File size: 19,588 Bytes
9030acd e7fc062 9030acd | 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 | """Tests for OpenSleuth Level 2: auto-fuzzer + TaskCatalog + open /reset.
These tests do *not* require Hub network access. The Hub-availability test
is opportunistic: it asserts ``>=15`` total tasks if the dataset loads, but
silently passes (with a marker) if the Hub is offline / the env is sandboxed.
"""
from __future__ import annotations
import os
import random
import typing
from typing import Optional, Literal
import pytest
from fastapi.testclient import TestClient
from opensleuth_env import (
BLACK_BOX_FUNCTIONS,
OpenSleuthEnv,
ProbeAction,
SubmitAction,
TaskCatalog,
TaskResolutionError,
auto_fuzz,
)
# ---------------------------------------------------------------------------
# Auto-fuzzer
# ---------------------------------------------------------------------------
class TestAutoFuzzerTypes:
def _rng(self, seed: int = 0) -> random.Random:
return random.Random(seed)
def test_int_inputs_are_ints(self):
def f(n: int) -> int:
return n
rng = self._rng()
outs = auto_fuzz(f, 50, rng)
assert len(outs) == 50
assert all(isinstance(t, tuple) and len(t) == 1 for t in outs)
assert all(isinstance(t[0], int) and not isinstance(t[0], bool) for t in outs)
def test_str_inputs_are_strs(self):
def f(s: str) -> int:
return len(s)
outs = auto_fuzz(f, 30, self._rng())
assert all(isinstance(t[0], str) for t in outs)
def test_list_int_inputs_are_lists_of_ints(self):
def f(xs: list[int]) -> int:
return sum(xs)
outs = auto_fuzz(f, 30, self._rng())
for (xs,) in outs:
assert isinstance(xs, list)
assert all(isinstance(x, int) for x in xs)
def test_homogeneous_tuple_inputs(self):
def f(xs: tuple[int, ...]) -> int:
return sum(xs)
outs = auto_fuzz(f, 30, self._rng())
for (xs,) in outs:
assert isinstance(xs, tuple)
assert all(isinstance(x, int) for x in xs)
def test_heterogeneous_tuple_inputs(self):
def f(t: tuple[int, str]) -> int:
return len(t[1])
outs = auto_fuzz(f, 30, self._rng())
for (t,) in outs:
assert isinstance(t, tuple) and len(t) == 2
assert isinstance(t[0], int)
assert isinstance(t[1], str)
def test_optional_inputs_sometimes_None(self):
def f(x: Optional[int]) -> int:
return 0
outs = auto_fuzz(f, 200, self._rng(seed=42))
seen_none = any(t[0] is None for t in outs)
seen_int = any(isinstance(t[0], int) and not isinstance(t[0], bool) for t in outs)
assert seen_none, "Optional[int] should occasionally yield None"
assert seen_int, "Optional[int] should also yield ints"
def test_literal_inputs_only_pick_listed_values(self):
def f(mode: Literal["a", "b", "c"]) -> int:
return 0
outs = auto_fuzz(f, 50, self._rng())
for (m,) in outs:
assert m in ("a", "b", "c")
def test_dict_str_int_inputs(self):
def f(d: dict[str, int]) -> int:
return len(d)
outs = auto_fuzz(f, 20, self._rng())
for (d,) in outs:
assert isinstance(d, dict)
for k, v in d.items():
assert isinstance(k, str)
assert isinstance(v, int)
def test_multi_arg_returns_full_tuples(self):
def f(a: int, b: str) -> int:
return 0
outs = auto_fuzz(f, 20, self._rng())
for t in outs:
assert isinstance(t, tuple)
assert len(t) == 2
assert isinstance(t[0], int)
assert isinstance(t[1], str)
def test_unannotated_param_falls_back_to_int(self):
def f(x): # no annotation
return x
outs = auto_fuzz(f, 30, self._rng())
for (x,) in outs:
assert isinstance(x, int)
class TestAutoFuzzerSpecOverride:
def test_int_min_max_overrides_default_range(self):
def f(n: int) -> int:
return n
outs = auto_fuzz(f, 100, random.Random(0), fuzz_spec={"n": {"type": "int", "min": 1, "max": 5}})
for (n,) in outs:
assert 1 <= n <= 5, f"expected n in [1, 5], got {n}"
def test_str_alphabet_override(self):
def f(s: str) -> int:
return len(s)
outs = auto_fuzz(
f, 100, random.Random(0),
fuzz_spec={"s": {"type": "str", "alphabet": "ab", "max_len": 4}},
)
for (s,) in outs:
assert len(s) <= 4
for ch in s:
assert ch in "ab", f"unexpected char {ch!r} in {s!r}"
def test_list_elem_override(self):
def f(xs: list[int]) -> int:
return sum(xs)
outs = auto_fuzz(
f, 80, random.Random(0),
fuzz_spec={"xs": {"type": "list", "elem": {"type": "int", "min": 0, "max": 3}, "max_len": 4}},
)
for (xs,) in outs:
assert len(xs) <= 4
for v in xs:
assert 0 <= v <= 3
def test_tuple_elems_override(self):
def f(t):
return t
outs = auto_fuzz(
f, 30, random.Random(0),
fuzz_spec={"t": {"type": "tuple", "elems": [
{"type": "int", "min": 0, "max": 1},
{"type": "str", "alphabet": "x", "max_len": 2},
]}},
)
for (t,) in outs:
assert isinstance(t, tuple) and len(t) == 2
assert 0 <= t[0] <= 1
for ch in t[1]:
assert ch == "x"
# ---------------------------------------------------------------------------
# TaskCatalog
# ---------------------------------------------------------------------------
class TestTaskCatalog:
def test_resolves_builtin_by_name(self):
cat = TaskCatalog(enable_hub=False)
spec = cat.resolve(target_name="fibonacci")
assert spec.name == "fibonacci"
assert spec is BLACK_BOX_FUNCTIONS["fibonacci"]
assert spec.unpack_args is False
assert spec.source == "builtin"
def test_resolves_caller_supplied_target_code(self):
cat = TaskCatalog(enable_hub=False)
code = "def add(a: int, b: int) -> int:\n return a + b\n"
spec = cat.resolve(target_code=code, target_function_name="add")
assert spec.name == "add"
assert spec.unpack_args is True # 2-arg
assert spec.source == "user"
# The wrapped fuzzer must produce calls that succeed end-to-end.
rng = random.Random(0)
inputs = spec.fuzzer(rng, 10)
for args in inputs:
assert isinstance(args, tuple) and len(args) == 2
assert spec.fn(*args) == args[0] + args[1]
def test_caller_supplied_unary_uses_unwrapped_call(self):
cat = TaskCatalog(enable_hub=False)
code = "def square(n: int) -> int:\n return n * n\n"
spec = cat.resolve(target_code=code, target_function_name="square")
assert spec.unpack_args is False
rng = random.Random(0)
inputs = spec.fuzzer(rng, 5)
for x in inputs:
assert isinstance(x, int)
assert spec.fn(x) == x * x
def test_resolve_with_no_source_raises(self):
cat = TaskCatalog(enable_hub=False)
with pytest.raises(TaskResolutionError):
cat.resolve()
def test_resolve_unknown_name_raises(self):
cat = TaskCatalog(enable_hub=False)
with pytest.raises(TaskResolutionError):
cat.resolve(target_name="this_does_not_exist")
def test_target_code_without_function_name_raises(self):
cat = TaskCatalog(enable_hub=False)
with pytest.raises(TaskResolutionError):
cat.resolve(target_code="def foo(): return 1\n")
def test_rejects_oracle_import(self):
cat = TaskCatalog(enable_hub=False)
bad = (
"import opensleuth_env\n"
"def f(x): return x\n"
)
with pytest.raises(TaskResolutionError):
cat.resolve(target_code=bad, target_function_name="f")
bad2 = (
"from opensleuth_env.black_box import _fibonacci\n"
"def f(x): return _fibonacci(x)\n"
)
with pytest.raises(TaskResolutionError):
cat.resolve(target_code=bad2, target_function_name="f")
def test_target_code_using_open_is_blocked_at_call_time(self):
"""`open` is not in the safe-builtins whitelist. The catalog will
compile the function (since `open` is only resolved at call-time
via NameError), but invoking it must fail safely."""
cat = TaskCatalog(enable_hub=False)
code = (
"def f(x):\n"
" open('/tmp/x', 'w')\n"
" return 0\n"
)
spec = cat.resolve(target_code=code, target_function_name="f")
with pytest.raises(NameError):
spec.fn(0)
def test_caller_supplied_edge_cases_are_parsed(self):
cat = TaskCatalog(enable_hub=False)
spec = cat.resolve(
target_code="def neg(n: int) -> int:\n return -n\n",
target_function_name="neg",
edge_cases=["0", "1", "-1", "100"],
)
assert spec.edge_cases == [0, 1, -1, 100]
def test_caller_supplied_fuzz_spec_is_used(self):
cat = TaskCatalog(enable_hub=False)
spec = cat.resolve(
target_code="def f(n: int) -> int:\n return n\n",
target_function_name="f",
fuzz_spec={"n": {"type": "int", "min": 7, "max": 9}},
)
rng = random.Random(0)
inputs = spec.fuzzer(rng, 50)
for x in inputs:
assert 7 <= x <= 9
def test_list_builtin_returns_nine_entries(self):
cat = TaskCatalog(enable_hub=False)
builtins_list = cat.list_builtin()
assert len(builtins_list) == 9
for entry in builtins_list:
assert entry["source"] == "builtin"
assert "name" in entry
assert "signature" in entry
assert "difficulty" in entry
# ---------------------------------------------------------------------------
# End-to-end via OpenSleuthEnv
# ---------------------------------------------------------------------------
class TestEnvOpenEnded:
def test_legacy_reset_by_target_name_unchanged(self):
env = OpenSleuthEnv(fuzz_count=10)
obs = env.reset(target_name="fibonacci")
assert obs.target_function_name == "fibonacci"
assert obs.difficulty == "easy"
assert obs.steps_taken == 0
# Probe via the same path as before.
resp = env.step(obs.episode_id, ProbeAction(input_repr="10"))
assert resp.observation.probe_history[-1].output_repr == "55"
def test_env_caller_supplied_unary_full_loop(self):
env = OpenSleuthEnv(fuzz_count=10)
obs = env.reset(
target_code="def square(n: int) -> int:\n return n * n\n",
target_function_name="square",
)
assert obs.target_function_name == "square"
# Probe.
resp = env.step(obs.episode_id, ProbeAction(input_repr="5"))
assert resp.observation.probe_history[-1].output_repr == "25"
# Submit a perfect implementation.
code = "def square(n):\n return n * n\n"
resp = env.step(obs.episode_id, SubmitAction(code=code))
assert resp.done is True
assert resp.info["execution_reward"] == pytest.approx(100.0)
assert resp.reward > 140.0
def test_env_caller_supplied_multi_arg_full_loop(self):
env = OpenSleuthEnv(fuzz_count=10)
obs = env.reset(
target_code="def add(a: int, b: int) -> int:\n return a + b\n",
target_function_name="add",
edge_cases=["(0, 0)", "(1, -1)", "(100, 0)"],
)
assert obs.target_function_name == "add"
# Probe with a 2-tuple.
resp = env.step(obs.episode_id, ProbeAction(input_repr="(2, 3)"))
assert resp.observation.probe_history[-1].output_repr == "5"
# Submit a perfect implementation.
code = "def add(a, b):\n return a + b\n"
resp = env.step(obs.episode_id, SubmitAction(code=code))
assert resp.done is True
assert resp.info["execution_reward"] == pytest.approx(100.0)
assert resp.reward > 140.0
def test_env_caller_supplied_buggy_submission_scored_negative(self):
env = OpenSleuthEnv(fuzz_count=10)
obs = env.reset(
target_code="def add(a: int, b: int) -> int:\n return a + b\n",
target_function_name="add",
)
bad = "def add(a, b):\n return a - b\n"
resp = env.step(obs.episode_id, SubmitAction(code=bad))
assert resp.done is True
assert resp.info["execution_reward"] < 50.0
assert resp.reward < 0.0
def test_env_caller_supplied_oracle_import_rejected(self):
env = OpenSleuthEnv()
with pytest.raises(ValueError):
env.reset(
target_code="import opensleuth_env\ndef f(x): return x\n",
target_function_name="f",
)
# ---------------------------------------------------------------------------
# HTTP layer
# ---------------------------------------------------------------------------
@pytest.fixture(scope="module")
def http_client():
from server import app
with TestClient(app) as client:
yield client
class TestHttpLayer:
def test_tasks_endpoint_lists_at_least_nine_builtin(self, http_client):
r = http_client.get("/tasks?source=builtin")
assert r.status_code == 200
body = r.json()
assert body["count"] >= 9
names = [t["name"] for t in body["tasks"]]
for name in BLACK_BOX_FUNCTIONS:
assert name in names
def test_tasks_all_includes_at_least_builtins(self, http_client):
r = http_client.get("/tasks?source=all")
assert r.status_code == 200
body = r.json()
# The builtins are always present. If the Hub is reachable we'd
# expect 15+, but the test must pass even if Hub is unavailable
# (e.g. CI sandboxes block egress).
assert body["count"] >= 9
if not body["hub"].get("enabled", False) or body["hub"].get("error"):
pytest.skip(f"hub not reachable: {body['hub']}")
# Hub reachable -> dataset should have 15+ rows after bootstrap.
assert body["count"] >= 15
def test_sample_inputs_returns_n_repr_strings_for_builtin(self, http_client):
r = http_client.get("/tasks/fibonacci/sample_inputs?n=5&seed=7")
assert r.status_code == 200, r.text
body = r.json()
assert body["name"] == "fibonacci"
assert body["n"] == 5
assert body["seed"] == 7
assert isinstance(body["inputs"], list)
assert len(body["inputs"]) == 5
# Every returned string must be ast.literal_eval-safe so the trainer
# can post it straight back to /step as a probe input_repr.
import ast
for s in body["inputs"]:
assert isinstance(s, str)
ast.literal_eval(s)
# Determinism: same seed -> identical inputs.
r2 = http_client.get("/tasks/fibonacci/sample_inputs?n=5&seed=7")
assert r2.json()["inputs"] == body["inputs"]
def test_sample_inputs_unknown_target_404s(self, http_client):
r = http_client.get("/tasks/__nope__/sample_inputs?n=2&seed=0")
assert r.status_code == 404
def test_obviously_wrong_submission_scores_low_under_thread_pool(self, http_client):
"""Regression: TestClient uses a worker thread, exercising the
same `signal.signal` -> ValueError path that uvicorn workers hit
in production. Before the verifier fix, this returned 100/100 for
any defined function (incl. ``def fibonacci(n): return n``).
After the fix, an obviously-wrong submission should score near
zero and trigger the floor penalty.
"""
ep = http_client.post("/reset", json={
"target_name": "fibonacci", "seed": 42, "max_steps": 2,
}).json()
eid = ep["episode_id"]
r = http_client.post("/step", json={
"episode_id": eid,
"action": {"action_type": "submit", "code": "def fibonacci(n):\n return n\n"},
})
assert r.status_code == 200, r.text
body = r.json()
info = body["info"]
# ``return n`` matches at most a couple of fixed points (n=1, n=2)
# out of 100+ random inputs; execution_reward should be tiny.
assert info["execution_reward"] < 20.0, info
assert info["matches"] < info["fuzz_count"] // 4, info
# Floor penalty should kick in.
assert info["floor_penalty"] == 25.0, info
# And the perfect-bonus must NOT fire.
assert info["perfect_bonus"] == 0.0, info
def test_reset_legacy_target_name_still_works(self, http_client):
r = http_client.post("/reset", json={
"target_name": "fibonacci", "seed": 0, "max_steps": 10,
})
assert r.status_code == 200
body = r.json()
assert body["target_function_name"] == "fibonacci"
assert "fibonacci" in body["target_function_signature"]
def test_reset_caller_supplied_target_code(self, http_client):
payload = {
"target_code": "def add(a: int, b: int) -> int:\n return a + b\n",
"target_function_name": "add",
"edge_cases": ["(0, 0)", "(1, -1)"],
"max_steps": 5,
}
r = http_client.post("/reset", json=payload)
assert r.status_code == 200, r.text
body = r.json()
assert body["target_function_name"] == "add"
eid = body["episode_id"]
# Probe -> verify wrapping.
r = http_client.post("/step", json={
"episode_id": eid,
"action": {"action_type": "probe", "input_repr": "(7, 8)"},
})
assert r.status_code == 200, r.text
body = r.json()
assert body["observation"]["probe_history"][-1]["output_repr"] == "15"
# Submit perfect.
r = http_client.post("/step", json={
"episode_id": eid,
"action": {"action_type": "submit", "code": "def add(a, b):\n return a + b\n"},
})
assert r.status_code == 200, r.text
body = r.json()
assert body["done"] is True
assert body["info"]["execution_reward"] == pytest.approx(100.0)
assert body["reward"] > 140.0
def test_reset_with_neither_target_returns_400(self, http_client):
r = http_client.post("/reset", json={"seed": 0})
assert r.status_code == 400
def test_reset_with_target_code_only_no_function_name_returns_400(self, http_client):
r = http_client.post("/reset", json={
"target_code": "def f(): return 1\n",
})
assert r.status_code == 400
def test_functions_endpoint_unchanged_for_trainer(self, http_client):
r = http_client.get("/functions")
assert r.status_code == 200
body = r.json()
assert "functions" in body
names = [f["name"] for f in body["functions"]]
for name in BLACK_BOX_FUNCTIONS:
assert name in names
# The original v0.3 fields must all be present.
for entry in body["functions"]:
for k in ("name", "signature", "description", "difficulty", "edge_case_count"):
assert k in entry
|