File size: 20,063 Bytes
ea8c728 | 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 | """
Tests for the meta-prompt evolution system.
Tests cover:
1. SystemPromptDatabase CRUD operations
2. SystemPrompt dataclass
3. Prompt fitness tracking and selection
4. Integration with Program tracking
"""
import pytest
import tempfile
from pathlib import Path
from shinka.database.prompt_dbase import (
SystemPromptDatabase,
SystemPromptConfig,
SystemPrompt,
create_system_prompt,
)
from shinka.database import Program
from shinka.core.prompt_evolver import SystemPromptSampler
class TestSystemPromptDataclass:
"""Test SystemPrompt dataclass functionality."""
def test_create_system_prompt(self):
"""Test creating a new SystemPrompt."""
prompt = create_system_prompt(
prompt_text="You are an expert programmer.",
generation=0,
patch_type="init",
)
assert prompt.id is not None
assert len(prompt.id) == 36 # UUID length
assert prompt.prompt_text == "You are an expert programmer."
assert prompt.generation == 0
assert prompt.patch_type == "init"
assert prompt.fitness == 0.0
assert prompt.program_count == 0
def test_prompt_to_dict(self):
"""Test serialization to dictionary."""
prompt = create_system_prompt(
prompt_text="Test prompt",
generation=1,
patch_type="diff",
metadata={"test_key": "test_value"},
)
data = prompt.to_dict()
assert data["prompt_text"] == "Test prompt"
assert data["generation"] == 1
assert data["patch_type"] == "diff"
assert data["metadata"]["test_key"] == "test_value"
def test_prompt_from_dict(self):
"""Test deserialization from dictionary."""
data = {
"id": "test-id-123",
"prompt_text": "Test prompt",
"generation": 2,
"patch_type": "full",
"fitness": 0.5,
"program_count": 10,
"metadata": {"key": "value"},
}
prompt = SystemPrompt.from_dict(data)
assert prompt.id == "test-id-123"
assert prompt.prompt_text == "Test prompt"
assert prompt.generation == 2
assert prompt.fitness == 0.5
assert prompt.program_count == 10
def test_update_fitness(self):
"""Test fitness update calculation using percentiles."""
prompt = create_system_prompt(
prompt_text="Test",
generation=0,
patch_type="init",
)
# Initial state
assert prompt.fitness == 0.0
assert prompt.program_count == 0
# First percentile update (e.g., program beats 50% of others)
prompt.update_fitness(percentile=0.5, program_score=0.75)
assert prompt.program_count == 1
assert prompt.correct_program_count == 1
assert prompt.fitness == 0.5
assert prompt.total_percentile == 0.5
assert prompt.program_scores == [0.75]
# Second percentile update (beats 80% of others)
prompt.update_fitness(percentile=0.8, program_score=0.95)
assert prompt.program_count == 2
assert prompt.correct_program_count == 2
assert prompt.fitness == 0.65 # (0.5 + 0.8) / 2
assert prompt.total_percentile == 1.3
assert prompt.program_scores == [0.75, 0.95]
# Third percentile update (beats 30% of others)
prompt.update_fitness(percentile=0.3, program_score=0.4)
assert prompt.program_count == 3
assert prompt.correct_program_count == 3
assert abs(prompt.fitness - 0.5333) < 0.01 # (0.5 + 0.8 + 0.3) / 3
assert prompt.program_scores == [0.75, 0.95, 0.4]
def test_update_fitness_correct_only(self):
"""Test that only correct programs contribute to fitness."""
prompt = create_system_prompt(
prompt_text="Test",
generation=0,
patch_type="init",
)
# Add a correct program with 80% percentile
prompt.update_fitness(percentile=0.8, correct=True, program_score=0.9)
assert prompt.program_count == 1
assert prompt.correct_program_count == 1
assert prompt.fitness == 0.8
assert prompt.program_scores == [0.9]
# Add an incorrect program - should not affect fitness or scores
prompt.update_fitness(percentile=0.0, correct=False, program_score=0.1)
assert prompt.program_count == 2
assert prompt.correct_program_count == 1 # Still 1
assert prompt.fitness == 0.8 # Unchanged
assert prompt.program_scores == [0.9] # No new score added
class TestSystemPromptDatabase:
"""Test SystemPromptDatabase operations."""
@pytest.fixture
def temp_db(self):
"""Create a temporary database for testing."""
with tempfile.TemporaryDirectory() as tmpdir:
db_path = Path(tmpdir) / "test_prompts.db"
config = SystemPromptConfig(
db_path=str(db_path),
archive_size=5,
)
db = SystemPromptDatabase(config)
yield db
db.close()
@pytest.fixture
def memory_db(self):
"""Create an in-memory database for testing."""
config = SystemPromptConfig(
db_path=None, # In-memory
archive_size=5,
)
db = SystemPromptDatabase(config)
yield db
db.close()
def test_add_and_get_prompt(self, memory_db):
"""Test adding and retrieving a prompt."""
prompt = create_system_prompt(
prompt_text="Test prompt",
generation=0,
patch_type="init",
)
memory_db.add(prompt)
retrieved = memory_db.get(prompt.id)
assert retrieved is not None
assert retrieved.id == prompt.id
assert retrieved.prompt_text == "Test prompt"
def test_get_nonexistent_prompt(self, memory_db):
"""Test getting a prompt that doesn't exist."""
result = memory_db.get("nonexistent-id")
assert result is None
def test_get_all_prompts(self, memory_db):
"""Test getting all prompts."""
prompts = []
for i in range(3):
prompt = create_system_prompt(
prompt_text=f"Test prompt {i}",
generation=i,
patch_type="init",
)
memory_db.add(prompt)
prompts.append(prompt)
all_prompts = memory_db.get_all_prompts()
assert len(all_prompts) == 3
def test_archive_management(self, memory_db):
"""Test archive size management."""
# Add prompts up to archive size
for i in range(5):
prompt = create_system_prompt(
prompt_text=f"Prompt {i}",
generation=i,
patch_type="init",
)
memory_db.add(prompt)
archive = memory_db.get_archive()
assert len(archive) == 5
# Add one more - should replace worst
prompt = create_system_prompt(
prompt_text="New prompt",
generation=5,
patch_type="init",
)
# Give it a higher fitness
prompt.fitness = 10.0
prompt.program_count = 5
memory_db.add(prompt)
archive = memory_db.get_archive()
assert len(archive) <= 5
def test_update_fitness(self, memory_db):
"""Test updating prompt fitness with percentiles."""
prompt = create_system_prompt(
prompt_text="Test",
generation=0,
patch_type="init",
)
memory_db.add(prompt)
# Update fitness using percentiles with program scores
memory_db.update_fitness(
prompt.id, percentile=0.5, program_id="prog1", program_score=0.7
)
memory_db.update_fitness(
prompt.id, percentile=0.9, program_id="prog2", program_score=0.95
)
updated = memory_db.get(prompt.id)
assert updated.program_count == 2
assert updated.correct_program_count == 2
assert updated.fitness == 0.7 # (0.5 + 0.9) / 2
assert len(updated.program_ids) == 2
assert updated.program_scores == [0.7, 0.95]
def test_sample_prompt(self, memory_db):
"""Test sampling prompts from archive."""
# Add prompts with different fitness values (using percentiles)
for i in range(3):
prompt = create_system_prompt(
prompt_text=f"Prompt {i}",
generation=i,
patch_type="init",
)
memory_db.add(prompt)
# Update fitness with increasing percentiles and scores
percentile = (i + 1) * 0.25 # 0.25, 0.5, 0.75
score = (i + 1) * 0.3 # 0.3, 0.6, 0.9
memory_db.update_fitness(
prompt.id, percentile=percentile, program_score=score
)
memory_db.update_fitness(
prompt.id, percentile=percentile, program_score=score
)
memory_db.update_fitness(
prompt.id, percentile=percentile, program_score=score
)
# Sample multiple times
samples = [memory_db.sample() for _ in range(10)]
assert all(s is not None for s in samples)
def test_get_best_prompt(self, memory_db):
"""Test getting the best prompt."""
# Add prompts with different fitness
prompt1 = create_system_prompt("Prompt 1", generation=0, patch_type="init")
prompt2 = create_system_prompt("Prompt 2", generation=1, patch_type="init")
prompt3 = create_system_prompt("Prompt 3", generation=2, patch_type="init")
memory_db.add(prompt1)
memory_db.add(prompt2)
memory_db.add(prompt3)
# Update fitness with different percentiles and scores
for _ in range(3):
memory_db.update_fitness(prompt1.id, percentile=0.3, program_score=0.4)
memory_db.update_fitness(prompt2.id, percentile=0.9, program_score=0.95)
memory_db.update_fitness(prompt3.id, percentile=0.5, program_score=0.6)
best = memory_db.get_best_prompt()
assert best is not None
assert best.id == prompt2.id
def test_prompt_lineage(self, memory_db):
"""Test getting prompt lineage."""
# Create chain of prompts
prompt0 = create_system_prompt("Prompt 0", generation=0, patch_type="init")
memory_db.add(prompt0)
prompt1 = create_system_prompt(
"Prompt 1", parent_id=prompt0.id, generation=1, patch_type="diff"
)
memory_db.add(prompt1)
prompt2 = create_system_prompt(
"Prompt 2", parent_id=prompt1.id, generation=2, patch_type="full"
)
memory_db.add(prompt2)
# Get lineage
lineage = memory_db.get_lineage(prompt2.id)
assert len(lineage) == 2
assert lineage[0].id == prompt0.id
assert lineage[1].id == prompt1.id
def test_persistence(self, temp_db):
"""Test database persistence."""
prompt = create_system_prompt(
prompt_text="Persistent prompt",
generation=0,
patch_type="init",
)
temp_db.add(prompt)
temp_db.save()
# Get the db_path before closing
db_path = temp_db.config.db_path
temp_db.close()
# Reopen and verify
config = SystemPromptConfig(db_path=db_path, archive_size=5)
db2 = SystemPromptDatabase(config)
retrieved = db2.get(prompt.id)
assert retrieved is not None
assert retrieved.prompt_text == "Persistent prompt"
db2.close()
class TestSystemPromptSampler:
"""Test SystemPromptSampler functionality."""
@pytest.fixture
def sampler_with_prompts(self):
"""Create a sampler with some prompts."""
config = SystemPromptConfig(db_path=None, archive_size=10)
db = SystemPromptDatabase(config)
# Add prompts with varying fitness (using percentiles)
for i in range(5):
prompt = create_system_prompt(
prompt_text=f"Prompt {i}",
generation=i,
patch_type="init",
)
db.add(prompt)
# Give each prompt different percentile-based fitness with scores
percentile = (i + 1) * 0.15 # 0.15, 0.30, 0.45, 0.60, 0.75
score = (i + 1) * 0.18 # Corresponding scores
for _ in range(3):
db.update_fitness(prompt.id, percentile=percentile, program_score=score)
sampler = SystemPromptSampler(prompt_db=db, exploration_constant=1.0)
yield sampler, db
db.close()
def test_sample(self, sampler_with_prompts):
"""Test sampling prompts."""
sampler, db = sampler_with_prompts
# Sample should return a prompt
prompt = sampler.sample()
assert prompt is not None
assert isinstance(prompt, SystemPrompt)
def test_get_best_prompt(self, sampler_with_prompts):
"""Test getting best prompt through sampler."""
sampler, db = sampler_with_prompts
best = sampler.get_best_prompt()
assert best is not None
def test_get_archive(self, sampler_with_prompts):
"""Test getting archive through sampler."""
sampler, db = sampler_with_prompts
archive = sampler.get_archive()
assert len(archive) == 5
def test_ucb_exploration_effect(self):
"""Test that UCB exploration constant affects selection."""
config = SystemPromptConfig(db_path=None, archive_size=10)
db = SystemPromptDatabase(config)
# Add prompts with very different percentile-based fitness
low_fit = create_system_prompt("Low", generation=0, patch_type="init")
high_fit = create_system_prompt("High", generation=1, patch_type="init")
db.add(low_fit)
db.add(high_fit)
# Give very different percentile-based fitness with scores
for _ in range(5):
db.update_fitness(low_fit.id, percentile=0.2, program_score=0.3)
db.update_fitness(high_fit.id, percentile=0.9, program_score=0.95)
# Low exploration = more greedy (exploit high fitness)
low_explore_sampler = SystemPromptSampler(
db, exploration_constant=0.1, epsilon=0.0
)
high_explore_sampler = SystemPromptSampler(
db, exploration_constant=10.0, epsilon=0.0
)
# Count how often high-fitness is selected
# With low exploration, UCB should favor exploitation (high fitness prompt)
low_explore_high_count = sum(
1 for _ in range(100) if low_explore_sampler.sample().id == high_fit.id
)
# With high exploration, UCB should explore more evenly
# (both prompts have same number of samples, so exploration term is equal,
# but high exploration makes fitness difference matter less relatively)
high_explore_high_count = sum(
1 for _ in range(100) if high_explore_sampler.sample().id == high_fit.id
)
# Low exploration should select high fitness more often
# (exploitation dominates over exploration)
assert low_explore_high_count >= high_explore_high_count
db.close()
class TestPromptEvolutionPrompts:
"""Test the prompt evolution prompt templates."""
def test_diff_prompt_construction(self):
"""Test constructing diff evolution prompts."""
from shinka.prompts.prompts_prompt_evo import (
construct_diff_evolution_prompt,
)
prompt = create_system_prompt(
prompt_text="Original prompt",
generation=0,
patch_type="init",
)
prompt.fitness = 0.5
prompt.program_count = 10
# New signature takes top_programs list instead of recent_improvements
sys_msg, user_msg = construct_diff_evolution_prompt(prompt, top_programs=[])
assert "expert prompt engineer" in sys_msg.lower()
assert "Original prompt" in user_msg
assert (
"targeted" in sys_msg.lower()
) # Diff should mention targeted modifications
def test_full_prompt_construction(self):
"""Test constructing full rewrite prompts."""
from shinka.prompts.prompts_prompt_evo import (
construct_full_evolution_prompt,
)
prompt = create_system_prompt(
prompt_text="Original prompt",
generation=0,
patch_type="init",
)
# New signature takes top_programs list instead of best_program_info
sys_msg, user_msg = construct_full_evolution_prompt(prompt, top_programs=[])
assert "rewrite" in sys_msg.lower()
assert "Original prompt" in user_msg
def test_inspiration_prompt_formatting(self):
"""Test formatting inspiration prompts for prompt evolution context."""
from shinka.prompts.prompts_prompt_evo import format_inspiration_prompts
inspirations = [
create_system_prompt(
prompt_text=f"Inspiration {i}",
generation=i,
patch_type="diff",
)
for i in range(3)
]
for i, insp in enumerate(inspirations):
insp.fitness = 0.3 * (i + 1)
insp.program_count = 5
formatted = format_inspiration_prompts(inspirations, max_inspirations=2)
assert "Inspiration 1" in formatted
assert "Inspiration 2" in formatted
assert "Inspiration 3" not in formatted
assert "Performance: fitness=" in formatted
class TestEvolutionConfigPromptSettings:
"""Test EvolutionConfig prompt evolution settings."""
def test_default_prompt_evolution_config(self):
"""Test default prompt evolution settings."""
from shinka.core.config import EvolutionConfig
config = EvolutionConfig()
assert config.evolve_prompts is False
assert config.prompt_patch_types == ["diff", "full"]
assert config.prompt_patch_type_probs == [0.7, 0.3]
assert config.prompt_evolution_interval is None
assert config.prompt_archive_size == 10
def test_custom_prompt_evolution_config(self):
"""Test custom prompt evolution settings."""
from shinka.core.config import EvolutionConfig
config = EvolutionConfig(
evolve_prompts=True,
prompt_patch_types=["diff", "full", "cross"],
prompt_patch_type_probs=[0.5, 0.3, 0.2],
prompt_evolution_interval=20,
prompt_archive_size=15,
)
assert config.evolve_prompts is True
assert "cross" in config.prompt_patch_types
assert config.prompt_evolution_interval == 20
assert config.prompt_archive_size == 15
class TestProgramSystemPromptId:
"""Test Program dataclass with system_prompt_id."""
def test_program_with_prompt_id(self):
"""Test creating a program with system_prompt_id."""
program = Program(
id="test-program-id",
code="print('hello')",
generation=1,
system_prompt_id="test-prompt-id",
)
assert program.system_prompt_id == "test-prompt-id"
def test_program_without_prompt_id(self):
"""Test creating a program without system_prompt_id."""
program = Program(
id="test-program-id",
code="print('hello')",
generation=1,
)
assert program.system_prompt_id is None
def test_program_to_dict_with_prompt_id(self):
"""Test program serialization with system_prompt_id."""
program = Program(
id="test-id",
code="code",
generation=1,
system_prompt_id="prompt-id",
)
data = program.to_dict()
assert data["system_prompt_id"] == "prompt-id"
if __name__ == "__main__":
pytest.main([__file__, "-v"])
|