forked from openai/openai-agents-python
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathtest_tracing_errors.py
More file actions
328 lines (263 loc) · 9.37 KB
/
Copy pathtest_tracing_errors.py
File metadata and controls
328 lines (263 loc) · 9.37 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
from __future__ import annotations
import json
from typing import Any
import pytest
from typing_extensions import TypedDict
from agents import (
Agent,
GuardrailFunctionOutput,
InputGuardrail,
InputGuardrailTripwireTriggered,
MaxTurnsExceeded,
ModelBehaviorError,
RunContextWrapper,
Runner,
TResponseInputItem,
)
from agents.tracing import AgentSpanData, FunctionSpanData, GenerationSpanData
from .fake_model import FakeModel
from .test_responses import (
get_final_output_message,
get_function_tool,
get_function_tool_call,
get_handoff_tool_call,
get_text_message,
)
from .testing_processor import fetch_ordered_spans, fetch_traces
@pytest.mark.asyncio
async def test_single_turn_model_error():
model = FakeModel(tracing_enabled=True)
model.set_next_output(ValueError("test error"))
agent = Agent(
name="test_agent",
model=model,
)
with pytest.raises(ValueError):
await Runner.run(agent, input="first_test")
traces = fetch_traces()
assert len(traces) == 1, f"Expected 1 trace, got {len(traces)}"
spans = fetch_ordered_spans()
assert len(spans) == 2, f"should have agent and generation spans, got {len(spans)}"
generation_span = spans[1]
assert isinstance(generation_span.span_data, GenerationSpanData)
assert generation_span.error, "should have error"
@pytest.mark.asyncio
async def test_multi_turn_no_handoffs():
model = FakeModel(tracing_enabled=True)
agent = Agent(
name="test_agent",
model=model,
tools=[get_function_tool("foo", "tool_result")],
)
model.add_multiple_turn_outputs(
[
# First turn: a message and tool call
[get_text_message("a_message"), get_function_tool_call("foo", json.dumps({"a": "b"}))],
# Second turn: error
ValueError("test error"),
# Third turn: text message
[get_text_message("done")],
]
)
with pytest.raises(ValueError):
await Runner.run(agent, input="first_test")
traces = fetch_traces()
assert len(traces) == 1, f"Expected 1 trace, got {len(traces)}"
spans = fetch_ordered_spans()
assert len(spans) == 4, (
f"should have agent, generation, tool, generation, got {len(spans)} with data: "
f"{[x.span_data for x in spans]}"
)
last_generation_span = [x for x in spans if isinstance(x.span_data, GenerationSpanData)][-1]
assert last_generation_span.error, "should have error"
@pytest.mark.asyncio
async def test_tool_call_error():
model = FakeModel(tracing_enabled=True)
agent = Agent(
name="test_agent",
model=model,
tools=[get_function_tool("foo", "tool_result", hide_errors=True)],
)
model.set_next_output(
[get_text_message("a_message"), get_function_tool_call("foo", "bad_json")],
)
with pytest.raises(ModelBehaviorError):
await Runner.run(agent, input="first_test")
traces = fetch_traces()
assert len(traces) == 1, f"Expected 1 trace, got {len(traces)}"
spans = fetch_ordered_spans()
assert len(spans) == 3, (
f"should have agent, generation, tool spans, got {len(spans)} with data: "
f"{[x.span_data for x in spans]}"
)
function_span = [x for x in spans if isinstance(x.span_data, FunctionSpanData)][0]
assert function_span.error, "should have error"
@pytest.mark.asyncio
async def test_multiple_handoff_doesnt_error():
model = FakeModel(tracing_enabled=True)
agent_1 = Agent(
name="test",
model=model,
)
agent_2 = Agent(
name="test",
model=model,
)
agent_3 = Agent(
name="test",
model=model,
handoffs=[agent_1, agent_2],
tools=[get_function_tool("some_function", "result")],
)
model.add_multiple_turn_outputs(
[
# First turn: a tool call
[get_function_tool_call("some_function", json.dumps({"a": "b"}))],
# Second turn: a message and 2 handoff
[
get_text_message("a_message"),
get_handoff_tool_call(agent_1),
get_handoff_tool_call(agent_2),
],
# Third turn: text message
[get_text_message("done")],
]
)
result = await Runner.run(agent_3, input="user_message")
assert result.last_agent == agent_1, "should have picked first handoff"
traces = fetch_traces()
assert len(traces) == 1, f"Expected 1 trace, got {len(traces)}"
spans = fetch_ordered_spans()
assert len(spans) == 7, (
f"should have 2 agent, 1 function, 3 generation, 1 handoff, got {len(spans)} with data: "
f"{[x.span_data for x in spans]}"
)
class Foo(TypedDict):
bar: str
@pytest.mark.asyncio
async def test_multiple_final_output_doesnt_error():
model = FakeModel(tracing_enabled=True)
agent_1 = Agent(
name="test",
model=model,
output_type=Foo,
)
model.set_next_output(
[
get_final_output_message(json.dumps(Foo(bar="baz"))),
get_final_output_message(json.dumps(Foo(bar="abc"))),
]
)
result = await Runner.run(agent_1, input="user_message")
assert result.final_output == Foo(bar="abc")
traces = fetch_traces()
assert len(traces) == 1, f"Expected 1 trace, got {len(traces)}"
spans = fetch_ordered_spans()
assert len(spans) == 2, (
f"should have 1 agent, 1 generation, got {len(spans)} with data: "
f"{[x.span_data for x in spans]}"
)
@pytest.mark.asyncio
async def test_handoffs_lead_to_correct_agent_spans():
model = FakeModel(tracing_enabled=True)
agent_1 = Agent(
name="test_agent_1",
model=model,
tools=[get_function_tool("some_function", "result")],
)
agent_2 = Agent(
name="test_agent_2",
model=model,
handoffs=[agent_1],
tools=[get_function_tool("some_function", "result")],
)
agent_3 = Agent(
name="test_agent_3",
model=model,
handoffs=[agent_1, agent_2],
tools=[get_function_tool("some_function", "result")],
)
agent_1.handoffs.append(agent_3)
model.add_multiple_turn_outputs(
[
# First turn: a tool call
[get_function_tool_call("some_function", json.dumps({"a": "b"}))],
# Second turn: a message and 2 handoff
[
get_text_message("a_message"),
get_handoff_tool_call(agent_1),
get_handoff_tool_call(agent_2),
],
# Third turn: tool call
[get_function_tool_call("some_function", json.dumps({"a": "b"}))],
# Fourth turn: handoff
[get_handoff_tool_call(agent_3)],
# Fifth turn: text message
[get_text_message("done")],
]
)
result = await Runner.run(agent_3, input="user_message")
assert result.last_agent == agent_3, (
f"should have ended on the third agent, got {result.last_agent.name}"
)
traces = fetch_traces()
assert len(traces) == 1, f"Expected 1 trace, got {len(traces)}"
spans = fetch_ordered_spans()
assert len(spans) == 12, (
f"should have 3 agents, 2 function, 5 generation, 2 handoff, got {len(spans)} with data: "
f"{[x.span_data for x in spans]}"
)
@pytest.mark.asyncio
async def test_max_turns_exceeded():
model = FakeModel(tracing_enabled=True)
agent = Agent(
name="test",
model=model,
output_type=Foo,
tools=[get_function_tool("foo", "result")],
)
model.add_multiple_turn_outputs(
[
[get_function_tool_call("foo")],
[get_function_tool_call("foo")],
[get_function_tool_call("foo")],
[get_function_tool_call("foo")],
[get_function_tool_call("foo")],
]
)
with pytest.raises(MaxTurnsExceeded):
await Runner.run(agent, input="user_message", max_turns=2)
traces = fetch_traces()
assert len(traces) == 1, f"Expected 1 trace, got {len(traces)}"
spans = fetch_ordered_spans()
assert len(spans) == 5, (
f"should have 1 agent span, 2 generations, 2 function calls, got "
f"{len(spans)} with data: {[x.span_data for x in spans]}"
)
agent_span = [x for x in spans if isinstance(x.span_data, AgentSpanData)][-1]
assert agent_span.error, "last agent should have error"
def guardrail_function(
context: RunContextWrapper[Any], agent: Agent[Any], input: str | list[TResponseInputItem]
) -> GuardrailFunctionOutput:
return GuardrailFunctionOutput(
output_info=None,
tripwire_triggered=True,
)
@pytest.mark.asyncio
async def test_guardrail_error():
agent = Agent(
name="test", input_guardrails=[InputGuardrail(guardrail_function=guardrail_function)]
)
model = FakeModel()
model.set_next_output([get_text_message("some_message")])
with pytest.raises(InputGuardrailTripwireTriggered):
await Runner.run(agent, input="user_message")
traces = fetch_traces()
assert len(traces) == 1, f"Expected 1 trace, got {len(traces)}"
spans = fetch_ordered_spans()
assert len(spans) == 2, (
f"should have 1 agent, 1 guardrail, got {len(spans)} with data: "
f"{[x.span_data for x in spans]}"
)
agent_span = [x for x in spans if isinstance(x.span_data, AgentSpanData)][-1]
assert agent_span.error, "last agent should have error"