google-agents-cli-extensions 1.14.2__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- core/claude/README.md +35 -0
- core/claude/template/.agents/skills/agents-cli-claude/SKILL.md +106 -0
- core/claude/template/.agents/skills/agents-cli-claude/references/claude.md +386 -0
- core/claude/template/.agents/skills/agents-cli-claude/references/migrate-claude-agent.md +318 -0
- core/claude/template/.agents/skills/agents-cli-claude/references/migrate-from-aws.md +132 -0
- core/claude/template/.dockerignore +28 -0
- core/claude/template/.env +18 -0
- core/claude/template/.env.example +17 -0
- core/claude/template/.template/templateconfig.yaml +31 -0
- core/claude/template/AGENTS.md +33 -0
- core/claude/template/Dockerfile +55 -0
- core/claude/template/agent_config/README.md +9 -0
- core/claude/template/agents-cli-extension.yaml +49 -0
- core/claude/template/app/__init__.py +15 -0
- core/claude/template/app/agent.py +49 -0
- core/claude/template/app/app_utils/a2a.py +245 -0
- core/claude/template/app/app_utils/bq_analytics/__init__.py +124 -0
- core/claude/template/app/app_utils/bq_analytics/recorder.py +593 -0
- core/claude/template/app/app_utils/bq_analytics/redaction.py +209 -0
- core/claude/template/app/app_utils/bq_analytics/schema.py +394 -0
- core/claude/template/app/app_utils/bq_analytics/tables.py +104 -0
- core/claude/template/app/app_utils/bq_analytics/uploader.py +188 -0
- core/claude/template/app/app_utils/claude.py +297 -0
- core/claude/template/app/app_utils/completions.py +225 -0
- core/claude/template/app/app_utils/conversation.py +29 -0
- core/claude/template/app/app_utils/memory_bank.py +308 -0
- core/claude/template/app/app_utils/project_settings.py +91 -0
- core/claude/template/app/app_utils/reasoning_engine.py +237 -0
- core/claude/template/app/app_utils/remote_agents.py +349 -0
- core/claude/template/app/app_utils/remote_auth.py +159 -0
- core/claude/template/app/app_utils/replies.py +143 -0
- core/claude/template/app/app_utils/session_store.py +174 -0
- core/claude/template/app/app_utils/settings.py +24 -0
- core/claude/template/app/app_utils/telemetry/__init__.py +114 -0
- core/claude/template/app/app_utils/telemetry/call_counts.py +44 -0
- core/claude/template/app/app_utils/telemetry/cli_relay.py +266 -0
- core/claude/template/app/app_utils/telemetry/config.py +133 -0
- core/claude/template/app/app_utils/telemetry/logs.py +175 -0
- core/claude/template/app/app_utils/telemetry/metrics.py +270 -0
- core/claude/template/app/app_utils/telemetry/request_flush.py +104 -0
- core/claude/template/app/app_utils/telemetry/subagent_spans.py +80 -0
- core/claude/template/app/app_utils/telemetry/traces.py +166 -0
- core/claude/template/app/app_utils/turn_events.py +463 -0
- core/claude/template/app/app_utils/turn_options.py +219 -0
- core/claude/template/app/app_utils/workspaces.py +65 -0
- core/claude/template/app/fast_api_app.py +71 -0
- core/claude/template/app/feature_tools.py +27 -0
- core/claude/template/pyproject.toml +100 -0
- core/claude/template/scripts/eval_generate.py +340 -0
- core/claude/template/scripts/publish.py +66 -0
- core/claude/template/scripts/run.py +85 -0
- core/claude/template/scripts/unsupported.py +48 -0
- core/claude/template/tests/eval/datasets/basic-dataset.json +31 -0
- core/claude/template/tests/eval/eval_config.yaml +17 -0
- core/claude/template/tests/eval/response_quality.py +60 -0
- core/claude/template/tests/integration/test_server_e2e.py +168 -0
- core/claude/template/tests/load_test/.results/.placeholder +0 -0
- core/claude/template/tests/load_test/README.md +24 -0
- core/claude/template/tests/load_test/load_test.py +118 -0
- core/claude/template/tests/unit/conftest.py +63 -0
- core/claude/template/tests/unit/helpers.py +32 -0
- core/claude/template/tests/unit/test_a2a.py +432 -0
- core/claude/template/tests/unit/test_bq_analytics.py +928 -0
- core/claude/template/tests/unit/test_bq_redaction.py +89 -0
- core/claude/template/tests/unit/test_claude.py +825 -0
- core/claude/template/tests/unit/test_completions.py +369 -0
- core/claude/template/tests/unit/test_fast_api_app.py +59 -0
- core/claude/template/tests/unit/test_feature_tools.py +55 -0
- core/claude/template/tests/unit/test_memory_bank.py +625 -0
- core/claude/template/tests/unit/test_reasoning_engine.py +295 -0
- core/claude/template/tests/unit/test_remote_agents.py +517 -0
- core/claude/template/tests/unit/test_telemetry.py +991 -0
- core/claude/template/tests/unit/test_turn_events.py +309 -0
- core/claude/template/tests/unit/test_turn_options.py +394 -0
- core/claude/template/uv.lock +3029 -0
- core/langchain/README.md +54 -0
- core/langchain/template/.agents/skills/agents-cli-langchain/SKILL.md +84 -0
- core/langchain/template/.agents/skills/agents-cli-langchain/references/langchain.md +120 -0
- core/langchain/template/.agents/skills/agents-cli-langchain/references/samples.md +42 -0
- core/langchain/template/.template/templateconfig.yaml +32 -0
- core/langchain/template/AGENTS.md +35 -0
- core/langchain/template/agents-cli-extension.yaml +54 -0
- core/langchain/template/app/__init__.py +19 -0
- core/langchain/template/app/agent.py +61 -0
- core/langchain/template/app/app_utils/a2a.py +176 -0
- core/langchain/template/app/app_utils/content.py +37 -0
- core/langchain/template/app/app_utils/telemetry.py +154 -0
- core/langchain/template/app/fast_api_app.py +68 -0
- core/langchain/template/langgraph.json +7 -0
- core/langchain/template/pyproject.toml +102 -0
- core/langchain/template/scripts/_content.py +40 -0
- core/langchain/template/scripts/eval_generate.py +202 -0
- core/langchain/template/scripts/publish.py +59 -0
- core/langchain/template/scripts/run.py +59 -0
- core/langchain/template/scripts/unsupported.py +48 -0
- core/langchain/template/tests/eval/datasets/basic-dataset.json +16 -0
- core/langchain/template/tests/eval/eval_config.yaml +32 -0
- core/langchain/template/tests/eval/response_quality.py +60 -0
- core/langchain/template/tests/integration/test_server_e2e.py +173 -0
- core/langchain/template/tests/load_test/.results/.placeholder +0 -0
- core/langchain/template/tests/load_test/README.md +20 -0
- core/langchain/template/tests/load_test/load_test.py +82 -0
- core/langchain/template/uv.lock +3820 -0
- google_agents_cli_extensions-1.14.2.dist-info/METADATA +133 -0
- google_agents_cli_extensions-1.14.2.dist-info/RECORD +107 -0
- google_agents_cli_extensions-1.14.2.dist-info/WHEEL +4 -0
- google_agents_cli_extensions-1.14.2.dist-info/licenses/LICENSE +201 -0
|
@@ -0,0 +1,369 @@
|
|
|
1
|
+
# Copyright 2026 Google LLC
|
|
2
|
+
#
|
|
3
|
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
4
|
+
# you may not use this file except in compliance with the License.
|
|
5
|
+
# You may obtain a copy of the License at
|
|
6
|
+
#
|
|
7
|
+
# http://www.apache.org/licenses/LICENSE-2.0
|
|
8
|
+
#
|
|
9
|
+
# Unless required by applicable law or agreed to in writing, software
|
|
10
|
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
11
|
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
12
|
+
# See the License for the specific language governing permissions and
|
|
13
|
+
# limitations under the License.
|
|
14
|
+
|
|
15
|
+
"""Prompt-response logging: what each model call uploads and logs."""
|
|
16
|
+
|
|
17
|
+
from __future__ import annotations
|
|
18
|
+
|
|
19
|
+
import json
|
|
20
|
+
import logging
|
|
21
|
+
import time
|
|
22
|
+
import uuid
|
|
23
|
+
|
|
24
|
+
import fsspec
|
|
25
|
+
import pytest
|
|
26
|
+
from claude_agent_sdk import (
|
|
27
|
+
AssistantMessage,
|
|
28
|
+
ClaudeAgentOptions,
|
|
29
|
+
StreamEvent,
|
|
30
|
+
TextBlock,
|
|
31
|
+
ThinkingBlock,
|
|
32
|
+
ToolResultBlock,
|
|
33
|
+
ToolUseBlock,
|
|
34
|
+
UserMessage,
|
|
35
|
+
)
|
|
36
|
+
from opentelemetry import trace
|
|
37
|
+
from opentelemetry.sdk.trace import TracerProvider
|
|
38
|
+
from opentelemetry.sdk.trace.export import SimpleSpanProcessor
|
|
39
|
+
from opentelemetry.sdk.trace.export.in_memory_span_exporter import (
|
|
40
|
+
InMemorySpanExporter,
|
|
41
|
+
)
|
|
42
|
+
from opentelemetry.util.genai.completion_hook import load_completion_hook
|
|
43
|
+
|
|
44
|
+
from app.app_utils import completions
|
|
45
|
+
from app.app_utils.telemetry.config import agent_name
|
|
46
|
+
from app.app_utils.telemetry.subagent_spans import SubagentSpans
|
|
47
|
+
from app.app_utils.turn_events import TurnParser
|
|
48
|
+
|
|
49
|
+
MODEL = "claude-haiku-4-5"
|
|
50
|
+
|
|
51
|
+
|
|
52
|
+
@pytest.fixture
|
|
53
|
+
def events(monkeypatch):
|
|
54
|
+
"""The completion events sent to Cloud Logging."""
|
|
55
|
+
sent: list = []
|
|
56
|
+
monkeypatch.setattr(completions.telemetry, "emit_event", sent.append)
|
|
57
|
+
return sent
|
|
58
|
+
|
|
59
|
+
|
|
60
|
+
@pytest.fixture
|
|
61
|
+
def upload(monkeypatch):
|
|
62
|
+
"""Turn on the upload hook, writing to an in-memory bucket."""
|
|
63
|
+
base = f"memory://bucket-{uuid.uuid4().hex}/completions"
|
|
64
|
+
monkeypatch.setenv("OTEL_INSTRUMENTATION_GENAI_COMPLETION_HOOK", "upload")
|
|
65
|
+
monkeypatch.setenv("OTEL_INSTRUMENTATION_GENAI_UPLOAD_BASE_PATH", base)
|
|
66
|
+
monkeypatch.setenv("OTEL_INSTRUMENTATION_GENAI_UPLOAD_FORMAT", "jsonl")
|
|
67
|
+
|
|
68
|
+
|
|
69
|
+
@pytest.fixture
|
|
70
|
+
def spans():
|
|
71
|
+
"""Where the turn's spans go: PromptLog gets this provider's tracer."""
|
|
72
|
+
exporter = InMemorySpanExporter()
|
|
73
|
+
provider = TracerProvider()
|
|
74
|
+
provider.add_span_processor(SimpleSpanProcessor(exporter))
|
|
75
|
+
return provider, exporter
|
|
76
|
+
|
|
77
|
+
|
|
78
|
+
def _call(message_id: str, blocks: list, parent: str | None = None, **fields):
|
|
79
|
+
return AssistantMessage(
|
|
80
|
+
content=blocks,
|
|
81
|
+
model=MODEL,
|
|
82
|
+
message_id=message_id,
|
|
83
|
+
parent_tool_use_id=parent,
|
|
84
|
+
**fields,
|
|
85
|
+
)
|
|
86
|
+
|
|
87
|
+
|
|
88
|
+
def _log(messages: list, spans, system_prompt=None) -> None:
|
|
89
|
+
"""Observe `messages` as one turn, under a span standing in for the agent's."""
|
|
90
|
+
provider, _ = spans
|
|
91
|
+
tracer = provider.get_tracer(__name__)
|
|
92
|
+
parser = TurnParser(agent_name())
|
|
93
|
+
turn = completions.PromptLog(
|
|
94
|
+
load_completion_hook(),
|
|
95
|
+
tracer,
|
|
96
|
+
"Weather in SF?",
|
|
97
|
+
ClaudeAgentOptions(system_prompt=system_prompt),
|
|
98
|
+
"conv-1",
|
|
99
|
+
parser.agent_name,
|
|
100
|
+
SubagentSpans(parser.agent_name),
|
|
101
|
+
)
|
|
102
|
+
with tracer.start_as_current_span("turn"):
|
|
103
|
+
for message in messages:
|
|
104
|
+
for event in parser.feed(message):
|
|
105
|
+
turn.handle(event)
|
|
106
|
+
for event in parser.finish():
|
|
107
|
+
turn.handle(event)
|
|
108
|
+
|
|
109
|
+
|
|
110
|
+
def _uploaded(path: str) -> list[dict]:
|
|
111
|
+
"""The JSONL file at `path`, once the hook's upload thread has written it."""
|
|
112
|
+
fs, _ = fsspec.url_to_fs(path)
|
|
113
|
+
deadline = time.monotonic() + 5
|
|
114
|
+
# The memory filesystem creates the file, empty, when it's opened to write.
|
|
115
|
+
while not fs.exists(path) or not fs.cat_file(path):
|
|
116
|
+
assert time.monotonic() < deadline, f"{path} was never uploaded"
|
|
117
|
+
time.sleep(0.01)
|
|
118
|
+
with fs.open(path) as file:
|
|
119
|
+
return [json.loads(line) for line in file.read().splitlines() if line]
|
|
120
|
+
|
|
121
|
+
|
|
122
|
+
def _messages(path: str) -> list[tuple[str, list[dict]]]:
|
|
123
|
+
"""(role, parts) of each message uploaded to `path`."""
|
|
124
|
+
return [(message["role"], message["parts"]) for message in _uploaded(path)]
|
|
125
|
+
|
|
126
|
+
|
|
127
|
+
_TOOL_TURN = [
|
|
128
|
+
_call(
|
|
129
|
+
"m1",
|
|
130
|
+
[
|
|
131
|
+
TextBlock(text="Checking."),
|
|
132
|
+
ToolUseBlock(id="t1", name="get_weather", input={"city": "SF"}),
|
|
133
|
+
],
|
|
134
|
+
stop_reason="tool_use",
|
|
135
|
+
usage={"input_tokens": 20, "output_tokens": 5},
|
|
136
|
+
),
|
|
137
|
+
UserMessage(content=[ToolResultBlock(tool_use_id="t1", content="60F")]),
|
|
138
|
+
_call("m2", [TextBlock(text="It's foggy.")], stop_reason="end_turn"),
|
|
139
|
+
]
|
|
140
|
+
|
|
141
|
+
|
|
142
|
+
def test_each_call_reads_the_turn_so_far(upload, events, spans):
|
|
143
|
+
_log(_TOOL_TURN, spans, system_prompt="You report the weather.")
|
|
144
|
+
|
|
145
|
+
first, second = events
|
|
146
|
+
attributes = first.attributes
|
|
147
|
+
# The log sink matches on this label.
|
|
148
|
+
assert attributes["event.name"] == completions.EVENT_NAME
|
|
149
|
+
assert attributes["gen_ai.request.model"] == MODEL
|
|
150
|
+
assert attributes["gen_ai.conversation.id"] == "conv-1"
|
|
151
|
+
assert attributes["gen_ai.response.finish_reasons"] == ["tool_use"]
|
|
152
|
+
assert attributes["gen_ai.usage.input_tokens"] == 20
|
|
153
|
+
prompt = ("user", [{"type": "text", "content": "Weather in SF?"}])
|
|
154
|
+
assert _messages(attributes["gen_ai.input.messages_ref"]) == [prompt]
|
|
155
|
+
[(_, output)] = _messages(attributes["gen_ai.output.messages_ref"])
|
|
156
|
+
tool_call = {
|
|
157
|
+
"type": "tool_call",
|
|
158
|
+
"arguments": {"city": "SF"},
|
|
159
|
+
"name": "get_weather",
|
|
160
|
+
"id": "t1",
|
|
161
|
+
}
|
|
162
|
+
assert output == [{"type": "text", "content": "Checking."}, tool_call]
|
|
163
|
+
[instruction] = _uploaded(attributes["gen_ai.system_instructions_ref"])
|
|
164
|
+
assert instruction["content"] == "You report the weather."
|
|
165
|
+
# The last call alone holds the whole turn.
|
|
166
|
+
assert _messages(second.attributes["gen_ai.input.messages_ref"]) == [
|
|
167
|
+
prompt,
|
|
168
|
+
("assistant", [{"type": "text", "content": "Checking."}, tool_call]),
|
|
169
|
+
(
|
|
170
|
+
"tool",
|
|
171
|
+
[{"type": "tool_call_response", "response": {"result": "60F"}, "id": "t1"}],
|
|
172
|
+
),
|
|
173
|
+
]
|
|
174
|
+
|
|
175
|
+
|
|
176
|
+
def test_a_custom_system_prompt_is_the_system_instruction(upload, events, spans):
|
|
177
|
+
_log(
|
|
178
|
+
[_call("m1", [TextBlock(text="It's foggy.")])],
|
|
179
|
+
spans,
|
|
180
|
+
system_prompt={"type": "custom", "prompt": "You report the weather."},
|
|
181
|
+
)
|
|
182
|
+
|
|
183
|
+
[event] = events
|
|
184
|
+
[instruction] = _uploaded(event.attributes["gen_ai.system_instructions_ref"])
|
|
185
|
+
assert instruction["content"] == "You report the weather."
|
|
186
|
+
|
|
187
|
+
|
|
188
|
+
def test_each_call_gets_a_chat_span_carrying_its_entry(upload, events, spans):
|
|
189
|
+
_log(_TOOL_TURN, spans)
|
|
190
|
+
|
|
191
|
+
_, exporter = spans
|
|
192
|
+
*calls, turn = exporter.get_finished_spans()
|
|
193
|
+
assert [span.name for span in calls] == [f"chat {MODEL}"] * 2
|
|
194
|
+
assert {span.parent.span_id for span in calls} == {turn.context.span_id}
|
|
195
|
+
for event, span in zip(events, calls, strict=True):
|
|
196
|
+
assert event.context is not None
|
|
197
|
+
span_context = trace.get_current_span(event.context).get_span_context()
|
|
198
|
+
assert span_context.span_id == span.context.span_id
|
|
199
|
+
assert (
|
|
200
|
+
span.attributes["gen_ai.output.messages_ref"]
|
|
201
|
+
== (event.attributes["gen_ai.output.messages_ref"])
|
|
202
|
+
)
|
|
203
|
+
assert calls[0].attributes["gen_ai.usage.input_tokens"] == 20
|
|
204
|
+
|
|
205
|
+
|
|
206
|
+
def test_a_failed_model_call_marks_its_chat_span_as_an_error(upload, events, spans):
|
|
207
|
+
_log([_call("bad", [TextBlock(text="API Error: 429")], error="rate_limit")], spans)
|
|
208
|
+
|
|
209
|
+
_, exporter = spans
|
|
210
|
+
chat, _ = exporter.get_finished_spans()
|
|
211
|
+
assert chat.status.status_code == trace.StatusCode.ERROR
|
|
212
|
+
assert chat.attributes["error.type"] == "rate_limit"
|
|
213
|
+
|
|
214
|
+
|
|
215
|
+
def test_a_subagent_is_logged_from_its_own_prompt(upload, events, spans):
|
|
216
|
+
task = ToolUseBlock(id="task-1", name="Task", input={"prompt": "Summarize"})
|
|
217
|
+
_log(
|
|
218
|
+
[
|
|
219
|
+
# The main agent's calls stream, so this one ends before the subagent runs.
|
|
220
|
+
_event("message_start", message={}),
|
|
221
|
+
_call("m1", [task]),
|
|
222
|
+
_event("message_stop"),
|
|
223
|
+
UserMessage(content="Summarize the forecast.", parent_tool_use_id="task-1"),
|
|
224
|
+
_call("s1", [TextBlock(text="Foggy all day.")], parent="task-1"),
|
|
225
|
+
UserMessage(
|
|
226
|
+
content=[ToolResultBlock(tool_use_id="task-1", content="Foggy.")]
|
|
227
|
+
),
|
|
228
|
+
_call("m2", [TextBlock(text="It's foggy.")]),
|
|
229
|
+
],
|
|
230
|
+
spans,
|
|
231
|
+
system_prompt={
|
|
232
|
+
"type": "preset",
|
|
233
|
+
"preset": "claude_code",
|
|
234
|
+
"append": "Be brief.",
|
|
235
|
+
},
|
|
236
|
+
)
|
|
237
|
+
|
|
238
|
+
agent_call, subagent_call, final_call = events
|
|
239
|
+
assert _messages(subagent_call.attributes["gen_ai.input.messages_ref"]) == [
|
|
240
|
+
("user", [{"type": "text", "content": "Summarize the forecast."}])
|
|
241
|
+
]
|
|
242
|
+
assert "gen_ai.system_instructions_ref" not in subagent_call.attributes
|
|
243
|
+
# Only what the agent adds to a preset prompt is its own.
|
|
244
|
+
[instruction] = _uploaded(agent_call.attributes["gen_ai.system_instructions_ref"])
|
|
245
|
+
assert instruction["content"] == "Be brief."
|
|
246
|
+
roles = [
|
|
247
|
+
role
|
|
248
|
+
for role, _ in _messages(final_call.attributes["gen_ai.input.messages_ref"])
|
|
249
|
+
]
|
|
250
|
+
assert roles == ["user", "assistant", "tool"]
|
|
251
|
+
|
|
252
|
+
|
|
253
|
+
def _event(kind: str, **fields) -> StreamEvent:
|
|
254
|
+
return StreamEvent(uuid="u", session_id="s", event={"type": kind, **fields})
|
|
255
|
+
|
|
256
|
+
|
|
257
|
+
def test_a_streamed_call_takes_its_final_usage_from_the_stream(upload, events, spans):
|
|
258
|
+
_log(
|
|
259
|
+
[
|
|
260
|
+
_event("message_start"),
|
|
261
|
+
_call(
|
|
262
|
+
"m1",
|
|
263
|
+
[ToolUseBlock(id="t1", name="get_weather", input={"city": "SF"})],
|
|
264
|
+
usage={"input_tokens": 20, "output_tokens": 1},
|
|
265
|
+
),
|
|
266
|
+
# The CLI runs the tool before the call's stream has ended.
|
|
267
|
+
UserMessage(content=[ToolResultBlock(tool_use_id="t1", content="60F")]),
|
|
268
|
+
_event(
|
|
269
|
+
"message_delta",
|
|
270
|
+
usage={"output_tokens": 12},
|
|
271
|
+
delta={"stop_reason": "tool_use"},
|
|
272
|
+
),
|
|
273
|
+
_event("message_stop"),
|
|
274
|
+
_call("m2", [TextBlock(text="It's foggy.")]),
|
|
275
|
+
],
|
|
276
|
+
spans,
|
|
277
|
+
)
|
|
278
|
+
|
|
279
|
+
first, second = events
|
|
280
|
+
assert first.attributes["gen_ai.usage.input_tokens"] == 20
|
|
281
|
+
assert first.attributes["gen_ai.usage.output_tokens"] == 12
|
|
282
|
+
assert first.attributes["gen_ai.response.finish_reasons"] == ["tool_use"]
|
|
283
|
+
[(role, _)] = _messages(first.attributes["gen_ai.input.messages_ref"])
|
|
284
|
+
assert role == "user"
|
|
285
|
+
# The tool's result, which came before message_stop, follows the reply.
|
|
286
|
+
roles = [
|
|
287
|
+
role for role, _ in _messages(second.attributes["gen_ai.input.messages_ref"])
|
|
288
|
+
]
|
|
289
|
+
assert roles == ["user", "assistant", "tool"]
|
|
290
|
+
|
|
291
|
+
|
|
292
|
+
@pytest.mark.parametrize(
|
|
293
|
+
("content", "is_error", "response"),
|
|
294
|
+
[
|
|
295
|
+
(
|
|
296
|
+
[{"type": "text", "text": "60"}, {"type": "text", "text": "F"}],
|
|
297
|
+
None,
|
|
298
|
+
{"result": "60F"},
|
|
299
|
+
),
|
|
300
|
+
("city not found", True, {"error": "city not found"}),
|
|
301
|
+
],
|
|
302
|
+
)
|
|
303
|
+
def test_tool_results_go_up_as_objects(
|
|
304
|
+
upload, events, spans, content, is_error, response
|
|
305
|
+
):
|
|
306
|
+
_log(
|
|
307
|
+
[
|
|
308
|
+
_call("m1", [ToolUseBlock(id="t1", name="get_weather", input={})]),
|
|
309
|
+
UserMessage(
|
|
310
|
+
content=[
|
|
311
|
+
ToolResultBlock(
|
|
312
|
+
tool_use_id="t1", content=content, is_error=is_error
|
|
313
|
+
)
|
|
314
|
+
]
|
|
315
|
+
),
|
|
316
|
+
_call("m2", [TextBlock(text="Done.")]),
|
|
317
|
+
],
|
|
318
|
+
spans,
|
|
319
|
+
)
|
|
320
|
+
|
|
321
|
+
*_, (role, parts) = _messages(events[1].attributes["gen_ai.input.messages_ref"])
|
|
322
|
+
assert (role, parts[0]["response"]) == ("tool", response)
|
|
323
|
+
|
|
324
|
+
|
|
325
|
+
def test_one_model_call_in_several_messages_is_one_event(upload, events, spans):
|
|
326
|
+
_log(
|
|
327
|
+
[
|
|
328
|
+
_call("m1", [ThinkingBlock(thinking="SF is foggy.", signature="s")]),
|
|
329
|
+
_call("m1", [TextBlock(text="It's foggy.")]),
|
|
330
|
+
],
|
|
331
|
+
spans,
|
|
332
|
+
)
|
|
333
|
+
|
|
334
|
+
[event] = events
|
|
335
|
+
[(_, output)] = _messages(event.attributes["gen_ai.output.messages_ref"])
|
|
336
|
+
assert output == [
|
|
337
|
+
{"type": "text", "content": "SF is foggy."},
|
|
338
|
+
{"type": "text", "content": "It's foggy."},
|
|
339
|
+
]
|
|
340
|
+
|
|
341
|
+
|
|
342
|
+
def test_off_unless_the_upload_hook_is_configured(monkeypatch):
|
|
343
|
+
monkeypatch.delenv("OTEL_INSTRUMENTATION_GENAI_COMPLETION_HOOK", raising=False)
|
|
344
|
+
names = TurnParser(agent_name()).agent_name
|
|
345
|
+
spans = SubagentSpans(names)
|
|
346
|
+
assert (
|
|
347
|
+
completions.start_turn("hi", ClaudeAgentOptions(), "conv-1", names, spans)
|
|
348
|
+
is None
|
|
349
|
+
)
|
|
350
|
+
|
|
351
|
+
|
|
352
|
+
@pytest.mark.parametrize(
|
|
353
|
+
"calls",
|
|
354
|
+
[
|
|
355
|
+
pytest.param(["m1"], id="at the end of the turn"),
|
|
356
|
+
pytest.param(["m1", "m2"], id="when the next call starts"),
|
|
357
|
+
],
|
|
358
|
+
)
|
|
359
|
+
def test_a_failure_never_fails_the_turn(upload, spans, monkeypatch, caplog, calls):
|
|
360
|
+
def refuse(record):
|
|
361
|
+
raise ConnectionError("logging.googleapis.com unreachable")
|
|
362
|
+
|
|
363
|
+
monkeypatch.setattr(completions.telemetry, "emit_event", refuse)
|
|
364
|
+
with caplog.at_level(logging.WARNING):
|
|
365
|
+
_log([_call(message_id, [TextBlock(text="hi")]) for message_id in calls], spans)
|
|
366
|
+
|
|
367
|
+
assert [r.getMessage() for r in caplog.records] == ["Couldn't log a model call"]
|
|
368
|
+
_, exporter = spans
|
|
369
|
+
assert [s.name for s in exporter.get_finished_spans()] == [f"chat {MODEL}", "turn"]
|
|
@@ -0,0 +1,59 @@
|
|
|
1
|
+
# Copyright 2026 Google LLC
|
|
2
|
+
#
|
|
3
|
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
4
|
+
# you may not use this file except in compliance with the License.
|
|
5
|
+
# You may obtain a copy of the License at
|
|
6
|
+
#
|
|
7
|
+
# http://www.apache.org/licenses/LICENSE-2.0
|
|
8
|
+
#
|
|
9
|
+
# Unless required by applicable law or agreed to in writing, software
|
|
10
|
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
11
|
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
12
|
+
# See the License for the specific language governing permissions and
|
|
13
|
+
# limitations under the License.
|
|
14
|
+
|
|
15
|
+
"""The local server: `app/fast_api_app.py` and the `.env` it loads."""
|
|
16
|
+
|
|
17
|
+
import os
|
|
18
|
+
import shutil
|
|
19
|
+
import subprocess
|
|
20
|
+
import sys
|
|
21
|
+
from pathlib import Path
|
|
22
|
+
|
|
23
|
+
_APP_DIR = Path(__file__).resolve().parents[2] / "app"
|
|
24
|
+
|
|
25
|
+
|
|
26
|
+
def test_features_turned_on_only_in_dotenv_add_their_tools(tmp_path):
|
|
27
|
+
# A copy of the app next to its own .env, run in a fresh interpreter: the
|
|
28
|
+
# tool lists are fixed when app.agent is first imported.
|
|
29
|
+
shutil.copytree(
|
|
30
|
+
_APP_DIR, tmp_path / "app", ignore=shutil.ignore_patterns("__pycache__")
|
|
31
|
+
)
|
|
32
|
+
(tmp_path / ".env").write_text(
|
|
33
|
+
"MEMORY_BANK_ENABLED=true\nGOOGLE_CLOUD_AGENT_ENGINE_ENABLE_TELEMETRY=false\n"
|
|
34
|
+
)
|
|
35
|
+
env = {
|
|
36
|
+
name: value
|
|
37
|
+
for name, value in os.environ.items()
|
|
38
|
+
if name
|
|
39
|
+
not in ("MEMORY_BANK_ENABLED", "GOOGLE_CLOUD_AGENT_ENGINE_ENABLE_TELEMETRY")
|
|
40
|
+
}
|
|
41
|
+
env["PYTHONPATH"] = str(tmp_path)
|
|
42
|
+
|
|
43
|
+
result = subprocess.run(
|
|
44
|
+
[
|
|
45
|
+
sys.executable,
|
|
46
|
+
"-c",
|
|
47
|
+
"import app.fast_api_app\n"
|
|
48
|
+
"from app.agent import root_agent\n"
|
|
49
|
+
"print(app.__file__, root_agent.allowed_tools)",
|
|
50
|
+
],
|
|
51
|
+
cwd=tmp_path,
|
|
52
|
+
env=env,
|
|
53
|
+
capture_output=True,
|
|
54
|
+
text=True,
|
|
55
|
+
check=True,
|
|
56
|
+
)
|
|
57
|
+
|
|
58
|
+
assert result.stdout.startswith(str(tmp_path / "app" / "__init__.py"))
|
|
59
|
+
assert "mcp__memory__search_memory" in result.stdout
|
|
@@ -0,0 +1,55 @@
|
|
|
1
|
+
# Copyright 2026 Google LLC
|
|
2
|
+
#
|
|
3
|
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
4
|
+
# you may not use this file except in compliance with the License.
|
|
5
|
+
# You may obtain a copy of the License at
|
|
6
|
+
#
|
|
7
|
+
# http://www.apache.org/licenses/LICENSE-2.0
|
|
8
|
+
#
|
|
9
|
+
# Unless required by applicable law or agreed to in writing, software
|
|
10
|
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
11
|
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
12
|
+
# See the License for the specific language governing permissions and
|
|
13
|
+
# limitations under the License.
|
|
14
|
+
|
|
15
|
+
"""The tools `app/feature_tools.py` adds to root_agent for the features that are on."""
|
|
16
|
+
|
|
17
|
+
import importlib
|
|
18
|
+
|
|
19
|
+
import pytest
|
|
20
|
+
|
|
21
|
+
from app import feature_tools
|
|
22
|
+
|
|
23
|
+
|
|
24
|
+
@pytest.fixture
|
|
25
|
+
def reload_feature_tools(monkeypatch):
|
|
26
|
+
"""Re-evaluates feature_tools, which reads the env at import, under the test's env.
|
|
27
|
+
|
|
28
|
+
Afterwards it is evaluated again under the original env, so no other test
|
|
29
|
+
sees the test's tools.
|
|
30
|
+
"""
|
|
31
|
+
yield lambda: importlib.reload(feature_tools)
|
|
32
|
+
monkeypatch.undo()
|
|
33
|
+
importlib.reload(feature_tools)
|
|
34
|
+
|
|
35
|
+
|
|
36
|
+
@pytest.mark.parametrize(
|
|
37
|
+
"enabled, servers, tools",
|
|
38
|
+
[
|
|
39
|
+
("true", ["memory"], ["mcp__memory__search_memory"]),
|
|
40
|
+
(None, [], []),
|
|
41
|
+
],
|
|
42
|
+
ids=["memory_on", "memory_off"],
|
|
43
|
+
)
|
|
44
|
+
def test_memory_bank_adds_its_search_tool_only_while_on(
|
|
45
|
+
monkeypatch, reload_feature_tools, enabled, servers, tools
|
|
46
|
+
):
|
|
47
|
+
if enabled is None:
|
|
48
|
+
monkeypatch.delenv("MEMORY_BANK_ENABLED", raising=False)
|
|
49
|
+
else:
|
|
50
|
+
monkeypatch.setenv("MEMORY_BANK_ENABLED", enabled)
|
|
51
|
+
|
|
52
|
+
module = reload_feature_tools()
|
|
53
|
+
|
|
54
|
+
assert list(module.FEATURE_MCP_SERVERS) == servers
|
|
55
|
+
assert module.FEATURE_TOOLS == tools
|