Spaces:
Paused
Paused
File size: 37,748 Bytes
4b03eed | 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 | # -*- coding: utf-8 -*-
# pylint: disable=abstract-method
"""Unit tests for middleware system."""
from unittest.async_case import IsolatedAsyncioTestCase
from unittest.mock import AsyncMock, patch
from typing import Any, AsyncGenerator, Callable, Union
from utils import MockModel
from pydantic import BaseModel
from agentscope.event import AgentEvent
from agentscope.agent import Agent, ContextConfig
from agentscope.middleware import MiddlewareBase
from agentscope.model import ChatResponse
from agentscope.message import (
TextBlock,
HintBlock,
UserMsg,
SystemMsg,
Msg,
ToolCallBlock,
)
from agentscope.tool import Toolkit, ToolBase, ToolChunk
from agentscope.permission import (
PermissionContext,
PermissionDecision,
PermissionBehavior,
)
class TestMiddleware(IsolatedAsyncioTestCase):
"""Test cases for middleware system."""
async def asyncSetUp(self) -> None:
"""Set up test fixtures."""
self.mock_model = MockModel()
self.toolkit = Toolkit()
self.execution_log = []
async def test_on_reply_middleware_pre_post_yield(self) -> None:
"""Test on_reply middleware pre, post and yield positions."""
class ReplyMiddleware(MiddlewareBase):
"""Middleware for testing on_reply hook."""
def __init__(self, log: list, name: str) -> None:
"""Initialize the reply middleware.
Args:
log: The execution log list.
name: The middleware name.
"""
self.log = log
self.name = name
async def on_reply(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable[[], AsyncGenerator],
) -> AsyncGenerator:
"""The on_reply middleware logic."""
self.log.append(f"{self.name}_pre")
async for item in next_handler():
if isinstance(item, AgentEvent):
self.log.append(f"{self.name}_{item.type}")
elif isinstance(item, Msg):
self.log.append(f"{self.name}_msg")
yield item
self.log.append(f"{self.name}_post")
middleware1 = ReplyMiddleware(self.execution_log, "mw1")
middleware2 = ReplyMiddleware(self.execution_log, "mw2")
self.mock_model.set_responses(
[
ChatResponse(
content=[TextBlock(text="test response")],
is_last=True,
),
],
)
agent = Agent(
name="test_agent",
system_prompt="test prompt",
model=self.mock_model,
toolkit=self.toolkit,
middlewares=[middleware1, middleware2],
)
await agent.reply(UserMsg("user", "test message"))
# Verify execution order
expected = [
"mw1_pre",
"mw2_pre",
"mw2_REPLY_START",
"mw1_REPLY_START",
"mw2_MODEL_CALL_START",
"mw1_MODEL_CALL_START",
"mw2_TEXT_BLOCK_START",
"mw1_TEXT_BLOCK_START",
"mw2_TEXT_BLOCK_DELTA",
"mw1_TEXT_BLOCK_DELTA",
"mw2_TEXT_BLOCK_END",
"mw1_TEXT_BLOCK_END",
"mw2_MODEL_CALL_END",
"mw1_MODEL_CALL_END",
"mw2_REPLY_END",
"mw1_REPLY_END",
"mw2_msg",
"mw1_msg",
"mw2_post",
"mw1_post",
]
self.assertListEqual(self.execution_log, expected)
async def test_on_reasoning_middleware_pre_yield(self) -> None:
"""Test on_reasoning middleware pre and yield positions."""
class ReasoningMiddleware(MiddlewareBase):
"""Middleware for testing on_reasoning hook."""
def __init__(self, log: list, name: str) -> None:
"""Initialize the reasoning middleware.
Args:
log: The execution log list.
name: The middleware name.
"""
self.log = log
self.name = name
async def on_reasoning(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable[[], AsyncGenerator],
) -> AsyncGenerator:
"""The on_reasoning middleware logic."""
self.log.append(f"{self.name}_pre")
async for item in next_handler():
if isinstance(item, AgentEvent):
self.log.append(f"{self.name}_{item.type}")
elif isinstance(item, Msg):
self.log.append(f"{self.name}_msg")
yield item
self.log.append(f"{self.name}_post")
middleware1 = ReasoningMiddleware(self.execution_log, "mw1")
middleware2 = ReasoningMiddleware(self.execution_log, "mw2")
self.mock_model.set_responses(
[
ChatResponse(
content=[TextBlock(text="test response")],
is_last=True,
),
],
)
agent = Agent(
name="test_agent",
system_prompt="test prompt",
model=self.mock_model,
toolkit=self.toolkit,
middlewares=[middleware1, middleware2],
)
await agent.reply(UserMsg("user", "test message"))
# Verify execution order
expected = [
"mw1_pre",
"mw2_pre",
"mw2_MODEL_CALL_START",
"mw1_MODEL_CALL_START",
"mw2_TEXT_BLOCK_START",
"mw1_TEXT_BLOCK_START",
"mw2_TEXT_BLOCK_DELTA",
"mw1_TEXT_BLOCK_DELTA",
"mw2_TEXT_BLOCK_END",
"mw1_TEXT_BLOCK_END",
"mw2_MODEL_CALL_END",
"mw1_MODEL_CALL_END",
"mw2_msg",
"mw1_msg",
]
self.assertListEqual(self.execution_log, expected)
async def test_on_model_call_middleware_non_streaming(self) -> None:
"""Test on_model_call middleware for non-streaming model."""
class ModelCallMiddleware(MiddlewareBase):
"""Middleware for testing on_model_call hook."""
def __init__(self, log: list, name: str) -> None:
"""Initialize the model call middleware.
Args:
log: The execution log list.
name: The middleware name.
"""
self.log = log
self.name = name
async def on_model_call(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable,
) -> Union[ChatResponse, AsyncGenerator[ChatResponse, None]]:
"""The on_model_call middleware logic."""
self.log.append(f"{self.name}_pre")
result = await next_handler()
self.log.append(f"{self.name}_post")
return result
middleware1 = ModelCallMiddleware(self.execution_log, "mw1")
middleware2 = ModelCallMiddleware(self.execution_log, "mw2")
self.mock_model.set_responses(
[
ChatResponse(
content=[TextBlock(text="test response")],
is_last=True,
),
],
)
agent = Agent(
name="test_agent",
system_prompt="test prompt",
model=self.mock_model,
toolkit=self.toolkit,
middlewares=[middleware1, middleware2],
)
await agent.reply(UserMsg("user", "test message"))
# Verify execution order: mw1_pre -> mw2_pre -> mw2_post -> mw1_post
expected = ["mw1_pre", "mw2_pre", "mw2_post", "mw1_post"]
self.assertListEqual(self.execution_log, expected)
async def test_on_model_call_middleware_streaming(self) -> None:
"""Test on_model_call middleware for streaming model."""
class ModelCallMiddleware(MiddlewareBase):
"""Middleware for testing on_model_call hook with streaming."""
def __init__(self, log: list, name: str) -> None:
"""Initialize the model call middleware.
Args:
log: The execution log list.
name: The middleware name.
"""
self.log = log
self.name = name
async def on_model_call(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable,
) -> Union[ChatResponse, AsyncGenerator[ChatResponse, None]]:
"""The on_model_call middleware logic for streaming."""
self.log.append(f"{self.name}_pre")
result = await next_handler()
async def wrapped_generator() -> AsyncGenerator[
ChatResponse,
None,
]:
"""Wrap the generator to log yields."""
async for chunk in result:
self.log.append(f"{self.name}_chunk")
yield chunk
self.log.append(f"{self.name}_post")
return wrapped_generator()
middleware1 = ModelCallMiddleware(self.execution_log, "mw1")
middleware2 = ModelCallMiddleware(self.execution_log, "mw2")
self.mock_model.set_responses(
[
[
ChatResponse(
content=[TextBlock(text="chunk1")],
is_last=False,
),
ChatResponse(
content=[TextBlock(text="chunk2")],
is_last=False,
),
ChatResponse(
content=[TextBlock(text="chunk3")],
is_last=True,
),
],
],
)
agent = Agent(
name="test_agent",
system_prompt="test prompt",
model=self.mock_model,
toolkit=self.toolkit,
middlewares=[middleware1, middleware2],
)
await agent.reply(UserMsg("user", "test message"))
# Verify execution order
expected = [
"mw1_pre",
"mw2_pre",
"mw2_chunk",
"mw1_chunk",
"mw2_chunk",
"mw1_chunk",
"mw2_chunk",
"mw1_chunk",
"mw2_post",
"mw1_post",
]
self.assertListEqual(self.execution_log, expected)
async def test_on_system_prompt_middleware(self) -> None:
"""Test on_system_prompt middleware (transformer pattern)."""
class SystemPromptMiddleware(MiddlewareBase):
"""Middleware for testing on_system_prompt hook."""
def __init__(self, log: list, name: str, suffix: str) -> None:
"""Initialize the system prompt middleware.
Args:
log: The execution log list.
name: The middleware name.
suffix: The suffix to append to the prompt.
"""
self.log = log
self.name = name
self.suffix = suffix
async def on_system_prompt(
self,
agent: Agent,
current_prompt: str,
) -> str:
"""The on_system_prompt middleware logic."""
self.log.append(f"{self.name}_executed")
return f"{current_prompt} {self.suffix}"
middleware1 = SystemPromptMiddleware(
self.execution_log,
"mw1",
"[MW1]",
)
middleware2 = SystemPromptMiddleware(
self.execution_log,
"mw2",
"[MW2]",
)
self.mock_model.set_responses(
[
ChatResponse(
content=[TextBlock(text="test response")],
is_last=True,
),
],
)
agent = Agent(
name="test_agent",
system_prompt="test prompt",
model=self.mock_model,
toolkit=self.toolkit,
middlewares=[middleware1, middleware2],
)
await agent.reply(UserMsg("user", "test message"))
# Verify execution order: mw1 -> mw2 (sequential transformer pattern)
# Note: system_prompt is called twice (once for initial setup, once
# during reasoning)
expected = [
"mw1_executed",
"mw2_executed", # First call
"mw1_executed",
"mw2_executed", # Second call during reasoning
]
self.assertListEqual(self.execution_log, expected)
async def test_multiple_middleware_types(self) -> None:
"""Test multiple middleware types working together."""
class MultiMiddleware(MiddlewareBase):
"""Middleware implementing multiple hooks."""
def __init__(self, log: list, name: str) -> None:
"""Initialize the multi middleware.
Args:
log: The execution log list.
name: The middleware name.
"""
self.log = log
self.name = name
async def on_reply(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable[[], AsyncGenerator],
) -> AsyncGenerator:
"""The on_reply middleware logic."""
self.log.append(f"{self.name}_reply_pre")
async for item in next_handler():
yield item
self.log.append(f"{self.name}_reply_post")
async def on_reasoning(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable[[], AsyncGenerator],
) -> AsyncGenerator:
"""The on_reasoning middleware logic."""
self.log.append(f"{self.name}_reasoning_pre")
async for item in next_handler():
yield item
self.log.append(f"{self.name}_reasoning_post")
async def on_model_call(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable,
) -> Union[ChatResponse, AsyncGenerator[ChatResponse, None]]:
"""The on_model_call middleware logic."""
self.log.append(f"{self.name}_model_call_pre")
result = await next_handler()
self.log.append(f"{self.name}_model_call_post")
return result
async def on_system_prompt(
self,
agent: Agent,
current_prompt: str,
) -> str:
"""The on_system_prompt middleware logic."""
self.log.append(f"{self.name}_system_prompt")
return current_prompt
middleware = MultiMiddleware(self.execution_log, "multi")
self.mock_model.set_responses(
[
ChatResponse(
content=[TextBlock(text="test response")],
is_last=True,
),
],
)
agent = Agent(
name="test_agent",
system_prompt="test prompt",
model=self.mock_model,
toolkit=self.toolkit,
middlewares=[middleware],
)
await agent.reply(UserMsg("user", "test message"))
# Verify all middleware hooks were called
expected = [
"multi_reply_pre",
"multi_system_prompt",
"multi_reasoning_pre",
"multi_system_prompt",
"multi_model_call_pre",
"multi_model_call_post",
"multi_reply_post",
]
self.assertListEqual(self.execution_log, expected)
async def test_on_reply_middleware_modify_input(self) -> None:
"""Test that on_reply middleware can modify msgs input."""
class ModifyMsgsMiddleware(MiddlewareBase):
"""Middleware that modifies the msgs input."""
async def on_reply(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable[..., AsyncGenerator],
) -> AsyncGenerator:
"""Modify inputs before passing to next handler."""
# Modify the message content
inputs = input_kwargs["inputs"]
if isinstance(inputs, Msg):
modified_msg = UserMsg(
name=inputs.name,
content="MODIFIED: " + inputs.get_text_content(),
)
async for item in next_handler(inputs=modified_msg):
yield item
else:
async for item in next_handler(**input_kwargs):
yield item
middleware = ModifyMsgsMiddleware()
# Track what message the model receives
received_messages = []
class TrackingModel(MockModel):
"""Model that tracks received messages."""
async def _call_api(
self,
*args: Any,
**kwargs: Any,
) -> ChatResponse:
"""Track the messages and return mock response."""
messages = kwargs.get("messages", [])
received_messages.extend(messages)
return await super()._call_api(*args, **kwargs)
tracking_model = TrackingModel()
tracking_model.set_responses(
[
ChatResponse(
content=[TextBlock(text="response")],
is_last=True,
),
],
)
agent_instance = Agent(
name="test_agent",
system_prompt="test prompt",
model=tracking_model,
toolkit=self.toolkit,
middlewares=[middleware],
)
await agent_instance.reply(UserMsg("user", "original message"))
# Verify the model received the modified message
user_messages = [m for m in received_messages if m.role == "user"]
self.assertTrue(len(user_messages) > 0)
self.assertIn(
"MODIFIED: original message",
user_messages[-1].get_text_content(),
)
async def test_on_reasoning_middleware_modify_input(self) -> None:
"""Test that on_reasoning middleware can modify tool_choice input."""
class ModifyToolChoiceMiddleware(MiddlewareBase):
"""Middleware that modifies the tool_choice input."""
async def on_reasoning(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable[..., AsyncGenerator],
) -> AsyncGenerator:
"""Force tool_choice to 'none' to prevent tool calls."""
# Override tool_choice to 'none'
async for item in next_handler(tool_choice="none"):
yield item
middleware = ModifyToolChoiceMiddleware()
# Track what tool_choice the model receives
received_tool_choices = []
class TrackingModel(MockModel):
"""Model that tracks received tool_choice."""
async def _call_api(
self,
*args: Any,
**kwargs: Any,
) -> ChatResponse:
"""Track the tool_choice and return mock response."""
tool_choice = kwargs.get("tool_choice")
received_tool_choices.append(tool_choice)
return await super()._call_api(*args, **kwargs)
tracking_model = TrackingModel()
tracking_model.set_responses(
[
ChatResponse(
content=[TextBlock(text="response without tools")],
is_last=True,
),
],
)
agent_instance = Agent(
name="test_agent",
system_prompt="test prompt",
model=tracking_model,
toolkit=self.toolkit,
middlewares=[middleware],
)
await agent_instance.reply(UserMsg("user", "test message"))
# Verify the model received tool_choice='none'
self.assertIn("none", received_tool_choices)
async def test_on_acting_middleware_intercepts_tool_execution(
self,
) -> None:
"""Test that on_acting middleware intercepts raw tool execution.
After the refactor, ``on_acting`` wraps only ``_acting_impl``
(i.e. ``toolkit.call_tool``). Permission checking and context
writes are handled by ``_execute_tool_call`` *outside* the hook.
This test verifies that the middleware can observe and modify the
``tool_call`` passed to the actual tool function.
"""
# ------------------------------------------------------------------ #
# A minimal tool that records the raw input it receives. #
# ------------------------------------------------------------------ #
received_inputs: list[str] = []
class _EchoParams(BaseModel):
value: str
class EchoTool(ToolBase):
"""Tool that echoes its input and records it."""
name: str = "echo"
description: str = "Echo the value."
input_schema: dict = _EchoParams.model_json_schema()
is_concurrency_safe: bool = True
is_read_only: bool = True
is_state_injected: bool = False
is_external_tool: bool = False
is_mcp: bool = False
mcp_name: str | None = None
async def check_permissions(
self,
tool_input: dict[str, Any],
context: PermissionContext,
) -> PermissionDecision:
"""Always allow."""
return PermissionDecision(
behavior=PermissionBehavior.ALLOW,
message="allowed",
)
async def __call__(
self,
value: str,
) -> ToolChunk:
"""Record the value and return it."""
received_inputs.append(value)
return ToolChunk(
content=[TextBlock(text=f"echo:{value}")],
)
toolkit_with_tool = Toolkit(tools=[EchoTool()])
# ------------------------------------------------------------------ #
# Middleware that renames the tool_call.input before forwarding. #
# ------------------------------------------------------------------ #
intercepted_tool_calls: list[str] = []
class ObserveActingMiddleware(MiddlewareBase):
"""Middleware that records the tool_call seen at acting level."""
async def on_acting(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable[..., AsyncGenerator],
) -> AsyncGenerator:
"""Record the tool_call and forward to next handler."""
tool_call = input_kwargs["tool_call"]
intercepted_tool_calls.append(tool_call.input)
# Modify the input before execution
import json
modified = ToolCallBlock(
id=tool_call.id,
name=tool_call.name,
input=json.dumps({"value": "MODIFIED"}),
state=tool_call.state,
)
async for chunk in next_handler(tool_call=modified):
yield chunk
middleware = ObserveActingMiddleware()
self.mock_model.set_responses(
[
ChatResponse(
content=[TextBlock(text="done")],
is_last=True,
),
],
)
agent_instance = Agent(
name="test_agent",
system_prompt="test prompt",
model=self.mock_model,
toolkit=toolkit_with_tool,
middlewares=[middleware],
)
# Call _execute_tool_call with a valid tool call.
tool_call = ToolCallBlock(
id="call_1",
name="echo",
input='{"value": "ORIGINAL"}',
)
events = []
# pylint: disable=protected-access
async for evt in agent_instance._execute_tool_call(tool_call):
events.append(evt)
# Middleware intercepted the tool call at execution level
self.assertEqual(len(intercepted_tool_calls), 1)
self.assertIn("ORIGINAL", intercepted_tool_calls[0])
# The tool actually received the MODIFIED value
self.assertEqual(len(received_inputs), 1)
self.assertEqual(received_inputs[0], "MODIFIED")
async def test_on_model_call_middleware_modify_input(self) -> None:
"""Test that on_model_call middleware can modify messages and model."""
class ModifyMessagesMiddleware(MiddlewareBase):
"""Middleware that modifies messages input."""
async def on_model_call(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable[..., Any],
) -> Union[ChatResponse, AsyncGenerator[ChatResponse, None]]:
"""Prepend a system message to the messages list."""
messages = input_kwargs["messages"]
modified_messages = [
SystemMsg(
name="system",
content="INJECTED SYSTEM MESSAGE",
),
] + messages
# Pass modified messages to next handler
return await next_handler(
current_model=input_kwargs["current_model"],
messages=modified_messages,
tools=input_kwargs["tools"],
tool_choice=input_kwargs["tool_choice"],
)
middleware = ModifyMessagesMiddleware()
# Track what messages the model receives
received_messages = []
class TrackingModel(MockModel):
"""Model that tracks received messages."""
async def _call_api(
self,
*args: Any,
**kwargs: Any,
) -> ChatResponse:
"""Track the messages and return mock response."""
messages = kwargs.get("messages", [])
received_messages.extend(messages)
return await super()._call_api(*args, **kwargs)
tracking_model = TrackingModel()
tracking_model.set_responses(
[
ChatResponse(
content=[TextBlock(text="response")],
is_last=True,
),
],
)
agent_instance = Agent(
name="test_agent",
system_prompt="test prompt",
model=tracking_model,
toolkit=self.toolkit,
middlewares=[middleware],
)
await agent_instance.reply(UserMsg("user", "test message"))
# Verify the injected system message is present
system_messages = [m for m in received_messages if m.role == "system"]
self.assertTrue(
any(
"INJECTED SYSTEM MESSAGE" in m.get_text_content()
for m in system_messages
),
)
async def test_on_compress_context_middleware(self) -> None:
"""Test on_compress_context middleware follows the onion chain pattern.
Verifies that:
- Multiple middlewares are chained in onion order (mw1 wraps mw2).
- ``input_kwargs`` carries the correct ``context_config`` and
``instructions``.
- The ``next_handler`` ultimately calls ``_compress_context_impl``.
- A middleware can short-circuit and skip the actual implementation.
"""
seen_instructions = []
# ------------------------------------------------------------------ #
# Middleware that records pre/post and forwards to next_handler. #
# ------------------------------------------------------------------ #
class CompressContextMiddleware(MiddlewareBase):
"""Middleware for testing on_compress_context hook."""
def __init__(self, log: list, name: str) -> None:
"""Initialize the compress context middleware.
Args:
log (`list`):
The execution log list.
name (`str`):
The middleware name.
"""
self.log = log
self.name = name
async def on_compress_context(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable[..., Any],
) -> None:
"""Forward to next handler, recording pre and post."""
self.log.append(f"{self.name}_pre")
seen_instructions.append(input_kwargs.get("instructions"))
await next_handler(**input_kwargs)
self.log.append(f"{self.name}_post")
middleware1 = CompressContextMiddleware(self.execution_log, "mw1")
middleware2 = CompressContextMiddleware(self.execution_log, "mw2")
self.mock_model.set_responses(
[
ChatResponse(
content=[TextBlock(text="test response")],
is_last=True,
),
],
)
context_config = ContextConfig(trigger_ratio=0.8, reserve_ratio=0.1)
agent = Agent(
name="test_agent",
system_prompt="test prompt",
model=self.mock_model,
toolkit=self.toolkit,
middlewares=[middleware1, middleware2],
context_config=context_config,
)
# Patch _compress_context_impl to avoid real token counting.
instructions = HintBlock(
hint="Keep user requirements while compressing.",
source="user",
)
with patch.object(
agent,
"_compress_context_impl",
new_callable=AsyncMock,
) as mock_impl:
await agent.compress_context(
context_config=context_config,
instructions=instructions,
)
# _compress_context_impl must have been called exactly once.
mock_impl.assert_awaited_once_with(
context_config=context_config,
instructions=instructions,
)
# Verify onion execution order: mw1_pre -> mw2_pre -> mw2_post ->
# mw1_post
expected = ["mw1_pre", "mw2_pre", "mw2_post", "mw1_post"]
self.assertListEqual(self.execution_log, expected)
self.assertListEqual(seen_instructions, [instructions, instructions])
async def test_on_compress_context_middleware_modify_instructions(
self,
) -> None:
"""Test that middleware can replace compress_context instructions."""
class ReplaceInstructionsMiddleware(MiddlewareBase):
"""Middleware that replaces the compression instructions."""
def __init__(self, replacement: HintBlock) -> None:
"""Initialize the middleware with replacement instructions."""
self.replacement = replacement
async def on_compress_context(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable[..., Any],
) -> None:
"""Replace instructions before forwarding."""
input_kwargs["instructions"] = self.replacement
await next_handler(**input_kwargs)
original = HintBlock(
hint="Keep all requirements.",
source="user",
)
replacement = HintBlock(
hint="Keep only unresolved requirements.",
source="middleware",
)
context_config = ContextConfig(trigger_ratio=0.8, reserve_ratio=0.1)
agent = Agent(
name="test_agent",
system_prompt="test prompt",
model=self.mock_model,
toolkit=self.toolkit,
middlewares=[ReplaceInstructionsMiddleware(replacement)],
context_config=context_config,
)
with patch.object(
agent,
"_compress_context_impl",
new_callable=AsyncMock,
) as mock_impl:
await agent.compress_context(
context_config=context_config,
instructions=original,
)
mock_impl.assert_awaited_once_with(
context_config=context_config,
instructions=replacement,
)
async def test_on_compress_context_middleware_short_circuit(
self,
) -> None:
"""Test that a middleware can skip _compress_context_impl entirely."""
class SkipCompressMiddleware(MiddlewareBase):
"""Middleware that skips the actual compress_context call."""
def __init__(self, log: list) -> None:
"""Initialize the skip compress middleware.
Args:
log (`list`):
The execution log list.
"""
self.log = log
async def on_compress_context(
self,
agent: Agent,
input_kwargs: dict,
next_handler: Callable[..., Any],
) -> None:
"""Record the call and skip forwarding to next_handler."""
self.log.append("skipped")
# Intentionally NOT calling next_handler.
middleware = SkipCompressMiddleware(self.execution_log)
self.mock_model.set_responses(
[
ChatResponse(
content=[TextBlock(text="test response")],
is_last=True,
),
],
)
agent = Agent(
name="test_agent",
system_prompt="test prompt",
model=self.mock_model,
toolkit=self.toolkit,
middlewares=[middleware],
)
with patch.object(
agent,
"_compress_context_impl",
new_callable=AsyncMock,
) as mock_impl:
await agent.compress_context()
# _compress_context_impl should NOT have been called.
mock_impl.assert_not_awaited()
self.assertListEqual(self.execution_log, ["skipped"])
async def test_on_compress_context_no_middleware(self) -> None:
"""Test that compress_context calls _compress_context_impl directly
when no middleware is registered."""
self.mock_model.set_responses(
[
ChatResponse(
content=[TextBlock(text="test response")],
is_last=True,
),
],
)
context_config = ContextConfig(trigger_ratio=0.8, reserve_ratio=0.1)
agent = Agent(
name="test_agent",
system_prompt="test prompt",
model=self.mock_model,
toolkit=self.toolkit,
# No middlewares registered.
context_config=context_config,
)
with patch.object(
agent,
"_compress_context_impl",
new_callable=AsyncMock,
) as mock_impl:
await agent.compress_context(context_config=context_config)
mock_impl.assert_awaited_once_with(
context_config=context_config,
instructions=None,
)
async def asyncTearDown(self) -> None:
"""Clean up test fixtures."""
self.execution_log.clear()
|