SCORE:** 0.88 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "[![: TON 618 doesn't break the rules of the universe. What they found left the astronomy world speechless 🤯 Here's what lives 10.4 billion light years away from us 🔴 Its mass is 66 billion times that of our Sun making it one of the most massive single objects ever discovered in the observable universe 🌀 Its event horizon is so wide that our entire solar system would fit inside it 11 times over with room to spare ⏳ Light falling into TON 618 takes thousands of years just to cross from one side of its event horizon to the other 💀 The gravitational time dilation near its edge is so extreme that minutes there could equal thousands of years passing in the outside universe 🔭 It powers a quasar, a jet of energy so bright it outshines entire galaxies containing hundreds of billions of stars 📡 Despite being 10.4 billion light years away it is still one of the brightest objects ever detected proving its energy output is almost incomprehensible INTERESTING FACT: TON 618 formed when the universe was only a fraction of its current age meaning something this massive existed before most galaxies we see today were even born 💫 Some things in this universe don't just exceed our imagination. If there are any problems/issues in the post please dm, we will fix it)]() [: mysteries_of_universe_'s profile picture]() mysteries_of_... 9.6M TON 618 doesn't break the rules of the universe. [![: TON 618 doesn't break the rules of the universe. If there are any problems/issues in the post please dm, we will fix it)]() [: futuregenspace's profile picture]() futuregenspace 17.2K TON 618 doesn't break the rules of the universe. [![: TON 618 doesn't break the rules of the universe. Please follow my page @nasa_void for more!]() [: nasa_void's profile picture]() nasa_void 201K TON 618 doesn't break the rules of the universe.\n",
      "\n",
      "---\n",
      "\n",
      "# 8 Search Results for \"Paper linked at bottom of Universe Today article by Carolyn Collins Petersen June 6, 2023\"\n",
      "**URL:** https://www.nasa.gov?search=Paper+linked+at+bottom+of+Universe+Today+article+by+Carolyn+Collins+Petersen+June+6%2C+2023\\\n",
      "**SCORE:** 0.85 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "article2 days ago5 min read. ### NASA’s Chandra Finds Small Galaxies May Buck the Black Hole Trend. article3 days ago5 min read. ### NASA’s Parker Solar Probe Spies Solar Wind ‘U-Turn’. ### About NASA. ### NASA en Español. article 2 days ago 7 min read. ### NASA Announces Plan to Map Milky Way With Roman Space Telescope. article 2 days ago 2 min read. ### NASA Selects 2 Instruments for Artemis IV Lunar Surface Science. ### NASA Astronaut Jonny Kim Advances Research Aboard Space Station. ### CHAPEA Crew Begins Stay Inside NASA’s Mars Habitat for Second Mission. article 3 days ago 5 min read. ### NASA’s Parker Solar Probe Spies Solar Wind ‘U-Turn’. article 3 days ago 4 min read. ### NASA’s Parker Solar Probe Spies Solar Wind ‘U-Turn’. article 3 days ago 5 min read. ### NASA JPL Unveils Rover Operations Center for Moon, Mars Missions. article 4 days ago 3 min read. ### NASA Wins Second Emmy Award for 2024 Total Solar Eclipse Broadcast. ### NASA Announces Plan to Map Milky Way With Roman Space Telescope. ### NASA’s Chandra Finds Small Galaxies May Buck the Black Hole Trend. ### NASA Works with Boeing, Other Collaborators Toward More Efficient Global Flights. ### NASA Demonstrates Safer Skies for Future Urban Air Travel. ### New NASA Sensor Goes Hunting for Critical Minerals. ### NASA Wins Second Emmy Award for 2024 Total Solar Eclipse Broadcast. ### Pódcast en español de la NASA estrena su tercera temporada. article 6 months ago 5 min read. ### Las carreras en la NASA despegan con las pasantías. article 7 months ago 4 min read. ### El X-59 de la NASA completa las pruebas electromagnéticas. # Search Results for: Paper linked at bottom of Universe Today article by Carolyn Collins Petersen June 6, 2023. * We're working continuously to improve our search results. #### Women of JSC: Apollo to Artemis - NASA. the legacy of women in spaceflight with panelists who worked at NASA during...employees who carry on that legacy today. Center - America’s gateway to the universe Spaceport News lands safely at NASA’s Dryden Flight Research Center at... Space Center America's gateway to the universe.\n",
      "\n",
      "---\n",
      "\n",
      "# 4 Search Results for \"Carolyn Collins Petersen Universe Today June 6 2023\"\n",
      "**URL:** https://www.nasa.gov?search=Carolyn+Collins+Petersen+Universe+Today+June+6+2023\\\n",
      "**SCORE:** 0.82 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "4 Search Results for \"Carolyn Collins Petersen Universe Today June 6 2023\". ### News & Events. ### Multimedia. * View All Topics A-Z. ### Featured. ### Missions. * A to Z List of Missions. ### Humans in Space. ### Earth. ### The Solar System. ### The Universe. * The Search for Life in the Universe. ### Science. ### Aeronautics. ### Technology. ### Learning Resources. * Requests for Exhibits, Artifacts, or Speakers. ### About NASA. ### NASA en Español. ### News & Events. ### Multimedia. ### Featured. ### Highlights. ### Highlights. ### Highlights. ### Featured. ### Highlights. ### Highlights. ### Highlights. ### Featured. ### Featured. ### Highlights. Search Results for: Carolyn Collins Petersen Universe Today June 6 2023. ### Content Type. ### Categories. ### Changed. ### Language. * We're working continuously to improve our search results. If you aren't finding what you expect, please use our feedback form to share the search query and expected result with us so we can improve. * #### Hubble Space Telescope Bibliography Journey Through Time: Exploring the Universe With the Hubble Space Telescope. New...Hubble Space Telescope: Imaging the Universe. * #### 1996.pdf then transported to Pad 39-B on Dec. 6. Endeavour last flew in September 1995...planned two-hour built-in hold at the T-6 hour mark (5:58 p.m.) *... * #### Aug 22 Center - America’s gateway to the universe Spaceport News heads of state. Commander Eileen Collins and the crew of the... * #### June6snews Space Center America's gateway to the universe. Leading the world in preparing and...launching missions to Earth and beyond. ### National Aeronautics and Space Administration. NASA explores the unknown in air and space, innovates for the benefit of humanity, and inspires the world through discovery. * NASA en Español (opens in new tab). ### Follow NASA. * Office of the IG (opens in new tab). * Responsible NASA Official: Abigail Bowman.\n",
      "\n",
      "---\n",
      "\n",
      "# Bibliographies: 'paper referenced in Carolyn Collins Petersen article Universe Today June 6, 2023' – Grafiati\n",
      "**URL:** https://www.grafiati.com/en/literature-selections?q=paper+referenced+in+Carolyn+Collins+Petersen+article+Universe+Today+June+6%2C+2023\\\n",
      "**SCORE:** 0.81 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "# Relevant bibliographies by topics. ## Search by topic. Select a research topic from the given list or using the search bar. Criterion-referenced tests Wolfgang Petersen Today-Tonight Carolyn Bremer Article 6 Petersen Gdpr article 6 America today USA Today Referenced alignment June 6 June Bjelke-Petersen Collins Erich Petersen Joy Petersen Petersen, Johann Wilhelm, Petersen, Johanna Eleonora Carolyn Carolyn Jones Petersen, Carl Today show Carolyn Mackler Christianity Today Collins, Wilkie, Collins, Wilkie Norm-Referenced Referenced mothers Referenced microscopy GATS Article VI:6 Creep in paper Criterion-referenced Russia Today Carolyn Slaughter Petersen Coil ANC Today Education White Paper 6 White Paper 6 Lincoln-Petersen Today Distance in the universe Article 6 de la CEDH Peter Petersen Carolyn Eisele Today (Beijing Lesotho Today Propagation in the Universe Carolyn Ellis 6 june 1944 Article- In-Cell Direct applicability of Article 6 Invocation of article 6 P.J. Petersen Article 6 CEDH Article 6 EHCR Carolyn Forche Carolyn lindsay Referenced Article VI (6) Brown Senier, Carolyn June june Chile.\n",
      "==================================\u001b[1m Ai Message \u001b[0m==================================\n",
      "Tool Calls:\n",
      "  web_search (call_qMncqoQFE1LQAcvwaNGOU5Bq)\n",
      " Call ID: call_qMncqoQFE1LQAcvwaNGOU5Bq\n",
      "  Args:\n",
      "    query: Universe Today Carolyn Collins Petersen June 6 2023 article paper link\n",
      "=================================\u001b[1m Tool Message \u001b[0m=================================\n",
      "Name: web_search\n",
      "\n",
      "# Beyond Parallel Sampling: Diverse Query Initialization for ...\n",
      "**URL:** https://arxiv.org/html/2606.17209v1\\\n",
      "**SCORE:** 0.86 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "Jun 15, 2026 — On June 6, 2023, an article by Carolyn Collins Petersen was published in Universe Today. This article mentions a team that produced a paper\n",
      "\n",
      "---\n",
      "\n",
      "# There Are Hundreds of Mysterious Filaments at the Center ...\n",
      "**URL:** https://www.universetoday.com/articles/there-are-hundreds-of-mysterious-filaments-at-the-center-of-the-milky-way\\\n",
      "**SCORE:** 0.79 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "There Are Hundreds of Mysterious Filaments at the Center of the Milky Way. By Carolyn Collins Petersen - June 06, 2023 06:00 PM UTC | Milky Way.Read more\n",
      "\n",
      "---\n",
      "\n",
      "# Carolyn Collins Petersen\n",
      "**URL:** https://www.universetoday.com/authors/cc-petersen\\\n",
      "**SCORE:** 0.60 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "Carolyn Collins Petersen is a long-time science writer and former astronomy researcher. She writes about astronomy and space exploration.Read more\n",
      "\n",
      "---\n",
      "\n",
      "# Searching Out Missing Links in Galaxy Evolution\n",
      "**URL:** https://www.universetoday.com/articles/searching-out-missing-links-in-galaxy-evolution\\\n",
      "**SCORE:** 0.59 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "Feb 20, 2026 — Carolyn Collins Petersen is a long-time science writer and former astronomy researcher. She writes about astronomy and space exploration and has ...Read more\n",
      "\n",
      "---\n",
      "\n",
      "# There's a universe full of black holes out there, spinning ...\n",
      "**URL:** https://www.facebook.com/groups/AstroCosmoNews/posts/3502905580014990\\\n",
      "**SCORE:** 0.53 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "Carolyn Collins Petersen, Universe Today; There's a universe full of black holes out there, spinning merrily away—some fast, others more slowly.\n",
      "==================================\u001b[1m Ai Message \u001b[0m==================================\n",
      "Tool Calls:\n",
      "  fetch_webpage (call_GAmslfJA3iflRY4cvsGJPiIL)\n",
      " Call ID: call_GAmslfJA3iflRY4cvsGJPiIL\n",
      "  Args:\n",
      "    url: https://www.universetoday.com/articles/there-are-hundreds-of-mysterious-filaments-at-the-center-of-the-milky-way\n",
      "    max_chars: 3000\n",
      "=================================\u001b[1m Tool Message \u001b[0m=================================\n",
      "Name: fetch_webpage\n",
      "\n",
      "There Are Hundreds of Mysterious Filaments at the Center of the Milky Way - Universe Today\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Several million years ago, the core of our galaxy experienced a powerful event. It blew out a huge bi-lobed bubble that blasted through the interstellar medium in two directions. Whatever it was, it released huge amounts of energy from the central supermassive black hole, Sagittarius A\\* (Sgr A\\* for short).\n",
      "\n",
      "These bubbles contain weird, spindly one-dimensional vertical filaments that emit radio signals of their own. In 1984, astronomer Farhad Yusef-Zadeh at Northwestern University and his collaborators were\n",
      "the first to observe these wispy trails\n",
      ". Now, they know of more than a thousand of them. They give off radio waves in the microwave portion of the electromagnetic spectrum.\n",
      "\n",
      "## Wait, There's More!\n",
      "\n",
      "Recently, the team found a smaller population of filaments near Sgr A\\*. However, these aren't vertical. Instead, they lie along the galactic plane and radiate away like spokes on a wheel. \"It was a surprise to suddenly find a new population of structures that seem to be pointing in the direction of the black hole,\" Yusef-Zadeh said. \"We have always been thinking about vertical filaments and their origin. I'm used to them being vertical. I never considered there might be others along the plane.\"\n",
      "\n",
      "He and the team were stunned when they saw these radially stretched structures. \"We had to do a lot of work to establish that we weren't fooling ourselves,\" he said. \"And we found that these filaments are not random but appear to be tied to the outflow of our black hole. By studying them, we could learn more about the black hole's spin and accretion disk orientation. It is satisfying when one finds order in a middle of a chaotic field of the nucleus of our galaxy.\"\n",
      "\n",
      "Yusef-Zadeh pointed out that something clearly occurred in the core to create the short radial filaments. \"We think they must have originated with some kind of outflow from an activity that happened a few million years ago,\" he said. \"It seems to be the result of an interaction of that outflowing material with objects near it.\"\n",
      "\n",
      "## Pinning Down Filament Histories\n",
      "\n",
      "Like other supermassive black holes, Sgr A\\* has been known to swallow up material occasionally. Then, it \"burps out\" an energetic, highly magnetized blast of energy and plasma. That seems the most likely explanation for the appearance of the bubbles and filaments. It's possible that such a jet collided with clouds of gas and dust in the immediate neighborhood. It could have led to twisted magnetic structures that form the filaments.\n",
      "\n",
      "Another idea is that molecular gas got stretched by the expansion of the material near the core. That somehow creates the radial filaments. Such action would also explain the lopsided appearance of the Sgr E complex. It lies to one side of the black hole at the intersection of the central molecular zone and a wide central dust lane. Finally, it's also possible that the gravitational potential of our galaxy's c\n",
      "==================================\u001b[1m Ai Message \u001b[0m==================================\n",
      "Tool Calls:\n",
      "  web_search (call_JHHjQrNeP5aubNJuQGhHxn5w)\n",
      " Call ID: call_JHHjQrNeP5aubNJuQGhHxn5w\n",
      "  Args:\n",
      "    query: paper linked at bottom of Universe Today article by Carolyn Collins Petersen June 6 2023\n",
      "=================================\u001b[1m Tool Message \u001b[0m=================================\n",
      "Name: web_search\n",
      "\n",
      "# Beyond Parallel Sampling: Diverse Query Initialization for ...\n",
      "**URL:** https://arxiv.org/html/2606.17209v1\\\n",
      "**SCORE:** 0.81 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "Jun 15, 2026 — On June 6, 2023, an article by Carolyn Collins Petersen was published in Universe Today. This article mentions a team that produced a paper\n",
      "\n",
      "---\n",
      "\n",
      "# There Are Hundreds of Mysterious Filaments at the Center ...\n",
      "**URL:** https://www.universetoday.com/articles/there-are-hundreds-of-mysterious-filaments-at-the-center-of-the-milky-way\\\n",
      "**SCORE:** 0.74 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "There Are Hundreds of Mysterious Filaments at the Center of the Milky Way. By Carolyn Collins Petersen - June 06, 2023 06:00 PM UTC | Milky Way.Read more\n",
      "\n",
      "---\n",
      "\n",
      "# Carolyn Collins Petersen\n",
      "**URL:** https://www.universetoday.com/authors/cc-petersen\\\n",
      "**SCORE:** 0.50 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "Carolyn Collins Petersen is a long-time science writer and former astronomy researcher. She writes about astronomy and space exploration.Read more\n",
      "\n",
      "---\n",
      "\n",
      "# There's a universe full of black holes out there, spinning ...\n",
      "**URL:** https://www.facebook.com/groups/AstroCosmoNews/posts/3502905580014990\\\n",
      "**SCORE:** 0.47 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "Carolyn Collins Petersen, Universe Today; A recent survey of supermassive black holes reveals that their spin rates reveal something about\n",
      "\n",
      "---\n",
      "\n",
      "# Space.com - \"There is a growing chasm between theory and...\n",
      "**URL:** https://www.facebook.com/spacecom/posts/there-is-a-growing-chasm-between-theory-and-observation-related-to-the-early-uni/1264329448891070\\\n",
      "**SCORE:** 0.27 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "## Space.com's post. ### **Space.com**. '/%3E%3Cpath d='M16.0001 7.9996c0 4.418-3.5815 7.9996-7.9995 7.9996S.001 12.4176.001 7.9996 3.5825 0 8.0006 0C12.4186 0 16 3.5815 16 7.9996Z' fill='url(%23paint1_radial_15251_63610)'/%3E%3Cpath d='M16.0001 7.9996c0 4.418-3.5815 7.9996-7.9995 7.9996S.001 12.4176.001 7.9996 3.5825 0 8.0006 0C12.4186 0 16 3.5815 16 7.9996Z' fill='url(%23paint2_radial_15251_63610)' fill-opacity='.5'/%3E%3Cpath d='M7.3014 3.8662a.6974.6974 0 0 1 .6974-.6977c.6742 0 1.2207.5465 1.2207 1.2206v1.7464a.101.101 0 0 0 .101.101h1.7953c.992 0 1.7232.9273 1.4917 1.892l-.4572 1.9047a2.301 2.301 0 0 1-2.2374 1.764H6.9185a.5752.5752 0 0 1-.5752-.5752V7.7384c0-.4168.097-.8278.2834-1.2005l.2856-.5712a3.6878 3.6878 0 0 0 .3893-1.6509l-.0002-.4496ZM4.367 7a.767.767 0 0 0-.7669.767v3.2598a.767.767 0 0 0 .767.767h.767a.3835.3835 0 0 0 .3835-.3835V7.3835A.3835.3835 0 0 0 5.134 7h-.767Z' fill='%23fff'/%3E%3Cdefs%3E%3CradialGradient id='paint1_radial_15251_63610' cx='0' cy='0' r='1' gradientUnits='userSpaceOnUse' gradientTransform='rotate(90 .0005 8) scale(7.99958)'%3E%3Cstop offset='.5618' stop-color='%230866FF' stop-opacity='0'/%3E%3Cstop offset='1' stop-color='%230866FF' stop-opacity='.1'/%3E%3C/radialGradient%3E%3CradialGradient id='paint2_radial_15251_63610' cx='0' cy='0' r='1' gradientUnits='userSpaceOnUse' gradientTransform='rotate(45 -4.5257 10.9237) scale(10.1818)'%3E%3Cstop offset='.3143' stop-color='%2302ADFC'/%3E%3Cstop offset='1' stop-color='%2302ADFC' stop-opacity='0'/%3E%3C/radialGradient%3E%3ClinearGradient id='paint0_linear_15251_63610' x1='2.3989' y1='2.3999' x2='13.5983' y2='13.5993' gradientUnits='userSpaceOnUse'%3E%3Cstop stop-color='%2302ADFC'/%3E%3Cstop offset='.5' stop-color='%230866FF'/%3E%3Cstop offset='1' stop-color='%232B7EFF'/%3E%3C/linearGradient%3E%3C/defs%3E%3C/svg%3E). '%3E%3Cpath d='M15.9963 8c0 4.4179-3.5811 7.9993-7.9986 7.9993-4.4176 0-7.9987-3.5814-7.9987-7.9992 0-4.4179 3.5811-7.9992 7.9987-7.9992 4.4175 0 7.9986 3.5813 7.9986 7.9992Z' fill='url(%23paint0_linear_15251_63610)'/%3E%3Cpath d='M15.9973 7.9992c0 4.4178-3.5811 7.9992-7.9987 7.9992C3.5811 15.9984 0 12.417 0 7.9992S3.5811 0 7.9986 0c4.4176 0 7.9987 3.5814 7.9987 7.9992Z' fill='url(%23paint1_radial_15251_63610)'/%3E%3Cpath d='M7.9996 5.9081c-.3528-.8845-1.1936-1.507-2.1748-1.507-1.4323 0-2.4254 1.328-2.4254 2.6797 0 2.2718 2.3938 4.0094 4.0816 5.1589.3168.2157.7205.2157 1.0373 0 1.6878-1.1495 4.0815-2.8871 4.0815-5.159 0-1.3517-.993-2.6796-2.4254-2.6796-.9811 0-1.822.6225-2.1748 1.507Z' fill='%23fff'/%3E%3C/g%3E%3Cdefs%3E%3CradialGradient id='paint1_radial_15251_63610' cx='0' cy='0' r='1' gradientUnits='userSpaceOnUse' gradientTransform='matrix(0 7.9992 -7.99863 0 7.9986 7.9992)'%3E%3Cstop offset='.5637' stop-color='%23E11731' stop-opacity='0'/%3E%3Cstop offset='1' stop-color='%23E11731' stop-opacity='.1'/%3E%3C/radialGradient%3E%3ClinearGradient id='paint0_linear_15251_63610' x1='2.3986' y1='2.4007' x2='13.5975' y2='13.5993' gradientUnits='userSpaceOnUse'%3E%3Cstop stop-color='%23FF74AE'/%3E%3Cstop offset='.5001' stop-color='%23FA2E3E'/%3E%3Cstop offset='1' stop-color='%23FF5758'/%3E%3C/linearGradient%3E%3CclipPath id='clip0_15251_63610'%3E%3Cpath fill='%23fff' d='M-.001.0009h15.9992v15.9984H-.001z'/%3E%3C/clipPath%3E%3C/defs%3E%3C/svg%3E). Find out more: https://space-com.visitlink.me/Ia0ETQ.\n",
      "==================================\u001b[1m Ai Message \u001b[0m==================================\n",
      "Tool Calls:\n",
      "  fetch_webpage (call_foJ3gQ9mqwuPfnbdOqyu1Spf)\n",
      " Call ID: call_foJ3gQ9mqwuPfnbdOqyu1Spf\n",
      "  Args:\n",
      "    url: https://www.universetoday.com/articles/there-are-hundreds-of-mysterious-filaments-at-the-center-of-the-milky-way\n",
      "    max_chars: 3000\n",
      "=================================\u001b[1m Tool Message \u001b[0m=================================\n",
      "Name: fetch_webpage\n",
      "\n",
      "There Are Hundreds of Mysterious Filaments at the Center of the Milky Way - Universe Today\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "Several million years ago, the core of our galaxy experienced a powerful event. It blew out a huge bi-lobed bubble that blasted through the interstellar medium in two directions. Whatever it was, it released huge amounts of energy from the central supermassive black hole, Sagittarius A\\* (Sgr A\\* for short).\n",
      "\n",
      "These bubbles contain weird, spindly one-dimensional vertical filaments that emit radio signals of their own. In 1984, astronomer Farhad Yusef-Zadeh at Northwestern University and his collaborators were\n",
      "the first to observe these wispy trails\n",
      ". Now, they know of more than a thousand of them. They give off radio waves in the microwave portion of the electromagnetic spectrum.\n",
      "\n",
      "## Wait, There's More!\n",
      "\n",
      "Recently, the team found a smaller population of filaments near Sgr A\\*. However, these aren't vertical. Instead, they lie along the galactic plane and radiate away like spokes on a wheel. \"It was a surprise to suddenly find a new population of structures that seem to be pointing in the direction of the black hole,\" Yusef-Zadeh said. \"We have always been thinking about vertical filaments and their origin. I'm used to them being vertical. I never considered there might be others along the plane.\"\n",
      "\n",
      "He and the team were stunned when they saw these radially stretched structures. \"We had to do a lot of work to establish that we weren't fooling ourselves,\" he said. \"And we found that these filaments are not random but appear to be tied to the outflow of our black hole. By studying them, we could learn more about the black hole's spin and accretion disk orientation. It is satisfying when one finds order in a middle of a chaotic field of the nucleus of our galaxy.\"\n",
      "\n",
      "Yusef-Zadeh pointed out that something clearly occurred in the core to create the short radial filaments. \"We think they must have originated with some kind of outflow from an activity that happened a few million years ago,\" he said. \"It seems to be the result of an interaction of that outflowing material with objects near it.\"\n",
      "\n",
      "## Pinning Down Filament Histories\n",
      "\n",
      "Like other supermassive black holes, Sgr A\\* has been known to swallow up material occasionally. Then, it \"burps out\" an energetic, highly magnetized blast of energy and plasma. That seems the most likely explanation for the appearance of the bubbles and filaments. It's possible that such a jet collided with clouds of gas and dust in the immediate neighborhood. It could have led to twisted magnetic structures that form the filaments.\n",
      "\n",
      "Another idea is that molecular gas got stretched by the expansion of the material near the core. That somehow creates the radial filaments. Such action would also explain the lopsided appearance of the Sgr E complex. It lies to one side of the black hole at the intersection of the central molecular zone and a wide central dust lane. Finally, it's also possible that the gravitational potential of our galaxy's c\n",
      "==================================\u001b[1m Ai Message \u001b[0m==================================\n",
      "Tool Calls:\n",
      "  web_search (call_ImjPSnWtwsHolSgDkeTnZClR)\n",
      " Call ID: call_ImjPSnWtwsHolSgDkeTnZClR\n",
      "  Args:\n",
      "    query: R. G. Arendt NASA award number support\n",
      "=================================\u001b[1m Tool Message \u001b[0m=================================\n",
      "Name: web_search\n",
      "\n",
      "# population of Galactic Centre filaments – III. Candidate radio and ...\n",
      "**URL:** https://academic.oup.com/mnras/article-abstract/517/1/294/6677408\\\n",
      "**SCORE:** 0.53 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "by F Yusef-Zadeh · 2022 · Cited by 16 — R G Arendt ,. R G Arendt. NASA/GSFC, Code 665. , 8800 Greenbelt Road, Greenbelt, MD 20771, UMBC/CRESST 2,. USA. E-mail: richard.g.arendt@nasa.gov · ORCID logo\n",
      "\n",
      "---\n",
      "\n",
      "# Awards: How to Nominate - Sciences and Exploration Directorate\n",
      "**URL:** https://science.gsfc.nasa.gov/sci/awardsnominate\\\n",
      "**SCORE:** 0.07 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "NASA Logo in the header. ## International Awards. Jansky Lectureship: (February 1)  The Shaw Prize in Astronomy: (September) http://www.shawprize.org/ The Nobel Prize: (February 1)  Templeton Prize: (July 1). ### External Awards and Prizes. + International Award (Earth & Space Sciences). + Athelstan Spilhaus Award (Public Outreach, Earth & Space Sciences). #### American Astronomical Society (AAS) Awards:. Tinsley Prize (Especially Innovative Research) Joseph Weber Award for Astronomical Instrumentation (Instrumentation) Dannie Heineman Prize for Astrophysics (Mid-career) George Van Biesbroeck Prize (Extraordinary Service) Education Prize (Contribution to Education) Annie J. #### AAS Division of Planetary Sciences Awards:. Trumpler Award (PhD thesis) 1 Jan 2013 Klumpke-Roberts Award (Contributions to the public understanding of astronomy) 1 Jan 2013 Maria and Eric Muhlmann Award (Recent significant observational results made possible by innovative advances in astronomical instrumentation, software, or observational infrastructure.) 1 Jan 2013 Thomas J. ## NASA Awards. + the William Nordberg Memorial Award for Earth Science, and.\n",
      "\n",
      "---\n",
      "\n",
      "# Funding Support\n",
      "**URL:** https://eventhorizontelescope.org/funding-support\\\n",
      "**SCORE:** 0.06 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "Awards by the EHT · Awards to the EHT · Funding Support · Industry Donors ... NASA (Fermi Grant). National Institute of Natural Sciences (NINS) of Japan.\n",
      "\n",
      "---\n",
      "\n",
      "# The population of Galactic Centre filaments – III. Candidate radio and ...\n",
      "**URL:** https://ntrs.nasa.gov/api/citations/20230002172/downloads/Arendt%20-%20The%20population%20of%20Galactic%20centre%20filaments%20III.pdf\\\n",
      "**SCORE:** 0.06 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "Candidate radio and stellar sources F. Received 2022 August 22; in original form 2022 June 27 A B S T R A C T Recent MeerKAT radio continuum observations of the Galactic Centre at 20 cm show a large population of non-thermal radio filaments (NRFs) in the inner few hundred pc of the Galaxy. We have selected a sample of 57 radio sources, mainly compact objects, in the MeerKAT mosaic image that appear to be associated with NRFs. The selected sources are about four times the number of radio point sources associated with filaments than would be expected by random chance. To examine if compact radio sources are related to compact IR sources, we have used archi v al 2MASS, and Spitzer data to make spectral energy distribution of individual stellar sources coincident or close to radio sources. We provide a catalogue of radio and IR sources for future detailed observations to investigate a potential three-way physical association between NRFs, compact radio and IR stellar sources.\n",
      "\n",
      "---\n",
      "\n",
      "# Awards - Sciences and Exploration Directorate\n",
      "**URL:** https://science.gsfc.nasa.gov/sci/awards\\\n",
      "**SCORE:** 0.05 (Relevance score from 0 to 1, higher is better)\n",
      "\n",
      "### Editor Award – The Journal of Atmospheric and Oceanic Technology Link Copied to Clipboard. ### Editor Award - Journal of Hydrometeorology Link Copied to Clipboard. ### Marzano Award Gutta Aurea Link Copied to Clipboard. ### GISS Best Publication Award Link Copied to Clipboard. ### AAS Goddard Symposium Earth Science and Applications Award Link Copied to Clipboard. Recognizing their groundbreaking contributions to Earth observations and celebrating the teams' enduring impact and NASA Goddard’s leadership in advancing Earth Science for the Nation. ### William Nordberg Memorial Award for Earth Science Link Copied to Clipboard. ### NASA Wallops Flight Facility Distinguished Public Service Award for contractors Link Copied to Clipboard. ### Water Resources Research Editor’s Choice Award Link Copied to Clipboard. Eagleson Award Link Copied to Clipboard. Lindsay Memorial Award for Space Science Link Copied to Clipboard. Wagner Memorial Award for Women in the Atmospheric Sciences Link Copied to Clipboard. Goddard Award – Leadership Link Copied to Clipboard. Goddard Award – Science Link Copied to Clipboard.\n",
      "==================================\u001b[1m Ai Message \u001b[0m==================================\n",
      "Tool Calls:\n",
      "  fetch_webpage (call_ChXYwXUZP33u1A09nnnFG4vb)\n",
      " Call ID: call_ChXYwXUZP33u1A09nnnFG4vb\n",
      "  Args:\n",
      "    url: https://academic.oup.com/mnras/article-abstract/517/1/294/6677408\n",
      "    max_chars: 3000\n",
      "=================================\u001b[1m Tool Message \u001b[0m=================================\n",
      "Name: fetch_webpage\n",
      "\n",
      "Failed to fetch the webpage. Error: 403 Client Error: Forbidden for url: https://academic.oup.com/mnras/article-abstract/517/1/294/6677408.\n",
      "==================================\u001b[1m Ai Message \u001b[0m==================================\n",
      "Tool Calls:\n",
      "  fetch_webpage (call_AgAhFkVz8dxx18m9aWRADGL0)\n",
      " Call ID: call_AgAhFkVz8dxx18m9aWRADGL0\n",
      "  Args:\n",
      "    url: https://ntrs.nasa.gov/api/citations/20230002172/downloads/Arendt%20-%20The%20population%20of%20Galactic%20centre%20filaments%20III.pdf\n",
      "    max_chars: 3000\n",
      "=================================\u001b[1m Tool Message \u001b[0m=================================\n",
      "Name: fetch_webpage\n",
      "\n",
      "%PDF-1.4\n",
      "%����\n",
      "2 0 obj\n",
      "<\n",
      ">stream\n",
      "x��y`\u0013��?>3{gw��4W�4GӴ@\n",
      "��\u001c�Jé�}�\u0016�\u00149\u0005��\u001b�\u001e\u001c\u0016T�P��[T$@�r�\u0015\u0015�DP�AAEE\u0011EET����$)��y�?���l�������>�<�� �\u0010��zġ��+�L��͎Q(y\u0001���c��\u001d\\��;\u0010\\*� $^9a��+�<��m\u0017!��'N�nBlջ�wA�vܤ�c�}7��\u0017B�\u001e�gt�\u0004\u0005���\\*8�\u0010��']1��F� ;���P����Ǝ\u0011��?F�\"8�rŘk��nQ� �\u001d\u0007���g���غ�?��i�Pƥ6\u001f��F��p\u0006T�N�۟w�\u0013 uE7�\u0004 �\u001a��\u000bs��k�y\\\u0003��{��͝�u~�P,T ���B�p����p\\� i��NN�NΝ��Z�۹ �<�\n",
      "��\u0001g 7�;0 pq�&pI�607�\u0018�\u0011�\u001b�\u0017�%p<�\u001c4�y��`�`Y�k�[�W�����u�y�{�/\u0005���PV�\u001d�\u000b\u0015�څ\u0006���.\n",
      "�\u001a�/�l\u001e�\u0013�<{�3/'/��:/�waޘ��a\u00126���\u00161#��'��G�\"e����H}��ȢȒ�=��\"/D�E�D�EvDލ�|\u0016���� VУ��`l���)\u0005ӊ�s?\u0013zf����x�}�ө�S�Nu?���+�\u0012�Ǟ>~���og��9�\u001cn��<��t�t\"A��V\u0012DBd$y���\u0006s�s�\u0002������p�V~��A�\u0011� �\u000bw {�c\"\u0012��`y�|(\u0017\u0001����m�#�����'\u0010\f\\\u0018\u0018����@}`c`g���\u0017��\u0002'�(h\u0007޵ �\u0004˃\u0015�w����wK�+S�s�x7 4,t �ni\u000b�l���<�wuy�\u0018���w�[x�4�2�\\\u000b��\u0006�}\n",
      "���»�\u0005��wu�;�3��w�T֩�S]�w�S=O�9�����;S�xw�9�<����zʻ����\u0001B����5\\_N��\u001e�ڀ�}d�k���e7B���~\u0007���� �\\_�\\_\u0006 td\u001eB��p�:b;b\u001cяhG�#�#�\u0011�xD8�\u001d!GЏt����l\u000f�q�&3z�O�T}���\\�\u000eL>p��-Gv}St��#\u000f\u001cx���/�}����\u0011��iZ����\u0019\\_����\\_ƾ,�2����W�/��i����[������؏����Ⱦ�����5�k��}/�{i�&�^��Ծ5�z�뱯���}y�B�������z��\u0010r�\u0007u|XzHzPZ�|O���Z�>�h�\u0000�\u00107�j0\u0006\u0004�\u0003�.����\u0007�����\u0000�\u0014[�6�,�d������L��H�%jNj�B��z�\n",
      "�nQw��MV��!;~������ޓtSO�����[:��=�\u0007-�{���6�LM�\u001e\u0000Fj�t��G���\u0005�QTE��l�2u�CO�[�|�R�\f}�\u0016�;�bZ��D&j\u0000�ނ�A�Я�vt?Z�1ڏ~A������7t\u001c=\u000eH�&ډV���X�\u0014�Co���\n",
      "�\u0016z\u000f���E���h\u0002z\u001f�F{Ћh\"�\u0019݅>D\u001f��h\u0012�\u0001\u001dA����r4\u0005]���+�J4\n",
      "�@��L4\u000b]�f���5�0�\u0016]��C7��h\u000eڄ\u001eC�Ѝ`onB?���f�\fߏ �0�\u0005t\n",
      "��\u000f���A�\u0010:����%,�\u0004~\u0018?�\u001f�+�J�\u0018V�\u0005�X�'� �'~\u0012?�����g�\\*�\u001c~\u001e��W�\u0017�\u001a\u001c�k�:�\u001e��>�\n",
      "x1n�\u001b�F� 7a\u001d[�f�\u0005\u001b��6lG\u0007�W8\u000b;�V�\n",
      ";�\u000b/���/��v�\n",
      "~\u0015��\u0007�Aq��s�kx\u0007�b\u001f��~�:މ�F'���\u001b\u001c�A\u001c�y�\n",
      "�&~\u000b������]�=\u001c��8�\u000b�n�\u0007��?�{�h\u000b.ĭpk�\u0006\u001dB���\u0006q��D�]�C�S\\\\*�%�-�#�+�'.\u0013�\u0017\u001f\u0010�����i�!�a�\u0011�Qq��R|L|\\|B|R|J|Z|���O\u0011�\u0015W�ωϋ/���\u0017�5b\\\\+�\u0013׋��T�\n",
      "q��Q�$6���-�Vq��/�%�eq��������C|]�)�!�)�%�-�#�+�\u0012�\u0013w�{���\u0019��H�\u0002\u00118�\u0017\u0004A\u0014$A\u0016\u0014�\"����\u0007�^�#�c�\u0013�S�3�sq��\\_�B�R< \u001e\u0014�\u0012�\u0016�\u0011\u000f�ߊ߉ߋ��\u001f�\u001f�#�O�Q�g�1�\u0004�?ß�}�]2%�d��$��\\�[�HْW�I��\\_\n",
      "HA)$�Y�Cu��z@=�~�~�~�\u001eR�U�S�W\u000fkh'�?�������)��vFk�\u0012:ұN���/E�\u0002�Pj%���HQ!��T�T/�$�,�\"\\*͗\u0016H\u000b�E�mR��XZ\"�.�!�)-������C\u001f�\u0003�2��t��\u001c��!@�G�G�\u0015�J�1�q� �I�\u0019�\u001c�C\\_�O�\u0017�S���3ҳ�\\*�9�y�\u0005i����F�Kk�u�z�Qڠz�l5G��>5W��\u00015���<5��\u0011�@-T[�w���=���r/���G���-\\_(��/�����\u0001�@y�\n",
      "y�|����\\~P~H~؈\u0019ݍ\u001eFO�\u0011�Q���[^�s:�\u000b�o����c�W�\u001f����� �'�\u000b��O�(>�O�3�\u0019\u0017�\u0017&�p�'\u0002\u0011�Dd�\u0010\u000bnKT�\u0011�X�ALb#v�E\u001c�\u001dq\u0012\u0017.�퉛xH6�!^�#��O\u0002��-\u0001�$\u000fw�%$�KI>��\u0002RHZ�֤\n",
      "�\u001a}�2���Oݯ~�����\u001eS�� E�-iG�I{ҁ��RRF:�N�����\u0016���@搹�F2�ԓ����\u0016r+�/�M\u0016���;��.�=y��G~\\_�@�+($,\"\\*&.���\\_�\\_�\u0007��W���7�!�[�;�{����|D�I>\\*�,�\"\u001f���{ȿ���?�\u0013��\\_���I���&��M>%�\u0016�B�|Fn\u0016\u001c�Sp n9� \u0005+D�\n",
      "�������Ȋ�X\u0014Uф\u001c�+��\\�/\u0004\u0014]�\\*�b\n",
      "A!$� aŦؕ,š8\u0015��V\n",
      "�>�>�>i\\hN��f�a}ݺ����7�I}Z}J}F}V]�>�>�� �\u001ayF��7\"F\u00019j�WO���B�\u000bw\u0013w\u000b7�[�����������-�\u001e���)n\u0015�<��[í�6p��q�p�soq���Ql�\u001a����1�}�#�3�\u000b�+�[�\u0007�(�\u000b�+���F~'��\u001f�\u0004���%t\u0011ʅ��j�Eu�zJ=��Q�Մ��J�79IN���\fi& \u000eq�#\u001c��&�\u0004��P$�'݄\u0018��C�%�\u0011.\u0004\u000fl�0\u0014b��\\@�T�L� L\u0016�\u0014f\n",
      "Ws���B�p�p��@X\u0004\u0011�\u0012�\u000ea�p�p��\f��\u0007�����RxRx\u0016<���(l\u0014[���� �zG�-�ϵ\u0015�\n",
      "��\u0003�7\\\u0007�{�����pJH��(��h�61KtsG�l1W\f�A1$��b��Jl#\u0016����\\G�D,\u0013��]�nbL�!��d���G�@�P�+^$�\u0013��\u0003ā� q�8D\u001c\\*\u000e\u0013��#�\\*�b�Z�\u0011G��x��V]��\u000fg�TNK�G\u001c%։��I���\u0006�h�fѬ�]si9�\\_��\n",
      "�VZ\u001b�Hk��i]�\n",
      "-���.��k���Z�6J���i����T�\u0001�7��G�G��Z��a��z��\u0006�d(�n�\f��9\n",
      "==================================\u001b[1m Ai Message \u001b[0m==================================\n",
      "Tool Calls:\n",
      "  download_and_read_pdf (call_Q360v8eddVGLsJft0pvwCPV3)\n",
      " Call ID: call_Q360v8eddVGLsJft0pvwCPV3\n",
      "  Args:\n",
      "    url: https://ntrs.nasa.gov/api/citations/20230002172/downloads/Arendt%20-%20The%20population%20of%20Galactic%20centre%20filaments%20III.pdf\n",
      "=================================\u001b[1m Tool Message \u001b[0m=================================\n",
      "Name: download_and_read_pdf\n",
      "\n",
      "MNRAS  517, 294–355  (2022) https://doi.org/10.1093/mnras/stac2415  \n",
      "Advance  Access  publication  2022 August  26 \n",
      "The  population  of Galactic  Centre  filaments  – III.  Candidate  radio  and  \n",
      "stellar  sources  \n",
      "F. Yusef-Zadeh  , 1 ‹R. G. Arendt  , 2 ‹M. Wardle,  3 I. Heywood  4 , 5 and W. Cotton  6 \n",
      "1 Department  of Physics and Astronomy,  Northwestern  University,  Evanston,  IL 60208, USA \n",
      "2 NASA/GSFC,  Code 665, 8800 Greenbelt  Road, Greenbelt,  MD 20771, UMBC/CRESST  2, USA \n",
      "3 School of Mathematical  and Physical  Sciences,  Research  Centre for Astronomy,  Astrophysics  and Astrophotonics,  Macquarie  University,  Sydney, NSW 2109, \n",
      "Australia  \n",
      "4 Royal Astronomical  Society, Burlington  House, Piccadilly,  London W1J 0BQ, UK \n",
      "5 Department  of Physics and Electronics,  Rhodes University,  PO Box 94, Makhanda,  6140, South Africa \n",
      "6 National  Radio Astronomy  Observatory,  Charlottesville,  VA 60208, USA \n",
      "Accepted  2022 August 22. Received  2022 August 22; in original form 2022 June 27 \n",
      "A B S T R A C T \n",
      "Recent  MeerKAT  radio continuum  observations  of the Galactic  Centre  at 20 cm show a large population  of non-thermal  radio \n",
      "filaments  (NRFs)  in the inner few hundred  pc of the Galaxy.  We have selected  a sample  of 57 radio sources,  mainly  compact  \n",
      "objects,  in the MeerKAT  mosaic  image that appear  to be associated  with NRFs.  The selected  sources  are about four times the \n",
      "number  of radio point sources  associated  with filaments  than would be expected  by random  chance.  Furthermore,  an apparent  \n",
      "correlation  between  bright IR stars and NRFs is inferred  from their similar  latitude  distributions,  suggesting  that they both \n",
      "co-exist  within  the same region.  To examine  if compact  radio sources  are related  to compact  IR sources,  we have used archi v al \n",
      "2MASS,  and Spitzer  data to make spectral  energy  distribution  of individual  stellar sources  coincident  or close to radio sources.  \n",
      "We provide  a catalogue  of radio and IR sources  for future detailed  observations  to investigate  a potential  three-way  physical  \n",
      "association  between  NRFs,  compact  radio and IR stellar sources.  This association  is suggested  by models  in which NRFs are \n",
      "cometary  tails produced  by the interaction  of a large-scale  nuclear  outflow  with stellar wind bubbles  in the Galactic  Centre.  \n",
      "Key words:  plasmas  – radiation  mechanisms:  non-thermal  – cosmic  rays –I S M :  magnetic  fields. \n",
      "1 INTRODUCTION  \n",
      "The inner thousand  light years of our Galaxy  are unusually  active \n",
      "compared  to the solar neighbourhood.  The dormant  supermassive  \n",
      "black hole, Sgr A ∗, is surrounded  by hot young stars with powerful  \n",
      "winds that irradiate  the nearby interstellar  gas. There is circumstantial  \n",
      "evidence  of two past explosions  within the last few million  years. \n",
      "First, vast γ-ray emitting  lobes filled with relativistic  particles  emerge  \n",
      "from the Galactic  Centre and extend above and below the disc \n",
      "for many thousands  of light years (i.e. the Fermi bubble)  (Su, \n",
      "Slatyer  & Finkbeiner  2010 ). Secondly,  a bipolar  MeerKAT  radio \n",
      "bubble filled by coronal  X-ray emitting  hot gas emerges  from the \n",
      "inner few hundred  light years of the Galaxy.  The radio bubble hosts \n",
      "hundreds  of mysterious  magnetized  radio filaments.  The filaments  \n",
      "occupy  a region consistent  with the radio bubble,  suggesting  a causal \n",
      "association  between  the tw o (Heyw ood et al. 2019 ). Theoretical  \n",
      "models  explain  the Fermi and MeerKAT  bubbles  by invoking  a \n",
      "correlated  burst of supernovae  or an explosion  from the black \n",
      "hole as it swallows  a cloud (Crocker  & Aharonian  2011 ; Yang, \n",
      "Ruszkowski  & Zweibel  2013 ). \n",
      "The high cosmic-ray  flux in the Galactic  Centre suggests  a global \n",
      "injection  of relativistic  particles  produced  as a result of a relic \n",
      "of a past activity  by Sgr A ∗, due to starburst  activity  producing  \n",
      "⋆ E-mail: zadeh@northwestern.edu  (FY-Z); richard.g.arendt@nasa.gov  \n",
      "(RGA) multiple  superno  va e xplosions  (He ywood et al. 2019 ). One scenario  \n",
      "for the origin of the bipolar  radio/X-ray  features  considers  high \n",
      "cosmic-ray  pressure  driving  large-scale  winds and expanding  the \n",
      "medium  away from the Galactic  plane. It has become  apparent  that \n",
      "relativistic  particles  permeate  the central molecular  zone (CMZ)  at \n",
      "levels a thousand  times that in the solar neighborhood  (Geballe  et al. \n",
      "1999 ; Oka et al. 2005 , 2019 ; Indriolo  & McCall  2012 ; Goto et al. \n",
      "2014 ; Le Petit et al. 2016 ). This provides  a significant  source of \n",
      "pressure  when compared  to thermal  pressure  in interstellar  space at \n",
      "the Galactic  Centre.  High cosmic-ray  pressure  drives winds in the \n",
      "nuclei of galaxies  (Everett  et al. 2008 ; Everett,  Schiller  & Zweibel  \n",
      "2010 ; Ruszkowski,  Yang & Zweibel  2017 ; Zweibel  2017 ) and play \n",
      "a role in feedback,  limiting  star formation  and the growth  of the \n",
      "central supermassive  black holes by transferring  their momentum  \n",
      "and energy into the surrounding  medium.  The high-cosmic-ray  \n",
      "flux in the Galactic  Centre suggests  global injection  of relativistic  \n",
      "particles  produced  as a result of a relic of a past activity  by Sgr A ∗, \n",
      "due to starburst  activity  producing  multiple  supernovae  explosions  \n",
      "(Heywood  et al. 2019 ). The interaction  of a cosmic-ray-driven  wind \n",
      "with stellar wind bubbles  create cometary  tails that could explain  the \n",
      "energetic  non-thermal  radio filaments  (NRFs)  found throughout  the \n",
      "Galactic  Centre (Shore & LaRosa  1999 ; Bicknell  & Li 2001 ; Yusef- \n",
      "Zadeh & Wardle  2019 ; Yusef-Zadeh  et al. 2022a ). In this scenario,  a \n",
      "compact  radio source is expected  to be located  along or at one end \n",
      "of NRFs where the wind interacts  with mass-losing  stars. \n",
      "Another  model suggests  a time-dependent  injecting  source,  such \n",
      "as a pulsar, crossing  spatially  intermittent  magnetic  bundles  and \n",
      "© 2022 The Author(s)  \n",
      "Published  by Oxford University  Press on behalf of Royal Astronomical  Society Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  295 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 1. A 1.28-GHz  mosaic image of the Galactic  Centre region (Heywood  et al. 2022 ). Red circles show the positions  of compact  radio sources with possible  \n",
      "infrared  counterparts.  A blow-up  of these sources is shown in Fig. 3 (a)–(eee).  There are 57 circles corresponding  to the number of sources listed in Table 1 . \n",
      "generates  filamentary  structure  (Thomas,  Pfrommer  & Enßlin 2020 ). \n",
      "This model predicts  the filaments  run perpendicular  to the direction  \n",
      "of the motion,  unlike cometary  tail model in which compact  sources  \n",
      "inject particles  in the direction  opposite  to their motion  (Shore & \n",
      "LaRosa  1999 ; Yusef-Zadeh  & Wardle  2019 ). \n",
      "Moti v ated by these scenarios,  we provide  a list of candidate  radio \n",
      "and stellar sources  that appear to be associated  with NRFs. Candidate  \n",
      "radio sources  are selected  only from 43 groupings  of 174 magnetized  \n",
      "filaments  that run parallel  to each other and are separated  from each \n",
      "other by mean spacing  of 16 arcsec (0.64 pc) (Yusef-Zadeh  et al. \n",
      "2022 ). We selected  the most spectacular  groupings  of filaments  \n",
      "ranging  from pairs to dozens in Yusef-Zadeh  et al. ( 2022a ). We \n",
      "defined  a group of filaments  or a grouping  having similar orientations,  \n",
      "similar curvature  or bending,  spatially  close to each other, and in \n",
      "some cases, converge  to a point, shift sideways  together,  and change  \n",
      "direction  coherently.  These characteristics  implied  that they are parts \n",
      "of the same system of filaments  with a common  origin. We a v oided \n",
      "groups of filaments  close to the Galactic  plane because  of confusing  \n",
      "thermal  sources  in this region. The groups of filaments  we selected  are \n",
      "distinct  from single filaments  along the line of sight at large physical  \n",
      "distances  from each other but which appear close in projection.  The \n",
      "analysis  of single filaments,  which may comprise  multiple  filaments  \n",
      "in future higher resolution  observations,  are postponed  elsewhere.  \n",
      "We present  radio and infrared  images  of each of the 57 candidate  \n",
      "sources  as well as the spectral  energy distributions  (SEDs)  of their \n",
      "IR counterparts.  The purpose  of this list is to moti v ate further high \n",
      "sensitivity,  spatial resolution  radio and IR observations  to examine  a \n",
      "three-way  spatial correlation  and to test the cometary  tail model of \n",
      "the origin of the filaments.  2 MEERKAT  AND VLA OBSERV ATIONS  \n",
      "Details  of MeerKAT  and the Karl G. Jansky Very Large Array \n",
      "(VLA) observations  are found in Heywood  et al. ( 2019 ), Heywood  \n",
      "et al. ( 2022 ). Figure 1 shows a mosaic  MeerKAT  image of the \n",
      "inner few degrees  of the Galactic  Centre (1 arcmin corresponds  \n",
      "to 2.4 pc at the Galactic  Centre distance)  with an FWHM  ∼4 \n",
      "arcsec resolution.  VLA obtained  higher resolution  images  of the \n",
      "region tow ard tw o fields centred  on Sgr C and Sgr A. Thus, only \n",
      "few sources  have been covered  with these limited  high-resolution  \n",
      "measurements.  VLA observation  was carried out at L band (1–2 GHz) \n",
      "with the array in the most extended  A-configuration  and centred  at \n",
      "Sgr C (J2000 17 h 44 m 35 . s 0, −29 ◦29 /prime 00 . /prime/prime 0). The initial flagging  and \n",
      "reference  calibration  was performed  using the VLA casa pipeline8  \n",
      "and processed  with the wsclean  multiscale  clean algorithm  with a \n",
      "resolution  of ∼1 arcsec. Further  details on the use of the VLA data \n",
      "can be found in Heywood  et al. ( 2022 ). \n",
      "3 RESULTS  AND DISCUSSION  \n",
      "3.1 Stellar  and NRF distributions  \n",
      "To examine  the possible  association  of NRFs and stellar sources  on \n",
      "a much larger scale, Figure 2(left- and right-hand  panels)  display  \n",
      "the histograms  of magnetized  filaments  longer than 66 arcsec and of \n",
      "bright IR stars (Ramirez  et al. 2008 ), as traced in four Spitzer bands, \n",
      "as a function  of Galactic  latitudes.  Bright stars are selected  mainly to \n",
      "a v oid confusion  limits and minimize  extinction.  A length of 66 arcsec \n",
      "is selected  because  it is a good constraint  for the exclusion  of thermal  Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "296 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) Figure 2. Left-hand  panel: A histogram  of the number of NRFs is shown as a function  of Galactic  latitude for filament  lengths longer than 66 arcsec and \n",
      "confined  to within Galactic  latitude | l | < 0.45 ◦(which excludes  most of the thermal features),  based on MeerKAT  data. Right-hand  panel: A histogram  of the \n",
      "number of bright stars is shown as a function  of Galactic  latitude using IRAC data in all four bands 3.6, 4.5, 5.8, and 8 μm (black, blue, green, and red). This \n",
      "plot is constructed  from data presented  in (Ramirez  et al. 2008 ). \n",
      "filamentary  features  versus NRFs (Yusef-Zadeh  et al. 2022a ). The \n",
      "similar scale heights  of the filaments  and stellar sources  indicate  their \n",
      "Galactic  population  is suggestive  of a spatial correlation  between  the \n",
      "two. In addition,  both distributions  tend to peak at slightly  ne gativ e \n",
      "latitudes,  suggesting  that both populations  are possibly  associated  \n",
      "with each other. Ho we ver, robust statistical  tests, e.g. the two-sample  \n",
      "KS test, simply confirm  that there are real differences  between  \n",
      "the distributions  in Fig. 2 (left- and right-hand  panels).  That’s no \n",
      "surprise  given that some fraction  of the stars will be ‘local’ while \n",
      "all of the NRFs are at the Galactic  Centre.  To investigate  this weak \n",
      "correlation  further,  we will next examine  the spatial relationship  \n",
      "between  compact  radio and IR stellar sources  as well as the SEDs of \n",
      "IR sources  closest to the peaks of compact  radio sources.  \n",
      "3.2 Compact  radio and stellar sources  \n",
      "We have identified  a sample  of small resolved  radio sources  and \n",
      "unresolved  compact  radio sources  in the MeerKAT  image of the \n",
      "Galactic  Centre that appear to coincide  with one end of or along \n",
      "NRFs. Examination  of modified  images  in which all radio sources  \n",
      "are artificially  shifted with respect  to the filaments,  1 indicates  that \n",
      "only ∼1/4 of the apparent  correlations  between  NRFs and compact  \n",
      "sources  can be attributed  to random  coincidence.  A single filament  \n",
      "with a length of 50pc and width of 0.1pc has a volume  of ∼0.5 pc 3 \n",
      "whereas  the large-scale  volume  of the CMZ is roughly  ∼10 7 pc 3 , \n",
      "giving a small filling factor of ∼10 −3 for NRFs. Given the large \n",
      "stellar density  of ∼10 6 IR sources  in the CMZ (Ramirez  et al. 2008 ), \n",
      "the average  number  of stellar sources  within the volume  of a filament  \n",
      "is expected  to be about one. In addition,  there are many striking  \n",
      "examples  of a single filament  splitting  into two prongs at a junction  \n",
      "where a resolved  compact  radio source is located,  suggestive  of a \n",
      "flow of plasma  along the filaments  (Yusef-Zadeh  et al. 2022 ). So, it \n",
      "is possible  that a source,  such as a mass-losing  star, is responsible  \n",
      "for injecting  cosmic-ray  particles  into the filaments  or acting as \n",
      "an obstacle  redirecting  plasma  flow (Yusef-Zadeh  & Wardle  2019 ; \n",
      "Yusef-Zadeh  et al. 2022a ). In this picture,  mass-losing  stars such as \n",
      "the red giants are impacted  by a cosmic-ray-driven  wind, creating  \n",
      "shocks at the wind–wind  interface,  and pushing  the shocked  stellar \n",
      "1 Tests were done by shifting point-source  images by ±64 pixels (69 arcsec) in \n",
      "l and b (four independent  tests), and then recounting  the number of apparent  \n",
      "associations  between  the shifted sources and the NRFs. wind into the tail, thus providing  additional  acceleration  of cosmic-  \n",
      "ray particles  (Yusef-Zadeh  & Wardle  2019 ). \n",
      "Not all filaments  show a compact  radio and infrared  candidate  \n",
      "source at their ends, so the cometary  model may not be viable to \n",
      "explain  their origin. It is possible  that this subclass  of filaments  \n",
      "could be explained  by an alternative  model in which a compact  \n",
      "radio source,  like a pulsar crossing  the filaments,  injects cosmic-ray  \n",
      "particles  intermittently.  In this picture,  no IR counterpart  is expected.  \n",
      "A more detailed  study of radio compact  sources  near this subclass  of \n",
      "filaments  need to be done in the future to test this model. \n",
      "To examine  whether  compact  radio sources  have stellar counter-  \n",
      "parts, we have used archi v al 2MASS  and Spitzer (IRAC and MIPS) \n",
      "data to examine  the SED of individual  stellar sources  that lie closest to \n",
      "the compact  radio sources.  We have also checked  the Gaia catalogue  \n",
      "and found 14 of the GALCEN  sources  have Gaia counterparts  within \n",
      "2 arcsec, 12 of which are within 1 arcsec. These sources  are likely \n",
      "foreground  sources.  In some cases there may be coincidences  of \n",
      "unrelated  sources.  Table 1 tabulates  the source number,  the positions  \n",
      "of radio and infrared  sources  in Galactic  coordinates  ( l , b ), the Gaia \n",
      "source offset, the separation  between  the peak radio and infrared  \n",
      "sources  and the peak intensity  of the compact  radio sources  in \n",
      "Columns  1–6, respectively.  The mosaic  image was blanked  below \n",
      "10 μJy beam −1 (Heywood  et al. 2022 ). The last column  shows the \n",
      "associated  groups of filaments  named in Yusef-Zadeh  et al. ( 2022 ). \n",
      "We present  the SEDs of the nearest  GALCEN  catalogue  (Ramirez  \n",
      "et al. 2008 ) IR source to each of the radio sources,  as listed in \n",
      "Table 1 . Figure 3 shows a close up view of all 57 compact  sources  \n",
      "at radio and IR and 1.25–24  μm (2MASS  + IRAC + MIPS) SEDs \n",
      "of stellar sources.  In spite of the lack of full wav elength  co v erage, \n",
      "Figure 3 shows a sample  illustrating  two different  classes of SEDs \n",
      "that may be IR counterparts  to compact  radio sources.  In one class, \n",
      "“Red Giant SEDs”,  show a distinct  dip at 4.5 μm (see Fig. 3j, t, \n",
      "v, oo and ww). The dip is caused by the CO fundamental,  which \n",
      "strengthens  with decreasing  temperature  (Engelke,  Price & Kraemer  \n",
      "2006 ). These sources  are not detected  at 24 μm. Other sources  do \n",
      "exhibit 24 μm emission.  A subset of these are visibly extended  in \n",
      "the images  and have very red SEDs (Fig. 3u, y, cc). Thus, these \n",
      "are likely young stellar objects  and H II regions.  The remainder  \n",
      "(Figs 3b, c, bb, and vv) are point-like  and typically  brighter  than \n",
      "those sources  with red giant SEDs. These stars may be dusty AGB \n",
      "stars or supergiants  (Reiter et al. 2015 ). It is possible  that all red \n",
      "giants have 24 μm emission,  but that it is only sufficiently  bright \n",
      "to be detected  by MIPS in these brighter  classes of giants. At Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  297 \n",
      "MNRAS  517, 294–355  (2022) Table 1. The positions  of infrared  sources closest to compact  radio sources,  both of which could be associated  with groups of NRFs. \n",
      "Radio selection  Closest IRAC/MIPS  counterpart  Gaia source offset Radio-IR  source offset 20-cm peak intensity  Associated  filament \n",
      "Number l ( o ) b ( o ) l ( o ) b ( o ) (arcsec) (arcsec) (mJy beam −1 ) (group names) \n",
      "1 358.732 69 −0 .228 66 358.73234  −0 .231 10 – 8.9 0.20 Horseshoe  \n",
      "2 358.775 59 0 .453 83 358.77572  0 .454 70 0.2 3.2 0.03 Pelican \n",
      "3 358.794 24 0 .479 40 358.79452  0 .478 98 0.4 1.8 0.07 ”\n",
      "4 358.798 32 0 .467 80 358.79845  0 .468 30 0.3 1.9 0.14 ”\n",
      "5 359.072 50 0 .735 48 359.07256  0 .735 26 – 0.8 3.30 Arrow \n",
      "6 359.119 72 −0 .264 79 359.119 41 −0 .265 23 – 1.9 0.57 Snake \n",
      "7 359.132 42 −0 .200 15 359.133 14 −0 .200 11 – 2.6 1.10 Snake \n",
      "8 359.211 00 −0 .085 00 359.210 45 −0 .085 41 – 2.5 0.73 Candle \n",
      "9 359.213 88 −0 .099 65 359.213 27 −0 .099 90 – 2.4 0.35 ”\n",
      "10 359.223 89 −0 .136 10 359.224 42 −0 .136 03 – 1.9 0.20 ”\n",
      "11 359.320 86 −0 .157 80 359.319 64 −0 .158 09 – 4.5 1.40 Hummingbird  \n",
      "12 359.321 39 −0 .430 90 359.322 11 −0 .430 14 – 3.8 0.26 Sausage \n",
      "13 359.323 53 −0 .429 52 359.324 78 −0 .430 18 0.9 5.1 0.24 ”\n",
      "14 359.343 95 −0 .41610 359.344 18 −0 .416 06 – 0.8 0.60 ”\n",
      "15 359.400 80 −0 .23797 359.400 81 −0 .238 23 – 1.0 0.01 ”\n",
      "16 359.404 94 −0 .25914 359.404 20 −0 .259 35 0.2 2.8 1.41 ”\n",
      "17 359.415 75 −0 .70605 359.414 95 −0 .705 14 – 4.4 0.20 Feather \n",
      "18 359.418 79 −0 .58340 359.419 81 −0 .582 45 – 5.0 0.04 ”\n",
      "19 359.420 24 −0 .66049 359.421 36 −0 .661 62 – 5.7 0.07 ”\n",
      "20 359.422 88 −0 .664 01 359.423 66 −0 .662 69 – 5.5 0.41 ”\n",
      "21 359.438 56 0 .005 95 359.438 54 0 .005 41 1.1 1.9 6.2 Sgr C \n",
      "22 359.435 80 0 .112 72 359.438 67 0 .102 42 – 2.9 0.01 French Knife \n",
      "23 359.452 86 −0 .033 31 359.45338  −0 .033 38 – 1.9 7.20 Sgr C \n",
      "24 359.454 76 −0 .054 66 359.454 23 −0 .05583 – 4.6 1.01 ”\n",
      "25 359.466 64 −0 .170 61 359.467 23 −0 .171 26 – 3.2 3.3 ”\n",
      "26 359.475 31 0 .126 72 359.475 24 0 .125 99 – 2.6 0.20 Bent Harp \n",
      "27 359.484 96 0 .120 52 359.485 14 0 .119 40 – 4.1 1.80 ”\n",
      "28 359.491 36 0 .121 41 359.490 23 0 .120 55 – 5.1 0.01 ”\n",
      "29 359.521 74 0 .096 97 359.521 46 0 .096 94 – 1.0 0.90 ”\n",
      "30 359.557 50 0 .140 30 359.557 49 0 .138 49 – 6.5 0.50 Ripple \n",
      "31 359.729 20 −0 .811 39 359.729 07 −0 .810 95 – 1.6 0.32 Knot \n",
      "32 359.733 72 −0 .801 22 359.734 48 −0 .801 24 – 2.7 0.12 ”\n",
      "33 359.745 74 −0 .783 98 359.744 98 −0 .784 00 – 2.8 1.00 ”\n",
      "34 359.748 29 −0 .780 59 359.747 85 −0 .781 35 0.3 3.2 0.54 ”\n",
      "35 359.757 64 0 .197 74 359.757 82 0 .197 85 0.4 0.8 0.78 Flamingo  \n",
      "36 359.807 90 0 .116 85 359.807 73 0 .117 74 – 3.3 0.31 ”\n",
      "37 359.838 44 0 .375 14 359.839 13 0 .374 89 – 2.7 0.18 Harp \n",
      "38 359.841 21 0 .365 71 359.840 95 0 .366 49 – 3.0 0.17 ”\n",
      "39 359.844 11 0 .35841 359.843 24 0 .359 19 – 4.2 0.23 ”\n",
      "40 359.848 14 0 .29777 359.84743  0 .29717 – 3.3 0.04 ”\n",
      "41 359.850 73 0 .29309 359.84970  0 .29374 – 4.4 0.20 ”\n",
      "42 359.873 91 −0 .264 30 359.874 04 −0 .26446 – 0.7 1.30 Cleaver Knife \n",
      "43 359.885 80 0 .101 55 359.885 86 0 .10101 – 2.0 5.60 Harp \n",
      "44 359.921 07 −0 .282 71 359.920 23 −0 .283 00 – 3.2 0.27 Cleaver Knife \n",
      "45 359.949 84 −0 .259 40 359.950 39 −0 .259 39 – 2.0 0.73 ”\n",
      "46 359.992 38 −0 .554 65 359.992 60 −0 .553 14 0.3 5.5 0.08 Comet \n",
      "47 0.023 33 −0 .625 74 0.023 43 −0 .626 68 0.05 3.4 0.07 ”\n",
      "48 0.029 52 −0 .659 87 0.029 86 −0 .660 06 1.9 1.4 0.29 ”\n",
      "49 0.034 34 −0 .661 20 0.034 86 −0 .660 23 – 4.0 0.28 ”\n",
      "50 0.045 31 −0 .667 85 0.045 83 −0 .666 76 – 4.4 0.25 ”\n",
      "51 0.224 57 0 .842 97 0.224 19 0 .843 30 0.8 (1.3) 1.8 0.04 Space Shuttle \n",
      "52 0.226 05 0 .791 78 0.226 51 0 .791 85 – 1.7 0.25 ”\n",
      "53 0.264 08 −0 .196 83 0.263 61 −0 .197 78 – 3.8 0.64 Porcupine  \n",
      "54 0.420 08 −0 .296 63 0.419 70 −0 .297 18 – 2.4 0.41 Contrail \n",
      "55 0.420 75 −0 .290 48 0.420 99 −0 .289 75 – 2.8 0.39 ”\n",
      "56 0.442 69 −0 .324 07 0.442 44 −0 .323 76 0.4 1.4 0.17 ”\n",
      "57 0.696 88 0 .148 27 0.697 03 0 .148 05 0.8 1.0 0.75 Bent Fork \n",
      "the shorter wavelengths,  the IRAC data are confusion  limited.  So \n",
      "spatial coincidence  of IR and radio sources  at these wavelengths  \n",
      "alone is not a strong indication  of the actual association  of the \n",
      "sources.  Some images  appear to show cometary  tails traced by the \n",
      "filaments  (see Fig. 3i). An interpretation  of the head–tail  mor- \n",
      "phology  is that the streaming  of cosmic  rays from the Galactic  \n",
      "plane interacts  with 24 μm sources  and create a tail oriented  in Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "298 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( a ) Each candidate  radio source numbered  1–57 in Table 1 is shown in six separate  panels. The left-hand  column shows a narrow view of, a \n",
      "composite  narrow view of IRAC 3.6, 4.5, 8 μm images, and a narro w vie w of MIPS 24 μm images from top to bottom, respecti vely. The right-hand  column \n",
      "shows a wider view of MeerKAT  image images (filtered  to highlight  filaments,  (Yusef-Zadeh  et al. 2022a )), unfiltered  VLA image (if available),  and the SED \n",
      "using 2MASS  + IRAC + MIPS data, (black symbols:  Ramirez  et al. 2008 ; Hinz et al. 2009 ). Red stars indicate  measurements  from the GLIMPSE  II Spring \n",
      "’08 Archive  (Benjamin  et al. 2003 ). Error bars are plotted for all measurements,  but are generally  smaller than the symbol size. 2MASS  95 per cent confidence  \n",
      "upper limits are now plotted as downward  triangles.  Upper limits are not stated in the other catalogues  on a source-by-source  basis. Red circles in the image \n",
      "panels show the positions  where the IR SED, the nearest source to the peak radio position,  is measured.  \n",
      "the directions  away from the source motion  and the nuclear  \n",
      "wind. \n",
      "4 SUMMARY  \n",
      "This short paper provides  a catalogue  of the positions  of radio and \n",
      "infrared  sources  that lie along groups of NFRs that have recently  been \n",
      "identified.  These sources  could potentially  be associated  with radio \n",
      "filaments.  We have displayed  radio and IR images  of 57 filaments  as \n",
      "well as the SEDs of stellar sources  coincident  with radio sources.  A true physical  association  will support  a picture in which the compact  \n",
      "radio sources  are a byproduct  of a large-scale  nuclear  wind interacting  \n",
      "with stellar systems,  thus producing  cometary  tails that trace NRFs. \n",
      "Near- to mid- IR SEDs illustrate  the various  classes of sources  that \n",
      "appear to be associated  with the radio filaments.  A trend we notice is \n",
      "the SEDs of mass-losing  dusty red giants with strong 24 μm emission,  \n",
      "or in some cases young, massive  stars. Future high-resolution  radio \n",
      "and infrared  observations  of the candidate  sources  are needed  to \n",
      "establish  their true association  with NRFs, thus testing cometary  tail \n",
      "model. Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  299 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( b ) Same as Fig. 3(a) except for source 2 in Table 1 . Open circles indicate  potentially  saturated  IRAC measurements.  Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "300 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( c ) Same as Fig. 3(a) except for source 3 Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  301 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( d ) Same as Fig. 3(a) except for source 4 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "302 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( e ) Same as Fig. 3(a) except for source 5 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  303 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (f) Same as Fig. 3(a) except for source 6 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "304 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( g ) Same as Fig. 3(a) except for source 7 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  305 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( h ) Same as Fig. 3(a) except for source 8 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "306 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( i ) Same as Fig. 3(a) except for source 9 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  307 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( j ) Same as Fig. 3(a) except for source 10 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "308 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( k ) Same as Fig. 3(a) except source 11 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  309 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( l ) Same as Fig. 3(a) except source 12 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "310 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( m ) Same as Fig. 3(a) except source 13 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  311 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( n ) Same as Fig. 3(a) except source 14 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "312 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( o ) Same as Fig. 3(a) except source 15 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  313 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( p ) Same as Fig. 3(a) except source 16 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "314 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( q ) Same as Fig. 3(a) except source 17 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  315 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (r) Same as Fig. 3(a) except source 18 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "316 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( s ) Same as Fig. 3(a) except source 19 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  317 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( t ) Same as Fig. 3(a) except source 20 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "318 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( u ) Same as Fig. 3(a) except source 21 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  319 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( v ) Same as Fig. 3(a) except source 22 listed in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "320 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( w ) Same as Fig. 3(a) except source 23 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  321 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( x ) Same as Fig. 3(a) except source 24 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "322 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( y ) Same as Fig. 3(a) except source 25 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  323 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( z ) Same as Fig. 3(a) except source 26 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "324 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (aa) Same as Fig. 3(a) except source 27 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  325 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (bb) Same as Fig. 3(a) except source 28 in Table 1 . Open circles indicate  potentially  saturated  IRAC measurements.  Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "326 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (cc) Same as Fig. 3(a) except source 29 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  327 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (dd) Same as Fig. 3(a) except source 30 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "328 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (ee) Same as Fig. 3(a) except source 31 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  329 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. –( f f ) Same as Fig. 3(a) except source 32 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "330 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (gg) Same as Fig. 3(a) except source 33 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  331 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (hh) Same as Fig. 3(a) except source 34 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "332 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (ii) Same as Fig. 3(a) except source 35 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  333 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (jj) Same as Fig. 3(a) except source 36 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "334 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (kk) Same as Fig. 3(a) except source 37 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  335 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (ll) Same as Fig. 3(a) except source 38 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "336 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (mm) Same as Fig. 3(a) except source 39 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  337 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (nn) Same as Fig. 3(a) except source 40 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "338 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (oo) Same as Fig. 3(a) except source 41 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  339 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (pp) Same as Fig. 3(a) except source 42 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "340 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (qq) Same as Fig. 3(a) except source 43 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  341 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (rr) Same as Fig. 3(a) except source 44 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "342 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (ss) Same as Fig. 3(a) except source 45 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  343 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (tt) Same as Fig. 3(a) except source 46 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "344 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (uu) Same as Fig. 3(a) except source 47 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  345 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (vv) Same as Fig. 3(a) except source 48 in Table 1 . Open circles indicate  potentially  saturated  IRAC measurements.  Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "346 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (ww) Same as Fig. 3(a) except source 49 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  347 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (xx) Same as Fig. 3(a) except source 50 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "348 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (yy) Same as Fig. 3(a) except source 51 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  349 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (zz) Same as Fig. 3(a) except source 52 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "350 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (aaa) Same as Fig. 3(a) except source 53 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  351 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (bbb) Same as Fig. 3(a) except source 54 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "352 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (ccc) Same as Fig. 3(a) except source 55 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  353 \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (ddd) Same as Fig. 3(a) except source 56 in Table 1 . Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "354 F. Yusef-Zadeh  et al. \n",
      "MNRAS  517, 294–355  (2022) \n",
      "Figure 3. – (eee) Same as Fig. 3(a) except source 57 in Table 1. Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "\n",
      "Stellar  counterparts  355 \n",
      "MNRAS  517, 294–355  (2022) ACKNOWLEDGEMENTS  \n",
      "Work by RGA was supported  by NASA under award number  \n",
      "80GSFC21M0002.  FYZ is partially  supported  by the grant AST- \n",
      "0807400  from the National  Science  Foundation.  The MeerKAT  \n",
      "telescope  is operated  by the South African  Radio Astronomy  Ob- \n",
      "servatory,  which is a facility  of the National  Research  Foundation,  \n",
      "an agency  of the Department  of Science  and Innovation.  The National  \n",
      "Radio Astronomy  Observatory  is a facility  of the National  Science  \n",
      "Foundation  operated  under cooperative  agreement  by Associated  \n",
      "Universities,  Inc. \n",
      "DATA  A V AILABILITY  \n",
      "All the data including  VLA and MeerKAT  that we used here are \n",
      "available  online and are not proprietary.  We have reduced  and \n",
      "calibrated  these data and they are available  if requested.  \n",
      "REFERENCES  \n",
      "Benjamin  R. A. et al., 2003, Publ. Astron. Soc. Aust. , 115, 953 \n",
      "Bicknell  G. V. , Li J., 2001, ApJ , 548, L69 \n",
      "Crocker  R. M., Aharonian  F., 2011, Phys. Rev. Lett , 106, 101102 \n",
      "Engelke  C. W., Price S. D., Kraemer  K. E., 2006, AJ , 132, 1445 \n",
      "Everett J. E., Zweibel  E. G., Benjamin  R. A., McCammon  D., Rocks L., \n",
      "Gallagher  J. S., 2008, ApJ , 674, 258 \n",
      "Everett J. E., Schiller Q. G., Zweibel  E. G., 2010, ApJ , 711, 13 Geballe T. R., McCall B. J., Hinkle K. H., Oka T., 1999, ApJ , 510, 251 \n",
      "Goto M., Geballe T. R., Indriolo  N., Yusef-Zadeh  F., Usuda T., Henning  T., \n",
      "Oka T., 2014, ApJ , 786, 96 \n",
      "Heywood  I. et al., 2019, Nature , 573, 235 \n",
      "Heywood  I. et al., 2022, ApJ , 925, 165 \n",
      "Hinz J. L., Rieke G. H., Yusef-Zadeh  F., Hewitt J., Balog Z., Block M., 2009, \n",
      "ApJS , 181, 227 \n",
      "Indriolo  N., McCall B. J., 2012, ApJ , 745, 91 \n",
      "Le Petit F. et al., 2016, A&A , 585, A105 \n",
      "Oka T., Geballe T. R., Goto M., Usuda T., McCall B. J., 2005, ApJ , 632, 882 \n",
      "Oka T., Geballe T. R., Goto M., Usuda T., McCall B. J., 2019, ApJ, 883, 54 \n",
      "Ramirez  S. V. , Arendt R. G., Sellgren  K., Stolovy S. R., Cotera A., Smith H. \n",
      "A., Yusef-Zadeh  F., 2008, ApJS , 175, 147 \n",
      "Reiter M. et al., 2015, MNRAS  , 447, 3909 \n",
      "Ruszkowski  M., Yang H.-Y . K., Zweibel  E., 2017, ApJ , 834, 208 \n",
      "Shore S. N., LaRosa T. N., 1999, ApJ , 521, 587 \n",
      "Su M., Slatyer T. R., Finkbeiner  D. P., 2010, ApJ , 724, 1044 \n",
      "Thomas  T., Pfrommer  C., Enßlin T., 2020, ApJ , 890, L18 \n",
      "Yang H.-Y . K., Ruszkowski  M., Zweibel  E., 2013, MNRAS  , 436, 2734 \n",
      "Yusef-Zadeh  F., Wardle M., 2019, MNRAS  , 490, L1 \n",
      "Yusef-Zadeh  F., Arendt R. G., Wardle M., Heywood  I., Cotton W., Camilo \n",
      "F., 2022, ApJ , 925, L18 \n",
      "Yusef-Zadeh  F., Arendt R. G., Wardle M., Boldyrev  S., Heywood  I., Cotton \n",
      "W., Camilo F., 2022, MNRAS  , 515, 3059 \n",
      "Zweibel  E. G., 2017, Phys. Plasmas,  24, 55402 \n",
      "This paper has been typeset from a T E X/L A T E X file prepared  by the author. Downloaded from https://academic.oup.com/mnras/article/517/1/294/6677408 by NASA Goddard Space Flight Ctr user on 15 February 2023\n",
      "==================================\u001b[1m Ai Message \u001b[0m==================================\n",
      "\n",
      "The work performed by R. G. Arendt was supported by NASA under award number 80GSFC21M0002.\n",
      "\n",
      "FINAL ANSWER: 80GSFC21M0002\n"
     ]
    }
   ],
   "source": [
    "# task_id, q_query = question_query(questions[14])\n",
    "# response = agent(q_query)\n",
    "# result = prettry_print_response(task_id, response)\n",
    "\n",
    "\n",
    "# for m in response[\"messages\"]:\n",
    "#     m.pretty_print()\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "gaia-agent (3.12.11)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.12.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}