Skip to content
Navigation Menu
{{ message }}
forked from ggml-org/llama.cpp
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathserver-test-function-call.py
More file actions
executable file
·1151 lines (1044 loc) · 38.1 KB
/
Copy pathserver-test-function-call.py
File metadata and controls
executable file
·1151 lines (1044 loc) · 38.1 KB
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
#!/usr/bin/env python3
"""
Test tool calling capability via chat completions endpoint.
Each test case contains:
- tools: list of tool definitions (OpenAI-compatible)
- messages: initial conversation messages
- mock_tool_responses: dict mapping tool_name -> callable(arguments) -> str (JSON)
- validate: callable(tool_calls_history, final_content) -> (passed: bool, reason: str)
"""
import argparse
import json
import requests
import sys
# ---------------------------------------------------------------------------
# Color / formatting helpers
# ---------------------------------------------------------------------------
RESET = "\x1b[0m"
BOLD = "\x1b[1m"
DIM = "\x1b[2m"
# Foreground colors
CYAN = "\x1b[36m"
YELLOW = "\x1b[33m"
GREEN = "\x1b[32m"
RED = "\x1b[31m"
BLUE = "\x1b[34m"
WHITE = "\x1b[97m"
def _print(text="", end="\n"):
sys.stdout.write(text + end)
sys.stdout.flush()
def print_header(title):
bar = "─" * 60
_print(f"\n{BOLD}{CYAN}┌{bar}┐{RESET}")
_print(
f"{BOLD}{CYAN}│ {WHITE}{title}{CYAN}{' ' * max(0, 58 - len(title))}│{RESET}"
)
_print(f"{BOLD}{CYAN}└{bar}┘{RESET}")
def print_tool_call(name, args):
args_str = json.dumps(args)
_print(
f"\n {BOLD}{YELLOW}⚙ tool call{RESET} {CYAN}{name}{RESET}{DIM}({args_str}){RESET}"
)
def print_tool_result(result):
preview = result[:160] + ("…" if len(result) > 160 else "")
_print(f" {DIM}{BLUE}↳ result{RESET} {DIM}{preview}{RESET}")
def print_model_output(text):
# printed inline during streaming; prefix with a visual marker on first chunk
sys.stdout.write(text)
sys.stdout.flush()
def print_pass(reason):
_print(f"\n{BOLD}{GREEN}✔ PASS{RESET} {reason}")
def print_fail(reason):
_print(f"\n{BOLD}{RED}✘ FAIL{RESET} {reason}")
def print_info(msg):
_print(f"{DIM}{msg}{RESET}")
# ---------------------------------------------------------------------------
# HTTP helpers
# ---------------------------------------------------------------------------
def chat_completion(url, messages, tools=None, stream=False, force_tools=False):
payload = {
"messages": messages,
"stream": stream,
"max_tokens": 4096,
}
if tools:
payload["tools"] = tools
if force_tools:
payload["tool_choice"] = "required"
else:
payload["tool_choice"] = "auto"
try:
response = requests.post(url, json=payload, stream=stream)
response.raise_for_status()
except requests.exceptions.RequestException as e:
body = e.response.content if (e.response is not None) else b""
print_fail(f"Request error: {e} | body: {body}")
return None
full_content = ""
reasoning_content = ""
tool_calls: list[dict] = []
if stream:
for line in response.iter_lines():
if not line:
continue
decoded = line.decode("utf-8")
if not decoded.startswith("data: "):
continue
data_str = decoded[6:]
if data_str == "[DONE]":
break
try:
data = json.loads(data_str)
except json.JSONDecodeError:
continue
choices = data.get("choices", [])
if not choices:
continue
delta = choices[0].get("delta", {})
if delta.get("reasoning_content"):
reasoning_content += delta["reasoning_content"]
if delta.get("content"):
full_content += delta["content"]
print_model_output(delta["content"])
for tc in delta.get("tool_calls", []):
idx = tc.get("index", 0)
while len(tool_calls) <= idx:
tool_calls.append(
{
"id": "",
"type": "function",
"function": {"name": "", "arguments": ""},
}
)
if "id" in tc:
tool_calls[idx]["id"] += tc["id"]
if "function" in tc:
if "name" in tc["function"]:
tool_calls[idx]["function"]["name"] += tc["function"]["name"]
if "arguments" in tc["function"]:
tool_calls[idx]["function"]["arguments"] += tc["function"][
"arguments"
]
else:
data = response.json()
choices = data.get("choices", [])
if choices:
msg = choices[0].get("message", {})
full_content = msg.get("content") or ""
reasoning_content = msg.get("reasoning_content") or ""
tool_calls = msg.get("tool_calls") or []
if full_content:
print_model_output(full_content)
result = {"content": full_content, "tool_calls": tool_calls}
if reasoning_content:
result["reasoning_content"] = reasoning_content
return result
def all_tools_called(tools, all_tool_calls):
all_tool_names = set([tc["function"]["name"] for tc in tools])
all_called_tool_names = set([tc["function"]["name"] for tc in all_tool_calls])
return all_tool_names == all_called_tool_names
def run_agentic_loop(url, messages, tools, mock_tool_responses, stream, max_turns=6, force_tools=False):
"""
Drive the multi-turn tool-call loop:
1. Send messages to model.
2. If the model returns tool calls, execute mocks and append results.
3. Repeat until no more tool calls or max_turns reached.
Returns (all_tool_calls, final_content).
"""
msgs = list(messages)
all_tool_calls: list[dict] = []
for t in range(max_turns):
result = chat_completion(url, msgs, tools=tools, stream=stream, force_tools=(force_tools and not all_tools_called(tools, all_tool_calls)))
if result is None:
return all_tool_calls, None
tcs = result.get("tool_calls") or []
content = result.get("content") or ""
if not tcs:
# Print a visual separator before the final model response
if content:
_print(f"\n{DIM}{'·'*60}{RESET}")
_print(f"{DIM} model response:{RESET}\n")
return all_tool_calls, content
# Record tool calls for validation
all_tool_calls.extend(tcs)
# Append assistant message with tool calls
assistant_msg: dict = {
"role": "assistant",
"content": content,
"tool_calls": tcs,
}
reasoning = result.get("reasoning_content")
if reasoning:
assistant_msg["reasoning_content"] = reasoning
msgs.append(assistant_msg)
# Execute each tool call via mock and append tool result messages
for tc in tcs:
tool_name = tc["function"]["name"]
try:
args = json.loads(tc["function"]["arguments"])
except json.JSONDecodeError:
args = {}
print_tool_call(tool_name, args)
mock_fn = mock_tool_responses.get(tool_name)
if mock_fn:
tool_result = mock_fn(args)
else:
tool_result = json.dumps({"error": f"Unknown tool: {tool_name}"})
print_tool_result(tool_result)
msgs.append(
{
"role": "tool",
"tool_call_id": tc.get("id", ""),
"content": tool_result,
}
)
return all_tool_calls, None
# ---------------------------------------------------------------------------
# Test case runner
# ---------------------------------------------------------------------------
def run_test(url, test_case, stream, force_tools):
name = test_case["name"]
mode = f"{'stream' if stream else 'non-stream'}"
print_header(f"{name} [{mode}, force_tools={force_tools}] ")
all_tool_calls, final_content = run_agentic_loop(
url,
messages=test_case["messages"],
tools=test_case["tools"],
mock_tool_responses=test_case["mock_tool_responses"],
stream=stream,
force_tools=force_tools
)
if final_content is None and not all_tool_calls:
print_fail("No response from server.")
return False
passed, reason = test_case["validate"](all_tool_calls, final_content)
if passed:
print_pass(reason)
else:
print_fail(reason)
return passed
# ---------------------------------------------------------------------------
# Test case definitions
# ---------------------------------------------------------------------------
# ---- Test 1: E-commerce multi-step search (Azzoo = anonymized marketplace) ----
_AZZOO_TOOLS = [
{
"type": "function",
"function": {
"name": "azzoo_search_products",
"description": (
"Search for products on Azzoo marketplace by keyword. "
"Returns a list of matching products with IDs, titles, ratings and prices."
),
"parameters": {
"type": "object",
"properties": {
"query": {
"type": "string",
"description": "Search keyword or phrase",
},
"page": {
"type": "string",
"description": "Page number (1-based)",
"default": "1",
},
},
"required": ["query"],
},
},
},
{
"type": "function",
"function": {
"name": "azzoo_get_product",
"description": "Retrieve detailed information about a specific Azzoo product including specs and price.",
"parameters": {
"type": "object",
"properties": {
"product_id": {
"type": "string",
"description": "Azzoo product identifier (e.g. AZB12345)",
},
},
"required": ["product_id"],
},
},
},
{
"type": "function",
"function": {
"name": "azzoo_get_reviews",
"description": "Fetch customer reviews for an Azzoo product.",
"parameters": {
"type": "object",
"properties": {
"product_id": {
"type": "string",
"description": "Azzoo product identifier",
},
"page": {
"type": "string",
"description": "Review page number",
"default": "1",
},
},
"required": ["product_id"],
},
},
},
]
_AZZOO_SEARCH_RESULT = {
"results": [
{
"product_id": "AZB00001",
"title": "SteelBrew Pro Kettle 1.7L",
"rating": 4.6,
"price": 34.99,
},
{
"product_id": "AZB00002",
"title": "HeatKeep Gooseneck Kettle",
"rating": 4.3,
"price": 27.50,
},
{
"product_id": "AZB00003",
"title": "QuickBoil Stainless Kettle",
"rating": 4.1,
"price": 21.00,
},
]
}
_AZZOO_PRODUCT_RESULT = {
"product_id": "AZB00001",
"title": "SteelBrew Pro Kettle 1.7L",
"price": 34.99,
"rating": 4.6,
"review_count": 2847,
"specs": {
"material": "18/8 stainless steel",
"capacity": "1.7 L",
"auto_shutoff": True,
"keep_warm": "30 min",
"warranty": "2 years",
},
}
_AZZOO_REVIEWS_RESULT = {
"product_id": "AZB00001",
"average_rating": 4.6,
"reviews": [
{
"rating": 5,
"title": "Excellent build quality",
"body": "Very sturdy, boils fast and stays warm longer than expected.",
},
{
"rating": 5,
"title": "Great for loose-leaf tea",
"body": "The wide spout makes filling a teapot easy. No leaks after months of use.",
},
{
"rating": 3,
"title": "Minor lid issue",
"body": "The lid doesn't always click shut properly, but overall happy with it.",
},
{
"rating": 4,
"title": "Good value",
"body": "Heats quickly and the auto shutoff works reliably.",
},
],
}
AZZOO_TEST_CASE = {
"name": "Azzoo E-commerce: search -> product detail -> reviews",
"messages": [
{
"role": "user",
"content": (
"I need a durable stainless steel tea kettle for my weekly tea gatherings. "
"Please search Azzoo for 'stainless steel tea kettle', then get full details "
"on the top-rated result, and finally fetch its customer reviews so I can "
"check for recurring complaints. Give me a summary with pros and cons."
),
}
],
"tools": _AZZOO_TOOLS,
"mock_tool_responses": {
"azzoo_search_products": lambda _: json.dumps(_AZZOO_SEARCH_RESULT),
"azzoo_get_product": lambda _: json.dumps(_AZZOO_PRODUCT_RESULT),
"azzoo_get_reviews": lambda _: json.dumps(_AZZOO_REVIEWS_RESULT),
},
"validate": lambda tcs, content: _validate_azzoo(tcs, content),
}
def _validate_azzoo(tcs, content):
names = [tc["function"]["name"] for tc in tcs]
if not names:
return False, "No tool calls made"
if "azzoo_search_products" not in names:
return False, f"Expected azzoo_search_products to be called, got: {names}"
# After search the model should look up product details
if "azzoo_get_product" not in names and "azzoo_get_reviews" not in names:
return False, f"Expected follow-up product/review lookup, got: {names}"
# Verify product lookup used an ID from search results
for tc in tcs:
if tc["function"]["name"] == "azzoo_get_product":
try:
args = json.loads(tc["function"]["arguments"])
pid = args.get("product_id", "")
if not pid:
return False, "azzoo_get_product called with empty product_id"
except json.JSONDecodeError:
return False, "azzoo_get_product arguments are not valid JSON"
if not content:
return False, "No final summary produced"
return True, f"All expected tools called in order: {names}"
# ---- Test 2: Fitness BMI + exercise recommendations ----
_FITNESS_TOOLS = [
{
"type": "function",
"function": {
"name": "calculate_bmi",
"description": "Calculate Body Mass Index (BMI) from weight and height.",
"parameters": {
"type": "object",
"properties": {
"weight_kg": {
"type": "number",
"description": "Body weight in kilograms",
},
"height_m": {"type": "number", "description": "Height in meters"},
},
"required": ["weight_kg", "height_m"],
},
},
},
{
"type": "function",
"function": {
"name": "get_exercises",
"description": (
"Fetch a list of exercises filtered by muscle group, difficulty, category, "
"and/or force type."
),
"parameters": {
"type": "object",
"properties": {
"muscle": {
"type": "string",
"description": "Target muscle group (e.g. chest, back, legs)",
},
"difficulty": {
"type": "string",
"description": "Difficulty level: beginner, intermediate, expert",
},
"category": {
"type": "string",
"description": "Exercise category (e.g. strength, cardio, stretching)",
},
"force": {
"type": "string",
"description": "Force type: push, pull, static",
},
},
"required": [],
},
},
},
]
_BMI_RESULT = {"bmi": 24.5, "category": "Normal weight", "healthy_range": "18.5 – 24.9"}
_EXERCISES_RESULT = {
"exercises": [
{
"name": "Push-Up",
"muscle": "chest",
"difficulty": "beginner",
"equipment": "none",
"instructions": "Keep body straight, lower chest to floor.",
},
{
"name": "Incline Dumbbell Press",
"muscle": "chest",
"difficulty": "beginner",
"equipment": "dumbbells, bench",
"instructions": "Press dumbbells up from chest on incline bench.",
},
{
"name": "Chest Fly (cables)",
"muscle": "chest",
"difficulty": "beginner",
"equipment": "cable machine",
"instructions": "Bring cables together in an arc motion.",
},
]
}
FITNESS_TEST_CASE = {
"name": "Fitness: BMI calculation + exercise suggestions",
"messages": [
{
"role": "user",
"content": (
"I'm a 32-year-old male, 78 kg and 1.80 m tall. "
"Please calculate my BMI and then suggest some beginner chest exercises I can do "
"to build strength. Give me a short personalised plan."
),
}
],
"tools": _FITNESS_TOOLS,
"mock_tool_responses": {
"calculate_bmi": lambda _: json.dumps(_BMI_RESULT),
"get_exercises": lambda _: json.dumps(_EXERCISES_RESULT),
},
"validate": lambda tcs, content: _validate_fitness(tcs, content),
}
def _validate_fitness(tcs, content):
names = [tc["function"]["name"] for tc in tcs]
if not names:
return False, "No tool calls made"
if "calculate_bmi" not in names:
return False, f"Expected calculate_bmi to be called, got: {names}"
# Validate BMI args contain plausible values
for tc in tcs:
if tc["function"]["name"] == "calculate_bmi":
try:
args = json.loads(tc["function"]["arguments"])
w = args.get("weight_kg")
h = args.get("height_m")
if w is None or h is None:
return False, f"calculate_bmi missing weight_kg or height_m: {args}"
if not (50 <= float(w) <= 200):
return False, f"calculate_bmi weight out of plausible range: {w}"
if not (1.0 <= float(h) <= 2.5):
return False, f"calculate_bmi height out of plausible range: {h}"
except (json.JSONDecodeError, ValueError) as e:
return False, f"calculate_bmi argument error: {e}"
if not content:
return False, "No final plan produced"
return True, f"Tools called: {names}"
# ---- Test 3: Community class planning (anonymised cooking/topic discovery) ----
_COMMUNITY_TOOLS = [
{
"type": "function",
"function": {
"name": "get_trending_questions",
"description": (
"Fetch commonly asked questions on a topic from search engine 'People Also Ask' boxes."
),
"parameters": {
"type": "object",
"properties": {
"query": {"type": "string", "description": "Topic to search for"},
"max_results": {
"type": "integer",
"description": "Maximum questions to return",
"default": 10,
},
},
"required": ["query"],
},
},
},
{
"type": "function",
"function": {
"name": "search_mobile_apps",
"description": "Search the mobile app store for apps matching a category or keyword.",
"parameters": {
"type": "object",
"properties": {
"keyword": {
"type": "string",
"description": "Search keyword (e.g. 'Italian cooking')",
},
"platform": {
"type": "string",
"enum": ["ios", "android", "both"],
"default": "both",
},
"max_results": {
"type": "integer",
"description": "Number of results",
"default": 10,
},
},
"required": ["keyword"],
},
},
},
]
_TRENDING_QUESTIONS_RESULT = {
"query": "Italian cuisine",
"questions": [
"What are the most popular Italian dishes?",
"What makes Italian food different from other cuisines?",
"How do you make authentic Italian pasta from scratch?",
"What are traditional Italian desserts?",
"What herbs are commonly used in Italian cooking?",
"Is Italian food healthy?",
"What wine pairs best with Italian pasta?",
],
}
_APPS_RESULT = {
"keyword": "Italian cooking",
"results": [
{
"name": "PastaPro",
"rating": 4.5,
"installs": "500K+",
"focus": "pasta recipes only",
},
{
"name": "CookEasy",
"rating": 4.2,
"installs": "1M+",
"focus": "general cooking, limited Italian content",
},
{
"name": "ItalianKitchen",
"rating": 3.8,
"installs": "100K+",
"focus": "regional Italian recipes, no video",
},
],
}
COMMUNITY_CLASS_TEST_CASE = {
"name": "Community class planning: trending topics + app gap analysis",
"messages": [
{
"role": "user",
"content": (
"I want to start teaching Italian cooking classes at my community centre. "
"First, find out what people commonly ask about Italian cuisine online. "
"Then search for existing Italian cooking apps to see what they cover. "
"Use both results to suggest three unique angles for my classes that fill gaps "
"in what apps already offer."
),
}
],
"tools": _COMMUNITY_TOOLS,
"mock_tool_responses": {
"get_trending_questions": lambda _: json.dumps(_TRENDING_QUESTIONS_RESULT),
"search_mobile_apps": lambda _: json.dumps(_APPS_RESULT),
},
"validate": lambda tcs, content: _validate_community(tcs, content),
}
def _validate_community(tcs, content):
names = [tc["function"]["name"] for tc in tcs]
if not names:
return False, "No tool calls made"
missing = [
t for t in ("get_trending_questions", "search_mobile_apps") if t not in names
]
if missing:
return False, f"Missing expected tool calls: {missing}; got: {names}"
if not content:
return False, "No class suggestion produced"
return True, f"Both discovery tools called: {names}"
# ---- Test 4: Multi-hostname geolocation filter (anonymized gallery discovery) ----
# Inspired by: checking gallery website server locations to find truly remote venues.
# Anonymized: galleryone.de → halle-eins.de, gallerytwo.fr → galerie-deux.fr,
# gallerythree.it → galleria-tre.it
_GEO_TOOLS = [
{
"type": "function",
"function": {
"name": "lookup_ip_geolocation",
"description": (
"Retrieve geolocation data for an IP address or hostname, including country, "
"city, coordinates, and network info. Useful for verifying physical server "
"locations or personalising regional content."
),
"parameters": {
"type": "object",
"properties": {
"host": {
"type": "string",
"description": "IP address or hostname to look up (e.g. '8.8.8.8' or 'example.com').",
},
},
"required": ["host"],
},
},
},
]
# Mock: one urban (Berlin → discard), two rural (keep)
_GEO_RESPONSES = {
"halle-eins.de": {
"host": "halle-eins.de",
"city": "Berlin",
"country": "DE",
"lat": 52.5200,
"lon": 13.4050,
"is_major_city": True,
},
"galerie-deux.fr": {
"host": "galerie-deux.fr",
"city": "Rocamadour",
"country": "FR",
"lat": 44.7994,
"lon": 1.6178,
"is_major_city": False,
},
"galleria-tre.it": {
"host": "galleria-tre.it",
"city": "Matera",
"country": "IT",
"lat": 40.6664,
"lon": 16.6044,
"is_major_city": False,
},
}
def _geo_mock(args):
host = args.get("host", "")
return json.dumps(_GEO_RESPONSES.get(host, {"error": f"unknown host: {host}"}))
GEO_TEST_CASE = {
"name": "Gallery geolocation: filter urban venues, keep remote ones",
"messages": [
{
"role": "user",
"content": (
"I have abstract paintings to exhibit in remote European galleries. "
"I received enquiries from three venues: halle-eins.de, galerie-deux.fr, "
"and galleria-tre.it. Please look up the geolocation of each website's server. "
"Discard any venue whose server is in a major city (e.g. Berlin, Paris, Rome). "
"For the remaining venues, report their exact coordinates so I can check "
"whether hiking trails are nearby — my work thrives where nature and art meet."
),
}
],
"tools": _GEO_TOOLS,
"mock_tool_responses": {
"lookup_ip_geolocation": _geo_mock,
},
"validate": lambda tcs, content: _validate_geo(tcs, content),
}
def _validate_geo(tcs, content):
names = [tc["function"]["name"] for tc in tcs]
if not names:
return False, "No tool calls made"
# Expect exactly one geolocation call per domain (3 total)
geo_calls = [tc for tc in tcs if tc["function"]["name"] == "lookup_ip_geolocation"]
if len(geo_calls) < 3:
return (
False,
f"Expected geolocation called 3 times (once per domain), got {len(geo_calls)}",
)
queried_hosts = set()
for tc in geo_calls:
try:
args = json.loads(tc["function"]["arguments"])
host = args.get("host", "")
if not host:
return False, f"lookup_ip_geolocation called with empty host: {args}"
queried_hosts.add(host)
except json.JSONDecodeError:
return False, "lookup_ip_geolocation arguments are not valid JSON"
expected = {"halle-eins.de", "galerie-deux.fr", "galleria-tre.it"}
if not expected.issubset(queried_hosts):
return (
False,
f"Not all domains queried. Expected {expected}, got {queried_hosts}",
)
if not content:
return False, "No final summary produced"
return True, f"All 3 domains geolocated: {sorted(queried_hosts)}"
# ---- Test 5: EV fleet expansion — stock → security → property → video ----
# Inspired by: multi-step business analysis combining finance, cybersecurity,
# real estate and educational content.
# Anonymized: Tesla → Voltara (VLTR), Rivian → Rivex (RVXN),
# Trenton → Halverton
_EV_TOOLS = [
{
"type": "function",
"function": {
"name": "get_stock_quote",
"description": "Retrieve the latest market quote for a financial instrument by ticker symbol.",
"parameters": {
"type": "object",
"properties": {
"symbol": {
"type": "string",
"description": "Ticker symbol (e.g. 'VLTR', 'RVXN')",
},
"interval": {
"type": "string",
"description": "Time interval: 1min, 5min, 1h, 1day, 1week",
"default": "1day",
},
},
"required": ["symbol"],
},
},
},
{
"type": "function",
"function": {
"name": "get_security_advisories",
"description": (
"Fetch current cybersecurity advisories from the national security agency, "
"covering known vulnerabilities and exploits for industrial and consumer systems."
),
"parameters": {
"type": "object",
"properties": {
"keyword": {
"type": "string",
"description": "Filter advisories by keyword or product name",
},
"limit": {
"type": "integer",
"description": "Maximum number of advisories to return",
"default": 5,
},
},
"required": [],
},
},
},
{
"type": "function",
"function": {
"name": "search_commercial_properties",
"description": "Search for commercial properties (offices, garages, warehouses) available for rent or sale in a given city.",
"parameters": {
"type": "object",
"properties": {
"city": {"type": "string", "description": "City name to search in"},
"property_type": {
"type": "string",
"description": "Type of property: office, garage, warehouse, premises",
},
"operation": {
"type": "string",
"enum": ["rent", "sale"],
"default": "rent",
},
"max_price": {
"type": "integer",
"description": "Maximum monthly rent or sale price",
},
},
"required": ["city", "property_type"],
},
},
},
{
"type": "function",
"function": {
"name": "get_video_recommendations",
"description": "Fetch a list of recommended videos related to a given topic or reference video.",
"parameters": {
"type": "object",
"properties": {
"topic": {
"type": "string",
"description": "Topic or keyword to search for related videos",
},
},
"required": ["topic"],
},
},
},
]
_STOCK_RESULT_VLTR = {
"symbol": "VLTR",
"company": "Voltara Inc.",
"price": 218.45,
"change_pct": "+2.3%",
"market_cap": "694B",
"currency": "USD",
}
_STOCK_RESULT_RVXN = {
"symbol": "RVXN",
"company": "Rivex Motors",
"price": 12.80,
"change_pct": "-1.1%",
"market_cap": "11B",
"currency": "USD",
}
_ADVISORIES_RESULT = {
"count": 2,
"advisories": [
{
"id": "ICSA-24-102-01",
"title": "Voltara In-Vehicle Infotainment System Authentication Bypass",
"severity": "Medium",
"summary": "Improper authentication in the OTA update module may allow an adjacent attacker to install unsigned firmware.",
"published": "2024-04-11",
},
{
"id": "ICSA-24-085-03",
"title": "Voltara Charging Management API Input Validation Flaw",
"severity": "Low",
"summary": "Insufficient input validation in the charging session API could expose internal error messages.",
"published": "2024-03-26",
},
],
}
_PROPERTIES_RESULT = {
"city": "Halverton",
"listings": [
{
"id": "HV-0041",
"type": "garage",
"area_sqm": 420,
"monthly_rent": 2800,
"ev_power_outlets": 12,
"address": "14 Ironworks Lane, Halverton",
},
{
"id": "HV-0089",
"type": "warehouse",
"area_sqm": 900,
"monthly_rent": 4200,
"ev_power_outlets": 30,
"address": "7 Depot Road, Halverton",
},
],
}
_VIDEOS_RESULT = {
"topic": "fleet electrification",
"recommendations": [
{
"title": "How to Build an EV Fleet from Scratch",
"channel": "Fleet Future",
"views": "182K",
},
{
"title": "EV Charging Infrastructure for Commercial Fleets",
"channel": "GreenDrive Pro",
"views": "94K",
},
{
"title": "Total Cost of Ownership: Electric vs Diesel Vans",
"channel": "LogisticsTech",
You can’t perform that action at this time.