forked from openai/openai-agents-python
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtest_run_step_execution.py
307 lines (252 loc) · 9.32 KB
/
test_run_step_execution.py
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
from __future__ import annotations
from typing import Any
import pytest
from pydantic import BaseModel
from agents import (
Agent,
MessageOutputItem,
ModelResponse,
RunConfig,
RunContextWrapper,
RunHooks,
RunItem,
Runner,
ToolCallItem,
ToolCallOutputItem,
TResponseInputItem,
Usage,
)
from agents._run_impl import (
NextStepFinalOutput,
NextStepHandoff,
NextStepRunAgain,
RunImpl,
SingleStepResult,
)
from .test_responses import (
get_final_output_message,
get_function_tool,
get_function_tool_call,
get_handoff_tool_call,
get_text_input_item,
get_text_message,
)
@pytest.mark.asyncio
async def test_empty_response_is_final_output():
agent = Agent[None](name="test")
response = ModelResponse(
output=[],
usage=Usage(),
referenceable_id=None,
)
result = await get_execute_result(agent, response)
assert result.original_input == "hello"
assert result.generated_items == []
assert isinstance(result.next_step, NextStepFinalOutput)
assert result.next_step.output == ""
@pytest.mark.asyncio
async def test_plaintext_agent_no_tool_calls_is_final_output():
agent = Agent(name="test")
response = ModelResponse(
output=[get_text_message("hello_world")],
usage=Usage(),
referenceable_id=None,
)
result = await get_execute_result(agent, response)
assert result.original_input == "hello"
assert len(result.generated_items) == 1
assert_item_is_message(result.generated_items[0], "hello_world")
assert isinstance(result.next_step, NextStepFinalOutput)
assert result.next_step.output == "hello_world"
@pytest.mark.asyncio
async def test_plaintext_agent_no_tool_calls_multiple_messages_is_final_output():
agent = Agent(name="test")
response = ModelResponse(
output=[
get_text_message("hello_world"),
get_text_message("bye"),
],
usage=Usage(),
referenceable_id=None,
)
result = await get_execute_result(
agent,
response,
original_input=[
get_text_input_item("test"),
get_text_input_item("test2"),
],
)
assert len(result.original_input) == 2
assert len(result.generated_items) == 2
assert_item_is_message(result.generated_items[0], "hello_world")
assert_item_is_message(result.generated_items[1], "bye")
assert isinstance(result.next_step, NextStepFinalOutput)
assert result.next_step.output == "bye"
@pytest.mark.asyncio
async def test_plaintext_agent_with_tool_call_is_run_again():
agent = Agent(name="test", tools=[get_function_tool(name="test", return_value="123")])
response = ModelResponse(
output=[get_text_message("hello_world"), get_function_tool_call("test", "")],
usage=Usage(),
referenceable_id=None,
)
result = await get_execute_result(agent, response)
assert result.original_input == "hello"
# 3 items: new message, tool call, tool result
assert len(result.generated_items) == 3
assert isinstance(result.next_step, NextStepRunAgain)
items = result.generated_items
assert_item_is_message(items[0], "hello_world")
assert_item_is_function_tool_call(items[1], "test", None)
assert_item_is_function_tool_call_output(items[2], "123")
assert isinstance(result.next_step, NextStepRunAgain)
@pytest.mark.asyncio
async def test_multiple_tool_calls():
agent = Agent(
name="test",
tools=[
get_function_tool(name="test_1", return_value="123"),
get_function_tool(name="test_2", return_value="456"),
get_function_tool(name="test_3", return_value="789"),
],
)
response = ModelResponse(
output=[
get_text_message("Hello, world!"),
get_function_tool_call("test_1"),
get_function_tool_call("test_2"),
],
usage=Usage(),
referenceable_id=None,
)
result = await get_execute_result(agent, response)
assert result.original_input == "hello"
# 5 items: new message, 2 tool calls, 2 tool call outputs
assert len(result.generated_items) == 5
assert isinstance(result.next_step, NextStepRunAgain)
items = result.generated_items
assert_item_is_message(items[0], "Hello, world!")
assert_item_is_function_tool_call(items[1], "test_1", None)
assert_item_is_function_tool_call(items[2], "test_2", None)
assert isinstance(result.next_step, NextStepRunAgain)
@pytest.mark.asyncio
async def test_handoff_output_leads_to_handoff_next_step():
agent_1 = Agent(name="test_1")
agent_2 = Agent(name="test_2")
agent_3 = Agent(name="test_3", handoffs=[agent_1, agent_2])
response = ModelResponse(
output=[get_text_message("Hello, world!"), get_handoff_tool_call(agent_1)],
usage=Usage(),
referenceable_id=None,
)
result = await get_execute_result(agent_3, response)
assert isinstance(result.next_step, NextStepHandoff)
assert result.next_step.new_agent == agent_1
assert len(result.generated_items) == 3
class Foo(BaseModel):
bar: str
@pytest.mark.asyncio
async def test_final_output_without_tool_runs_again():
agent = Agent(name="test", output_type=Foo, tools=[get_function_tool("tool_1", "result")])
response = ModelResponse(
output=[get_function_tool_call("tool_1")],
usage=Usage(),
referenceable_id=None,
)
result = await get_execute_result(agent, response)
assert isinstance(result.next_step, NextStepRunAgain)
assert len(result.generated_items) == 2, "expected 2 items: tool call, tool call output"
@pytest.mark.asyncio
async def test_final_output_leads_to_final_output_next_step():
agent = Agent(name="test", output_type=Foo)
response = ModelResponse(
output=[
get_text_message("Hello, world!"),
get_final_output_message(Foo(bar="123").model_dump_json()),
],
usage=Usage(),
referenceable_id=None,
)
result = await get_execute_result(agent, response)
assert isinstance(result.next_step, NextStepFinalOutput)
assert result.next_step.output == Foo(bar="123")
@pytest.mark.asyncio
async def test_handoff_and_final_output_leads_to_handoff_next_step():
agent_1 = Agent(name="test_1")
agent_2 = Agent(name="test_2")
agent_3 = Agent(name="test_3", handoffs=[agent_1, agent_2], output_type=Foo)
response = ModelResponse(
output=[
get_final_output_message(Foo(bar="123").model_dump_json()),
get_handoff_tool_call(agent_1),
],
usage=Usage(),
referenceable_id=None,
)
result = await get_execute_result(agent_3, response)
assert isinstance(result.next_step, NextStepHandoff)
assert result.next_step.new_agent == agent_1
@pytest.mark.asyncio
async def test_multiple_final_output_leads_to_final_output_next_step():
agent_1 = Agent(name="test_1")
agent_2 = Agent(name="test_2")
agent_3 = Agent(name="test_3", handoffs=[agent_1, agent_2], output_type=Foo)
response = ModelResponse(
output=[
get_final_output_message(Foo(bar="123").model_dump_json()),
get_final_output_message(Foo(bar="456").model_dump_json()),
],
usage=Usage(),
referenceable_id=None,
)
result = await get_execute_result(agent_3, response)
assert isinstance(result.next_step, NextStepFinalOutput)
assert result.next_step.output == Foo(bar="456")
# === Helpers ===
def assert_item_is_message(item: RunItem, text: str) -> None:
assert isinstance(item, MessageOutputItem)
assert item.raw_item.type == "message"
assert item.raw_item.role == "assistant"
assert item.raw_item.content[0].type == "output_text"
assert item.raw_item.content[0].text == text
def assert_item_is_function_tool_call(
item: RunItem, name: str, arguments: str | None = None
) -> None:
assert isinstance(item, ToolCallItem)
assert item.raw_item.type == "function_call"
assert item.raw_item.name == name
assert not arguments or item.raw_item.arguments == arguments
def assert_item_is_function_tool_call_output(item: RunItem, output: str) -> None:
assert isinstance(item, ToolCallOutputItem)
assert item.raw_item["type"] == "function_call_output"
assert item.raw_item["output"] == output
async def get_execute_result(
agent: Agent[Any],
response: ModelResponse,
*,
original_input: str | list[TResponseInputItem] | None = None,
generated_items: list[RunItem] | None = None,
hooks: RunHooks[Any] | None = None,
context_wrapper: RunContextWrapper[Any] | None = None,
run_config: RunConfig | None = None,
) -> SingleStepResult:
output_schema = Runner._get_output_schema(agent)
handoffs = Runner._get_handoffs(agent)
processed_response = RunImpl.process_model_response(
agent=agent,
response=response,
output_schema=output_schema,
handoffs=handoffs,
)
return await RunImpl.execute_tools_and_side_effects(
agent=agent,
original_input=original_input or "hello",
new_response=response,
pre_step_items=generated_items or [],
processed_response=processed_response,
output_schema=output_schema,
hooks=hooks or RunHooks(),
context_wrapper=context_wrapper or RunContextWrapper(None),
run_config=run_config or RunConfig(),
)