Spaces:
Sleeping
Sleeping
File size: 12,692 Bytes
116524e | 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 | """Tests for ace runners: ACE, TraceAnalyser, ACERunner, ACELiteLLM."""
from __future__ import annotations
from typing import Any, Optional
from unittest.mock import MagicMock, patch
import pytest
from ace.core.context import ACEStepContext, SkillbookView
from ace.core.environments import Sample, SimpleEnvironment
from ace.core.insight_source import TRACE_IDENTITY_METADATA_KEY
from ace.core.outputs import (
AgentOutput,
ReflectorOutput,
SkillManagerOutput,
)
from ace.core.skillbook import Skillbook, UpdateBatch, UpdateOperation
from ace.runners.base import ACERunner
# ------------------------------------------------------------------ #
# Mock roles — satisfy protocols without any LLM dependency
# ------------------------------------------------------------------ #
class MockAgent:
"""Minimal mock satisfying AgentLike."""
def generate(
self,
*,
question: str,
context: Optional[str],
skillbook: Any,
reflection: Optional[str] = None,
**kwargs: Any,
) -> AgentOutput:
return AgentOutput(reasoning="mock reasoning", final_answer="mock answer")
class MockReflector:
"""Minimal mock satisfying ReflectorLike."""
def reflect(
self,
*,
question: str,
agent_output: AgentOutput,
skillbook: Any,
ground_truth: Optional[str] = None,
feedback: Optional[str] = None,
**kwargs: Any,
) -> ReflectorOutput:
return ReflectorOutput(
reasoning="mock reflection",
correct_approach="mock approach",
key_insight="mock insight",
)
class MockSkillManager:
"""Minimal mock satisfying SkillManagerLike.
The real agentic SkillManager mutates the skillbook directly via
tools; this mock does the same so ``UpdateStep`` behaves realistically
without a live LLM.
"""
def update_skills(
self,
*,
reflections: tuple[ReflectorOutput, ...],
skillbook: Any,
question_context: str,
progress: str,
**kwargs: Any,
) -> SkillManagerOutput:
skill = skillbook.add_skill(section="learned", issue="mock skill")
return SkillManagerOutput(
update=UpdateBatch(
reasoning="mock update",
operations=[
UpdateOperation(
type="ADD",
section="learned",
issue="mock skill",
skill_id=skill.id,
)
],
),
)
# ------------------------------------------------------------------ #
# ACERunner base class
# ------------------------------------------------------------------ #
class TestACERunnerBase:
def test_save_and_load(self, tmp_path):
"""save() and load() should round-trip the skillbook."""
sb = Skillbook()
sb.add_skill("sec", "content", skill_id="s-001")
pipeline = MagicMock()
runner = ACERunner(pipeline=pipeline, skillbook=sb)
path = str(tmp_path / "sb.json")
runner.save(path)
# Modify skillbook
sb.add_skill("sec", "new", skill_id="s-002")
assert len(runner.skillbook.skills()) == 2
# Load should replace the skillbook
runner.load(path)
assert len(runner.skillbook.skills()) == 1
assert runner.skillbook.get_skill("s-001") is not None
def test_multi_epoch_requires_sequence(self):
"""Multi-epoch with non-Sequence should raise ValueError."""
pipeline = MagicMock()
sb = Skillbook()
runner = ACERunner(pipeline=pipeline, skillbook=sb)
def gen():
yield "item"
with pytest.raises(ValueError, match="Sequence"):
runner._run(gen(), epochs=2)
# ------------------------------------------------------------------ #
# load_skillbook alias correctness
# ------------------------------------------------------------------ #
class TestLoadSkillbookAlias:
def test_langchain_alias(self):
from ace.runners.langchain import LangChain
assert LangChain.load_skillbook is ACERunner.load
assert LangChain.save_skillbook is ACERunner.save
def test_browser_use_alias(self):
from ace.runners.browser_use import BrowserUse
assert BrowserUse.load_skillbook is ACERunner.load
assert BrowserUse.save_skillbook is ACERunner.save
def test_claude_code_alias(self):
from ace.runners.claude_code import ClaudeCode
assert ClaudeCode.load_skillbook is ACERunner.load
assert ClaudeCode.save_skillbook is ACERunner.save
def test_litellm_alias(self):
from ace.runners.litellm import ACELiteLLM
assert ACELiteLLM.load_skillbook is ACELiteLLM.load
assert ACELiteLLM.save_skillbook is ACELiteLLM.save
def test_load_not_save(self):
"""Critical: load_skillbook must NOT point to save."""
from ace.runners.langchain import LangChain
assert LangChain.load_skillbook is not ACERunner.save
assert LangChain.load_skillbook is not LangChain.save_skillbook
# ------------------------------------------------------------------ #
# ACE runner (full pipeline) with mocks
# ------------------------------------------------------------------ #
class TestACERunner:
def test_from_roles_run(self):
"""ACE.from_roles().run() should complete without error with mock roles."""
from ace.runners.ace import ACE
env = SimpleEnvironment()
runner = ACE.from_roles(
agent=MockAgent(),
reflector=MockReflector(),
skill_manager=MockSkillManager(),
environment=env,
)
samples = [
Sample(question="What is 2+2?", ground_truth="4"),
Sample(question="Capital of France?", ground_truth="Paris"),
]
results = runner.run(samples, epochs=1)
assert len(results) == 2
# After learning, skillbook should have skills
assert len(runner.skillbook.skills()) > 0
def test_multi_epoch(self):
from ace.runners.ace import ACE
env = SimpleEnvironment()
runner = ACE.from_roles(
agent=MockAgent(),
reflector=MockReflector(),
skill_manager=MockSkillManager(),
environment=env,
)
samples = [Sample(question="Q1", ground_truth="A1")]
results = runner.run(samples, epochs=2)
assert len(results) == 2 # 1 sample × 2 epochs
def test_build_context_adds_trace_identity_metadata(self):
from ace.runners.ace import ACE
runner = ACE.from_roles(
agent=MockAgent(),
reflector=MockReflector(),
skill_manager=MockSkillManager(),
)
sample = Sample(
question="Why did pagination stop early?",
id="conv-123",
metadata={
"source_system": "kayba-hosted",
"trace_id": "conv-123",
"display_name": "checkout-failure.md",
},
)
ctx = runner._build_context(
sample,
epoch=1,
total_epochs=1,
index=1,
total=1,
global_sample_index=1,
)
identity = ctx.metadata[TRACE_IDENTITY_METADATA_KEY]
assert identity["trace_uid"] == "kayba-hosted:conv-123"
assert identity["display_name"] == "checkout-failure.md"
# ------------------------------------------------------------------ #
# TraceAnalyser runner
# ------------------------------------------------------------------ #
class TestTraceAnalyser:
def test_from_roles_run(self):
"""TraceAnalyser.from_roles().run() with mock roles should work."""
from ace.runners.trace_analyser import TraceAnalyser
runner = TraceAnalyser.from_roles(
reflector=MockReflector(),
skill_manager=MockSkillManager(),
)
traces = [
{
"question": "What is 2+2?",
"answer": "4",
"reasoning": "simple",
"ground_truth": "4",
"feedback": "Correct!",
},
]
results = runner.run(traces)
assert len(results) == 1
assert len(runner.skillbook.skills()) > 0
def test_build_context_adds_inferred_trace_identity(self):
from ace.runners.trace_analyser import TraceAnalyser
runner = TraceAnalyser.from_roles(
reflector=MockReflector(),
skill_manager=MockSkillManager(),
)
ctx = runner._build_context(
{"sample_id": "trace-001", "question": "Q"},
epoch=1,
total_epochs=1,
index=1,
total=1,
global_sample_index=1,
)
identity = ctx.metadata[TRACE_IDENTITY_METADATA_KEY]
assert identity["trace_uid"] == "trace:trace-001"
assert identity["trace_id"] == "trace-001"
# ------------------------------------------------------------------ #
# ACELiteLLM
# ------------------------------------------------------------------ #
class TestACELiteLLM:
def _make_ace(self, **kwargs):
from ace.runners.litellm import ACELiteLLM
return ACELiteLLM(
"test-model",
agent=MockAgent(),
reflector=MockReflector(),
skill_manager=MockSkillManager(),
**kwargs,
)
def test_ask(self):
ace = self._make_ace()
answer = ace.ask("What is 2+2?")
assert answer == "mock answer"
def test_learn_from_feedback_no_prior_ask(self):
"""learn_from_feedback with no prior ask() should return False."""
ace = self._make_ace()
assert ace.learn_from_feedback("good answer") is False
def test_learn_from_feedback_after_ask(self):
"""learn_from_feedback after ask() should return True."""
ace = self._make_ace()
ace.ask("What is 2+2?")
result = ace.learn_from_feedback("Correct!", ground_truth="4")
assert result is True
assert len(ace.skillbook.skills()) > 0
def test_learn_from_feedback_disabled(self):
"""learn_from_feedback with learning disabled should return False."""
ace = self._make_ace(is_learning=False)
ace.ask("What is 2+2?")
assert ace.learn_from_feedback("Correct!") is False
def test_learn(self):
ace = self._make_ace(environment=SimpleEnvironment())
samples = [Sample(question="Q", ground_truth="A")]
results = ace.learn(samples)
assert len(results) == 1
def test_learn_disabled(self):
ace = self._make_ace(is_learning=False)
with pytest.raises(RuntimeError, match="disabled"):
ace.learn([Sample(question="Q", ground_truth="A")])
def test_save_and_load(self, tmp_path):
ace = self._make_ace()
ace.ask("Q")
ace.learn_from_feedback("good", ground_truth="A")
path = str(tmp_path / "sb.json")
ace.save(path)
skills_before = len(ace.skillbook.skills())
# Load into same instance
ace.load(path)
assert len(ace.skillbook.skills()) == skills_before
def test_enable_disable_learning(self):
ace = self._make_ace()
assert ace.is_learning is True
ace.disable_learning()
assert ace.is_learning is False
ace.enable_learning()
assert ace.is_learning is True
def test_get_strategies_empty(self):
ace = self._make_ace()
assert ace.get_strategies() == ""
def test_skillbook_path_loading(self, tmp_path):
"""Constructor with skillbook_path should load from file."""
sb = Skillbook()
sb.add_skill("test", "content", skill_id="t-001")
path = str(tmp_path / "sb.json")
sb.save_to_file(path)
from ace.runners.litellm import ACELiteLLM
ace = ACELiteLLM(
"test-model",
agent=MockAgent(),
reflector=MockReflector(),
skill_manager=MockSkillManager(),
skillbook_path=path,
)
assert ace.skillbook.get_skill("t-001") is not None
|