Added web agent
This commit is contained in:
@@ -2,6 +2,7 @@ import unittest
|
||||
from datetime import datetime
|
||||
from unittest.mock import Mock, MagicMock, patch
|
||||
import xml.etree.ElementTree as ET
|
||||
import logging
|
||||
|
||||
from sia.web_io_buffer import WebIOBuffer
|
||||
from sia.working_memory import WorkingMemory
|
||||
@@ -47,38 +48,48 @@ class BaseAgentTest(unittest.TestCase):
|
||||
self.mock_metrics = Mock(spec=SystemMetrics)
|
||||
self.mock_validator = Mock(spec=XMLValidator)
|
||||
self.io_buffer = WebIOBuffer()
|
||||
self.working_memory = WorkingMemory()
|
||||
|
||||
# Mock metrics context
|
||||
metrics_elem = ET.Element("context")
|
||||
metrics_elem.set("time", "2024-10-31T12:00:00Z")
|
||||
metrics_elem.set("cpu", "10")
|
||||
metrics_elem.set("gpu", "20")
|
||||
metrics_elem.set("memory_used", "1000")
|
||||
metrics_elem.set("memory_total", "2000")
|
||||
metrics_elem.set("disk_used", "5000")
|
||||
metrics_elem.set("disk_total", "10000")
|
||||
metrics_elem.set("context", "50")
|
||||
metrics_elem.set("stdin", "0")
|
||||
self.mock_metrics.generate_context.return_value = metrics_elem
|
||||
# Create parser with IO buffer
|
||||
self.parser = ResponseParser(self.io_buffer)
|
||||
|
||||
# Create test agent with mocked components
|
||||
with patch('sia.base_agent.LlmEngine') as mock_llm_class, \
|
||||
patch('sia.base_agent.SystemMetrics') as mock_metrics_class, \
|
||||
patch('sia.base_agent.XMLValidator') as mock_validator_class:
|
||||
# Mock metrics context generator
|
||||
def generate_context(context_size):
|
||||
metrics_elem = ET.Element("context")
|
||||
metrics_elem.set("time", "2024-10-31T12:00:00Z")
|
||||
metrics_elem.set("cpu", "10")
|
||||
metrics_elem.set("gpu", "20")
|
||||
metrics_elem.set("memory_used", "1000")
|
||||
metrics_elem.set("memory_total", "2000")
|
||||
metrics_elem.set("disk_used", "5000")
|
||||
metrics_elem.set("disk_total", "10000")
|
||||
metrics_elem.set("context", str(int(context_size * 100)))
|
||||
metrics_elem.set("stdin", "0")
|
||||
return metrics_elem
|
||||
|
||||
mock_llm_class.return_value = self.mock_llm
|
||||
mock_metrics_class.return_value = self.mock_metrics
|
||||
mock_validator_class.return_value = self.mock_validator
|
||||
|
||||
self.agent = TestBaseAgent(
|
||||
model_path="/test/model",
|
||||
system_prompt="test prompt",
|
||||
action_schema="test schema",
|
||||
io_buffer=self.io_buffer
|
||||
)
|
||||
self.mock_metrics.generate_context.side_effect = generate_context
|
||||
|
||||
# Create test agent with required components
|
||||
self.agent = TestBaseAgent(
|
||||
action_schema="test schema",
|
||||
working_memory=self.working_memory,
|
||||
system_metrics=self.mock_metrics,
|
||||
llm=self.mock_llm,
|
||||
validator=self.mock_validator,
|
||||
parser=self.parser
|
||||
)
|
||||
|
||||
# Set timestamp for entries
|
||||
self.test_timestamp = datetime(2024, 10, 31, 12, 0, 0)
|
||||
|
||||
# Capture logging output
|
||||
self.log_records = []
|
||||
logging.getLogger().handlers = []
|
||||
logging.getLogger().addHandler(self.log_handler)
|
||||
|
||||
def log_handler(self, record):
|
||||
"""Capture log records for testing."""
|
||||
self.log_records.append(record)
|
||||
|
||||
def tearDown(self):
|
||||
"""Clean up resources."""
|
||||
@@ -87,13 +98,11 @@ class BaseAgentTest(unittest.TestCase):
|
||||
|
||||
def test_initialization(self):
|
||||
"""Test agent initialization and component setup."""
|
||||
self.assertIsInstance(self.agent._working_memory, WorkingMemory)
|
||||
self.assertEqual(self.agent._working_memory, self.working_memory)
|
||||
self.assertEqual(self.agent._metrics, self.mock_metrics)
|
||||
self.assertEqual(self.agent._llm, self.mock_llm)
|
||||
self.assertEqual(self.agent._validator, self.mock_validator)
|
||||
self.assertEqual(self.agent._io_buffer, self.io_buffer)
|
||||
self.assertIsInstance(self.agent._parser, ResponseParser)
|
||||
self.assertEqual(self.agent._system_prompt, "test prompt")
|
||||
self.assertEqual(self.agent._parser, self.parser)
|
||||
self.assertEqual(self.agent._action_schema, "test schema")
|
||||
|
||||
def test_cleanup(self):
|
||||
@@ -107,16 +116,17 @@ class BaseAgentTest(unittest.TestCase):
|
||||
|
||||
# Parse context and verify structure
|
||||
root = ET.fromstring(context)
|
||||
self.assertEqual(root.tag, "state")
|
||||
self.assertEqual(root.tag, "context")
|
||||
|
||||
# Check metrics element
|
||||
context_elem = root.find("context")
|
||||
self.assertIsNotNone(context_elem)
|
||||
self.assertEqual(context_elem.get("cpu"), "10")
|
||||
self.assertEqual(context_elem.get("gpu"), "20")
|
||||
# Check metrics
|
||||
self.assertEqual(root.get("cpu"), "10")
|
||||
self.assertEqual(root.get("gpu"), "20")
|
||||
|
||||
# Check context size is 0 (empty memory)
|
||||
self.assertEqual(root.get("context"), "0")
|
||||
|
||||
# Check no memory entries
|
||||
self.assertEqual(len(root.findall("*")), 1) # Only metrics element
|
||||
self.assertEqual(len(list(root)), 0)
|
||||
|
||||
def test_compile_context_with_entries(self):
|
||||
"""Test context compilation with working memory entries."""
|
||||
@@ -129,8 +139,8 @@ class BaseAgentTest(unittest.TestCase):
|
||||
# Parse and verify context
|
||||
root = ET.fromstring(context)
|
||||
|
||||
# Check metrics and entry are present
|
||||
self.assertEqual(len(root.findall("*")), 2) # Metrics + reasoning
|
||||
# Check context size reflects one entry
|
||||
self.assertEqual(root.get("context"), "1")
|
||||
|
||||
# Verify entry content
|
||||
reasoning_elem = root.find("reasoning")
|
||||
@@ -152,11 +162,12 @@ class BaseAgentTest(unittest.TestCase):
|
||||
context = self.agent._compile_context()
|
||||
root = ET.fromstring(context)
|
||||
|
||||
# Verify all entries present
|
||||
self.assertEqual(len(root.findall("reasoning")), 3)
|
||||
# Check context size reflects three entries
|
||||
self.assertEqual(root.get("context"), "3")
|
||||
|
||||
# Verify entry order maintained
|
||||
reasoning_elems = root.findall("reasoning")
|
||||
self.assertEqual(len(reasoning_elems), 3)
|
||||
for i, elem in enumerate(reasoning_elems):
|
||||
self.assertEqual(elem.get("id"), f"id-{i}")
|
||||
self.assertEqual(elem.text, f"<![CDATA[test {i}]]>")
|
||||
233
test/web_agent_test.py
Normal file
233
test/web_agent_test.py
Normal file
@@ -0,0 +1,233 @@
|
||||
import unittest
|
||||
from datetime import datetime
|
||||
from unittest.mock import Mock, MagicMock, patch, call
|
||||
import xml.etree.ElementTree as ET
|
||||
from typing import Iterator
|
||||
|
||||
from sia.web_io_buffer import WebIOBuffer
|
||||
from sia.working_memory import WorkingMemory
|
||||
from sia.system_metrics import SystemMetrics
|
||||
from sia.llm_engine import LlmEngine
|
||||
from sia.xml_validator import XMLValidator
|
||||
from sia.response_parser import ResponseParser
|
||||
from sia.web_agent import WebAgent, WebAgentState
|
||||
from sia.command import Command
|
||||
from sia.command_result import CommandResult
|
||||
from sia.reasoning_entry import ReasoningEntry
|
||||
from sia.parse_error_entry import ParseErrorEntry
|
||||
from sia.entry import Entry
|
||||
|
||||
class MockCommand(Command):
|
||||
"""Mock command for testing execution flow."""
|
||||
def __init__(self, result: CommandResult):
|
||||
self.result = result
|
||||
self.executed = False
|
||||
|
||||
def execute(self, memory: WorkingMemory) -> CommandResult:
|
||||
self.executed = True
|
||||
return self.result
|
||||
|
||||
class WebAgentTest(unittest.TestCase):
|
||||
def setUp(self):
|
||||
"""Set up test cases with mocked components."""
|
||||
# Create mocks
|
||||
self.mock_llm = Mock(spec=LlmEngine)
|
||||
self.mock_metrics = Mock(spec=SystemMetrics)
|
||||
self.mock_validator = Mock(spec=XMLValidator)
|
||||
self.io_buffer = WebIOBuffer()
|
||||
self.working_memory = WorkingMemory()
|
||||
|
||||
# Mock validator to pass by default
|
||||
self.mock_validator.validate.return_value = None
|
||||
|
||||
# Create parser with IO buffer
|
||||
self.parser = ResponseParser(self.io_buffer)
|
||||
|
||||
# Mock metrics context generator
|
||||
def generate_context(context_size):
|
||||
metrics_elem = ET.Element("context")
|
||||
metrics_elem.set("time", "2024-10-31T12:00:00Z")
|
||||
metrics_elem.set("cpu", "10")
|
||||
metrics_elem.set("gpu", "20")
|
||||
metrics_elem.set("memory_used", "1000")
|
||||
metrics_elem.set("memory_total", "2000")
|
||||
metrics_elem.set("disk_used", "5000")
|
||||
metrics_elem.set("disk_total", "10000")
|
||||
metrics_elem.set("context", str(int(context_size * 100)))
|
||||
metrics_elem.set("stdin", "0")
|
||||
return metrics_elem
|
||||
|
||||
self.mock_metrics.generate_context.side_effect = generate_context
|
||||
|
||||
# Create agent with all components
|
||||
self.agent = WebAgent(
|
||||
system_prompt="test prompt",
|
||||
action_schema="test schema",
|
||||
working_memory=self.working_memory,
|
||||
system_metrics=self.mock_metrics,
|
||||
llm=self.mock_llm,
|
||||
validator=self.mock_validator,
|
||||
parser=self.parser
|
||||
)
|
||||
|
||||
# Set timestamp for entries
|
||||
self.test_timestamp = datetime(2024, 10, 31, 12, 0, 0)
|
||||
|
||||
# Handler tracking
|
||||
self.state_changes = []
|
||||
self.response_changes = []
|
||||
|
||||
def state_change_handler(self, state: WebAgentState):
|
||||
"""Track state changes for verification."""
|
||||
self.state_changes.append(state)
|
||||
|
||||
def response_change_handler(self, response: str):
|
||||
"""Track response changes for verification."""
|
||||
self.response_changes.append(response)
|
||||
|
||||
def test_initialization(self):
|
||||
"""Test initial state of web agent."""
|
||||
self.assertEqual(self.agent.current_state, WebAgentState.UPDATE)
|
||||
self.assertEqual(self.agent.context, "")
|
||||
self.assertEqual(self.agent.response, "")
|
||||
self.assertIsNone(self.agent._validation_error)
|
||||
self.assertEqual(len(self.agent._state_change_handlers), 0)
|
||||
self.assertEqual(len(self.agent._response_change_handlers), 0)
|
||||
|
||||
def test_handler_registration(self):
|
||||
"""Test adding state and response change handlers."""
|
||||
# Add handlers
|
||||
self.agent.add_state_change_handler(self.state_change_handler)
|
||||
self.agent.add_response_change_handler(self.response_change_handler)
|
||||
|
||||
# Verify handlers added
|
||||
self.assertEqual(len(self.agent._state_change_handlers), 1)
|
||||
self.assertEqual(len(self.agent._response_change_handlers), 1)
|
||||
|
||||
# Adding same handler twice should not duplicate
|
||||
self.agent.add_state_change_handler(self.state_change_handler)
|
||||
self.agent.add_response_change_handler(self.response_change_handler)
|
||||
|
||||
self.assertEqual(len(self.agent._state_change_handlers), 1)
|
||||
self.assertEqual(len(self.agent._response_change_handlers), 1)
|
||||
|
||||
def test_approve_context_state_error(self):
|
||||
"""Test approving context in wrong state raises error."""
|
||||
with self.assertRaises(Exception) as context:
|
||||
self.agent.approve_context()
|
||||
self.assertIn("Not in CONTEXT_APPROVAL state", str(context.exception))
|
||||
|
||||
def test_approve_response_state_error(self):
|
||||
"""Test approving response in wrong state raises error."""
|
||||
with self.assertRaises(Exception) as context:
|
||||
self.agent.approve_response()
|
||||
self.assertIn("Not in RESPONSE_APPROVAL state", str(context.exception))
|
||||
|
||||
def test_context_approval_flow(self):
|
||||
"""Test complete context approval flow with state transitions."""
|
||||
# Register handlers
|
||||
self.agent.add_state_change_handler(self.state_change_handler)
|
||||
self.agent.add_response_change_handler(self.response_change_handler)
|
||||
|
||||
# Setup LLM response
|
||||
def mock_infer(prompt: str, context: str) -> Iterator[str]:
|
||||
yield "<reasoning>test reasoning</reasoning>"
|
||||
self.mock_llm.infer.side_effect = mock_infer
|
||||
|
||||
# Set initial state
|
||||
self.agent._current_state = WebAgentState.CONTEXT_APPROVAL
|
||||
|
||||
# Approve context
|
||||
self.agent.approve_context()
|
||||
|
||||
# Verify state transitions
|
||||
expected_states = [
|
||||
WebAgentState.INFERENCE,
|
||||
WebAgentState.RESPONSE_APPROVAL
|
||||
]
|
||||
self.assertEqual(self.state_changes, expected_states)
|
||||
|
||||
# Verify response updates
|
||||
self.assertEqual(len(self.response_changes), 2) # Empty + final response
|
||||
self.assertEqual(self.response_changes[0], "")
|
||||
self.assertEqual(self.response_changes[1], "<reasoning>test reasoning</reasoning>")
|
||||
|
||||
def test_response_approval_flow_command(self):
|
||||
"""Test response approval flow with command execution."""
|
||||
# Register handlers
|
||||
self.agent.add_state_change_handler(self.state_change_handler)
|
||||
|
||||
# Setup command result
|
||||
command_result = CommandResult.success()
|
||||
mock_command = MockCommand(command_result)
|
||||
|
||||
# Mock parser to return command
|
||||
self.parser.parse = Mock(return_value=mock_command)
|
||||
|
||||
# Set initial state and response
|
||||
self.agent._current_state = WebAgentState.RESPONSE_APPROVAL
|
||||
self.agent._response = "<stop/>"
|
||||
|
||||
# Approve response
|
||||
self.agent.approve_response()
|
||||
|
||||
# Verify command executed
|
||||
self.assertTrue(mock_command.executed)
|
||||
|
||||
# Verify state transitions
|
||||
expected_states = [
|
||||
WebAgentState.UPDATE,
|
||||
WebAgentState.CONTEXT_APPROVAL
|
||||
]
|
||||
self.assertEqual(self.state_changes, expected_states)
|
||||
|
||||
# Verify command result stored
|
||||
self.assertEqual(self.agent.command_result, command_result)
|
||||
|
||||
def test_response_approval_flow_entry(self):
|
||||
"""Test response approval flow with entry creation."""
|
||||
# Register handlers
|
||||
self.agent.add_state_change_handler(self.state_change_handler)
|
||||
|
||||
# Set initial state and response
|
||||
self.agent._current_state = WebAgentState.RESPONSE_APPROVAL
|
||||
self.agent._response = "<reasoning>test reasoning</reasoning>"
|
||||
|
||||
# Approve response
|
||||
self.agent.approve_response()
|
||||
|
||||
# Verify entry added to working memory
|
||||
self.assertEqual(self.working_memory.get_entries_count(), 1)
|
||||
entry = self.working_memory.get_entries()[0]
|
||||
self.assertIsInstance(entry, ReasoningEntry)
|
||||
self.assertEqual(entry.content, "test reasoning")
|
||||
|
||||
# Verify state transitions
|
||||
expected_states = [
|
||||
WebAgentState.UPDATE,
|
||||
WebAgentState.CONTEXT_APPROVAL
|
||||
]
|
||||
self.assertEqual(self.state_changes, expected_states)
|
||||
|
||||
def test_response_validation(self):
|
||||
"""Test response validation during response setting."""
|
||||
# Register handlers
|
||||
self.agent.add_response_change_handler(self.response_change_handler)
|
||||
|
||||
# Mock validator to fail
|
||||
error_message = "Invalid XML"
|
||||
self.mock_validator.validate.return_value = error_message
|
||||
|
||||
# Set response
|
||||
self.agent._set_response("<invalid>")
|
||||
|
||||
# Verify validation error stored
|
||||
self.assertEqual(self.agent._validation_error, error_message)
|
||||
|
||||
# Verify response updated
|
||||
self.assertEqual(self.response_changes, ["<invalid>"])
|
||||
|
||||
def tearDown(self):
|
||||
"""Clean up resources."""
|
||||
if hasattr(self, 'agent'):
|
||||
del self.agent
|
||||
Reference in New Issue
Block a user