File size: 34,715 Bytes
5e5d6c5 | 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 | from __future__ import annotations
import hashlib
import json
from collections.abc import Sequence
from dataclasses import dataclass, field
from typing import cast
import pytest
import ctx.runtime.planning_v3 as planning_v3_module
from ctx.engine.benefit import (
BenefitCandidate,
BenefitSelectionResult,
EvidenceSummary,
NetBenefitPolicy,
ResourceCosts,
)
from ctx.engine.content import MaterialDescriptor, MaterialIdentity
from ctx.engine.installation import InstallPlanDescriptor, InstallPlanningBundle
from ctx.engine.lineage import CatalogCapabilityIdentity
from ctx.engine.planner import (
CandidateAuthorityUnavailable,
CandidateSourceUnavailable,
CapabilityCandidate,
PlannerValidationError,
WorkObservation,
)
from ctx.engine.planning_v3 import (
AuthenticatedNetBenefitPlanner,
BenefitAuditStore,
)
from ctx.engine.protocol import ScopeRef
from ctx.engine.replay import PlanningContext, StructuredSurrogate
from ctx.engine.state import CapabilityState, EngineState, LeaseRef
from ctx.runtime.planning_v3 import (
AuthenticatedReplayDecisionPlannerV3,
CatalogLoadPlanningBundle,
)
def _digest(value: str) -> str:
return hashlib.sha256(value.encode()).hexdigest()
def _material(capability_id: str, salt: str | None = None) -> MaterialIdentity:
kind = capability_id.split(":", 1)[0]
return MaterialIdentity.create(
capability_id=capability_id,
kind=kind,
content_sha256=_digest(f"content:{salt or capability_id}"),
content_bytes=32,
)
def _load_descriptor(
capability_id: str,
*,
schema_version: int = 2,
material_salt: str | None = None,
provenance_salt: str = "material-snapshot",
) -> MaterialDescriptor:
kind = capability_id.split(":", 1)[0]
material = _material(capability_id, material_salt)
return MaterialDescriptor.create(
capability_id=capability_id,
kind=kind,
actionability="load",
content_sha256=material.content_sha256,
content_bytes=material.content_bytes,
estimated_tokens=8,
provenance_digest=_digest(provenance_salt),
material_identity_digest=(material.identity_digest if schema_version == 2 else None),
)
def _install_bundle(
capability_id: str,
*,
provenance_salt: str = "install-snapshot",
result_salt: str = "installed",
) -> InstallPlanningBundle:
kind = capability_id.split(":", 1)[0]
result = _material(capability_id, result_salt)
descriptor = InstallPlanDescriptor.create(
capability_id=capability_id,
kind=kind,
installer_id="ctx-installer-v1",
plan_digest=_digest(f"plan:{capability_id}"),
provenance_digest=_digest(provenance_salt),
result_material_identity_digest=result.identity_digest,
)
return InstallPlanningBundle(descriptor=descriptor, result_material=result)
def _candidate(
capability_id: str,
actionability: str,
*,
source_salt: str | None = None,
) -> CapabilityCandidate:
kind, name = capability_id.split(":", 1)
bundle = _install_bundle(capability_id) if actionability == "install" else None
return CapabilityCandidate(
capability_id=capability_id,
kind=kind,
name=name,
source_digest=_digest(f"source:{source_salt or capability_id}"),
normalized_score_ppm=900_000,
matching_signals=("python",),
reason_codes=("signal-match",),
actionability=actionability,
install_descriptor_digest=(
None
if actionability != "install"
else (
bundle.descriptor.descriptor_digest
if bundle is not None
else _digest(f"descriptor:{capability_id}")
)
),
install_plan_digest=(
None
if actionability != "install"
else (
bundle.descriptor.plan_digest
if bundle is not None
else _digest(f"plan:{capability_id}")
)
),
)
def _benefit(
candidate: CapabilityCandidate,
*,
availability: str | None = None,
source_digest: str | None = None,
expected_task_benefit_ppm: int = 700_000,
) -> BenefitCandidate:
tier = availability or ("advisory" if candidate.actionability == "manual" else "executable")
return BenefitCandidate(
capability_id=candidate.capability_id,
source_digest=source_digest or candidate.source_digest,
resource_profile_digest=_digest(f"profile:{candidate.capability_id}"),
availability=tier,
expected_task_benefit_ppm=expected_task_benefit_ppm,
relevance_ppm=1_000_000,
trust_ppm=1_000_000,
costs=ResourceCosts(),
evidence=EvidenceSummary(
capability_id=candidate.capability_id,
kind=candidate.kind,
source_digest=source_digest or candidate.source_digest,
evidence_window_digest=_digest(f"window:{candidate.capability_id}"),
opportunity_observable=False,
),
source_trusted=True,
security_approved=True,
permissions_allowed=True,
credentials_available=True,
)
@dataclass
class _AuditStore(BenefitAuditStore):
results: dict[str, BenefitSelectionResult] = field(default_factory=dict)
def store(self, result: BenefitSelectionResult) -> str:
self.results[result.result_digest] = result
return result.result_digest
@dataclass
class _Source:
values: Sequence[CapabilityCandidate]
catalog_snapshot_digest: str = field(default_factory=lambda: _digest("catalog-snapshot"))
observations: list[WorkObservation] = field(default_factory=list)
def retrieve(self, observation: WorkObservation) -> Sequence[CapabilityCandidate]:
self.observations.append(observation)
return self.values
@dataclass
class _Facts:
values: dict[str, BenefitCandidate | None]
benefit_facts_snapshot_digest: str = field(default_factory=lambda: _digest("benefit-facts"))
calls: list[str] = field(default_factory=list)
def benefit_candidate(
self,
presentation: CapabilityCandidate,
_observation: WorkObservation,
) -> BenefitCandidate | None:
self.calls.append(presentation.capability_id)
return self.values.get(presentation.capability_id)
@dataclass
class _Materials:
values: dict[str, CatalogLoadPlanningBundle | MaterialDescriptor | None]
material_snapshot_digest: str = field(default_factory=lambda: _digest("material-snapshot"))
calls: list[str] = field(default_factory=list)
def load_bundle(
self,
presentation: CapabilityCandidate,
) -> CatalogLoadPlanningBundle | None:
self.calls.append(presentation.capability_id)
return cast(
CatalogLoadPlanningBundle | None,
self.values.get(presentation.capability_id),
)
@dataclass
class _Installs:
values: dict[str, InstallPlanningBundle | None]
installation_snapshot_digest: str = field(default_factory=lambda: _digest("install-snapshot"))
calls: list[str] = field(default_factory=list)
def describe_bundle(self, capability_id: str, _kind: str) -> InstallPlanningBundle | None:
self.calls.append(capability_id)
return self.values.get(capability_id)
def _observation(
*,
baseline: tuple[str, ...] = (),
active: tuple[str, ...] = (),
rejected: tuple[str, ...] = (),
requested_limit: int = 5,
) -> StructuredSurrogate:
return StructuredSurrogate.create(
schema_id="ctx.observation.current-work",
schema_version=1,
value={
"signals": ["python", "testing"],
"languages": ["python"],
"baseline_capability_ids": list(baseline),
"active_capability_ids": list(active),
"rejected_capability_ids": list(rejected),
"requested_limit": requested_limit,
},
)
def _catalog_load_bundle(
presentation: CapabilityCandidate,
*,
descriptor: MaterialDescriptor | None = None,
catalog_snapshot_digest: str | None = None,
catalog_namespace_digest: str | None = None,
) -> CatalogLoadPlanningBundle:
value = descriptor or _load_descriptor(presentation.capability_id)
return CatalogLoadPlanningBundle(
presentation=presentation,
catalog_identity=CatalogCapabilityIdentity.create(
capability_id=presentation.capability_id,
kind=presentation.kind,
catalog_namespace_digest=(catalog_namespace_digest or _digest("catalog-namespace")),
),
descriptor=value,
catalog_snapshot_digest=(catalog_snapshot_digest or _digest("catalog-snapshot")),
material_snapshot_digest=value.provenance_digest,
)
def _adapter(
candidates: Sequence[CapabilityCandidate],
*,
facts: dict[str, BenefitCandidate | None] | None = None,
materials: dict[str, CatalogLoadPlanningBundle | MaterialDescriptor | None] | None = None,
installs: dict[str, InstallPlanningBundle | None] | None = None,
) -> tuple[
AuthenticatedReplayDecisionPlannerV3,
_Source,
_Facts,
_Materials,
_Installs,
]:
source = _Source(candidates)
facts_port = _Facts(
facts
if facts is not None
else {candidate.capability_id: _benefit(candidate) for candidate in candidates}
)
material_port = _Materials(
materials
if materials is not None
else {
candidate.capability_id: _catalog_load_bundle(candidate)
for candidate in candidates
if candidate.actionability == "load"
}
)
install_port = _Installs(
installs
if installs is not None
else {
candidate.capability_id: _install_bundle(candidate.capability_id)
for candidate in candidates
if candidate.actionability == "install"
}
)
adapter = AuthenticatedReplayDecisionPlannerV3(
planner=AuthenticatedNetBenefitPlanner(
policy=NetBenefitPolicy(
calibration_digest=_digest("calibration"),
minimum_relevance_ppm=1,
),
audit_store=_AuditStore(),
),
source=source,
benefit_facts_port=facts_port,
material_port=material_port,
install_bundle_port=install_port,
planner_version="planner-v3",
catalog_namespace_digest=_digest("catalog-namespace"),
)
return adapter, source, facts_port, material_port, install_port
def _context(adapter: AuthenticatedReplayDecisionPlannerV3) -> PlanningContext:
return PlanningContext(
planner_version="planner-v3",
catalog_snapshot_digest=adapter.catalog_snapshot_digest,
)
def _active_state(candidate: CapabilityCandidate) -> EngineState:
lease = LeaseRef(
lease_id="lease-active",
owner_id="owner-active",
exposure_id="exposure-1",
)
return EngineState(
revision=1,
scope=ScopeRef(
tenant_id="tenant-1",
workspace_id="workspace-1",
repository_id="repository-1",
session_id="session-1",
exposure_id="exposure-1",
host_context_id="host-1",
),
host_level="activating",
host_descriptor_digest=_digest("host"),
capabilities=(
CapabilityState(
capability_id=candidate.capability_id,
source_digest=candidate.source_digest,
plan_id="plan-active",
catalog_snapshot_id=_digest("catalog-snapshot"),
kind=candidate.kind,
actionability=candidate.actionability,
leases=(lease,),
activation="active",
activation_lease_id=lease.lease_id,
),
),
_contract_version=2,
)
def test_emits_only_schema_v3_for_mixed_load_install_and_manual_authorities() -> None:
candidates = (
_candidate("agent:advisor", "manual"),
_candidate("skill:remote", "install"),
_candidate("skill:local", "load"),
)
adapter, _source, _facts, _materials, _installs = _adapter(candidates)
decision = adapter(_observation(), None, _context(adapter))
assert (decision.schema_id, decision.schema_version) == (
"ctx.decision.capability-plan",
3,
)
rows = decision.to_dict()["value"]["capabilities"]
assert [row["capability_id"] for row in rows] == [
"skill:local",
"skill:remote",
"agent:advisor",
]
assert [row["authority"]["type"] for row in rows] == ["load", "install", "manual"]
assert rows[0]["authority"]["material"]["origin"] == "catalog"
assert rows[1]["authority"]["descriptor"]["schema"] == "ctx.install-plan-descriptor-v2"
assert rows[2]["authority"] == {"type": "manual"}
@pytest.mark.parametrize(
("candidates", "requested_limit", "code"),
[
((), 5, "no-feasible-capability"),
((_candidate("skill:local", "load"),), 0, "limit-zero"),
],
)
def test_preserves_zero_selection_abstention(
candidates: Sequence[CapabilityCandidate],
requested_limit: int,
code: str,
) -> None:
adapter, _source, _facts, _materials, _installs = _adapter(candidates)
value = adapter(
_observation(requested_limit=requested_limit),
None,
_context(adapter),
).to_dict()["value"]
assert value["status"] == "abstained"
assert value["abstention_code"] == code
assert value["capabilities"] == []
assert value["benefit_audit"] is not None
def test_declared_candidate_source_unavailability_emits_schema_v3_degradation() -> None:
class _UnavailableSource(_Source):
def retrieve(self, observation: WorkObservation) -> Sequence[CapabilityCandidate]:
self.observations.append(observation)
raise CandidateSourceUnavailable("private graph failure")
adapter, source, facts, materials, installs = _adapter(())
unavailable = _UnavailableSource((), catalog_snapshot_digest=source.catalog_snapshot_digest)
adapter = AuthenticatedReplayDecisionPlannerV3(
planner=adapter.planner,
source=unavailable,
benefit_facts_port=facts,
material_port=materials,
install_bundle_port=installs,
planner_version="planner-v3",
catalog_namespace_digest=_digest("catalog-namespace"),
)
value = adapter(_observation(), None, _context(adapter)).to_dict()["value"]
assert value == {
"status": "degraded",
"abstention_code": "catalog-unavailable",
"benefit_audit": None,
"capabilities": [],
}
assert "private graph failure" not in str(value)
@pytest.mark.parametrize("mode", ["missing", "mismatched-id", "mismatched-source"])
def test_rejects_missing_or_mismatched_authenticated_benefit_facts(mode: str) -> None:
candidate = _candidate("skill:local", "load")
facts: dict[str, BenefitCandidate | None]
if mode == "missing":
facts = {candidate.capability_id: None}
elif mode == "mismatched-id":
other = _candidate("skill:other", "load")
facts = {candidate.capability_id: _benefit(other)}
else:
facts = {
candidate.capability_id: _benefit(
candidate,
source_digest=_digest("substituted-source"),
)
}
adapter, _source, _facts, _materials, _installs = _adapter((candidate,), facts=facts)
with pytest.raises(PlannerValidationError, match="benefit"):
adapter(_observation(), None, _context(adapter))
def test_isolates_v1_load_descriptor_without_downgrading_to_manual() -> None:
candidate = _candidate("skill:local", "load")
adapter, _source, _facts, _materials, _installs = _adapter(
(candidate,),
materials={
candidate.capability_id: _load_descriptor(candidate.capability_id, schema_version=1)
},
)
value = adapter(_observation(), None, _context(adapter)).to_dict()["value"]
assert value["status"] == "abstained"
assert value["capabilities"] == []
def test_rejects_load_descriptor_from_a_different_material_snapshot() -> None:
candidate = _candidate("skill:local", "load")
descriptor = _load_descriptor(candidate.capability_id)
materials = _Materials(
{candidate.capability_id: _catalog_load_bundle(candidate, descriptor=descriptor)},
material_snapshot_digest=_digest("different-material-snapshot"),
)
adapter, source, facts, _default_materials, installs = _adapter((candidate,))
adapter = AuthenticatedReplayDecisionPlannerV3(
planner=adapter.planner,
source=source,
benefit_facts_port=facts,
material_port=materials,
install_bundle_port=installs,
planner_version="planner-v3",
catalog_namespace_digest=_digest("catalog-namespace"),
)
value = adapter(_observation(), None, _context(adapter)).to_dict()["value"]
assert value["status"] == "abstained"
assert value["capabilities"] == []
@pytest.mark.parametrize("mode", ["missing", "substituted"])
def test_isolates_missing_or_substituted_install_bundle_instead_of_manual(
mode: str,
) -> None:
candidate = _candidate("skill:remote", "install")
bundle = None if mode == "missing" else _install_bundle("skill:other")
adapter, _source, _facts, _materials, _installs = _adapter(
(candidate,),
installs={candidate.capability_id: bundle},
)
value = adapter(_observation(), None, _context(adapter)).to_dict()["value"]
assert value["status"] == "abstained"
assert value["capabilities"] == []
def test_rejects_install_bundle_from_a_different_installation_snapshot() -> None:
candidate = _candidate("skill:remote", "install")
adapter, _source, _facts, _materials, _installs = _adapter(
(candidate,),
installs={
candidate.capability_id: _install_bundle(
candidate.capability_id,
provenance_salt="different-install-snapshot",
)
},
)
value = adapter(_observation(), None, _context(adapter)).to_dict()["value"]
assert value["status"] == "abstained"
assert value["capabilities"] == []
def test_rejects_planner_catalog_and_injected_port_snapshot_drift() -> None:
candidate = _candidate("skill:local", "load")
adapter, source, facts, materials, installs = _adapter((candidate,))
with pytest.raises(PlannerValidationError, match="planner version"):
adapter(
_observation(),
None,
PlanningContext(
planner_version="different-planner",
catalog_snapshot_digest=source.catalog_snapshot_digest,
),
)
with pytest.raises(PlannerValidationError, match="catalog snapshot"):
adapter(
_observation(),
None,
PlanningContext(
planner_version="planner-v3",
catalog_snapshot_digest=_digest("different-catalog"),
),
)
for port, attribute, message in (
(source, "catalog_snapshot_digest", "catalog snapshot"),
(facts, "benefit_facts_snapshot_digest", "benefit facts snapshot"),
(materials, "material_snapshot_digest", "material snapshot"),
(installs, "installation_snapshot_digest", "installation snapshot"),
):
setattr(port, attribute, _digest(f"changed:{attribute}"))
with pytest.raises(PlannerValidationError, match=message):
adapter(_observation(), None, _context(adapter))
setattr(port, attribute, getattr(adapter, f"_{attribute}"))
def test_excludes_baseline_and_rejected_before_authenticated_fact_lookup() -> None:
baseline = _candidate("skill:baseline", "load")
rejected = _candidate("agent:rejected", "manual")
selected = _candidate("skill:selected", "load")
adapter, _source, facts, materials, _installs = _adapter((baseline, rejected, selected))
decision = adapter(
_observation(
baseline=(baseline.capability_id,),
rejected=(rejected.capability_id,),
),
None,
_context(adapter),
)
rows = decision.to_dict()["value"]["capabilities"]
assert [row["capability_id"] for row in rows] == [selected.capability_id]
assert facts.calls == [selected.capability_id, selected.capability_id]
assert materials.calls == [selected.capability_id, selected.capability_id]
def test_authoritative_state_active_ids_replace_observed_active_ids_without_exclusion() -> None:
active = _candidate("skill:active", "load")
adapter, source, _facts, _materials, _installs = _adapter((active,))
decision = adapter(
_observation(active=("skill:forged",)),
_active_state(active),
_context(adapter),
)
assert source.observations[0].active_capability_ids == (active.capability_id,)
assert decision.to_dict()["value"]["capabilities"][0]["capability_id"] == active.capability_id
def test_excludes_ambiguous_duplicate_identity_but_deduplicates_exact_rows() -> None:
exact = _candidate("skill:exact", "load")
ambiguous_a = _candidate("skill:ambiguous", "load", source_salt="a")
ambiguous_b = _candidate("skill:ambiguous", "load", source_salt="b")
adapter, _source, facts, _materials, _installs = _adapter(
(ambiguous_a, exact, ambiguous_b, exact),
facts={exact.capability_id: _benefit(exact)},
)
decision = adapter(_observation(), None, _context(adapter))
assert [row["capability_id"] for row in decision.to_dict()["value"]["capabilities"]] == [
exact.capability_id
]
assert facts.calls == [exact.capability_id, exact.capability_id]
@pytest.mark.parametrize("requested_limit", [-1, 6, True])
def test_rejects_invalid_requested_limit_before_retrieval(requested_limit: int) -> None:
candidate = _candidate("skill:local", "load")
adapter, source, _facts, _materials, _installs = _adapter((candidate,))
with pytest.raises(PlannerValidationError, match="requested_limit"):
adapter(
_observation(requested_limit=requested_limit),
None,
_context(adapter),
)
assert source.observations == []
def test_rejects_oversized_candidate_pool() -> None:
repeated = tuple(_candidate(f"skill:item-{index}", "manual") for index in range(513))
adapter, _source, _facts, _materials, _installs = _adapter(repeated)
with pytest.raises(PlannerValidationError, match="candidate pool"):
adapter(_observation(), None, _context(adapter))
def test_rejects_non_current_work_and_unknown_observation_fields() -> None:
adapter, source, _facts, _materials, _installs = _adapter(())
wrong = StructuredSurrogate.create(
schema_id="ctx.observation.other",
schema_version=1,
value={},
)
unknown = StructuredSurrogate.create(
schema_id="ctx.observation.current-work",
schema_version=1,
value={**_observation().to_dict()["value"], "raw_prompt": "raw-prompt"},
)
with pytest.raises(PlannerValidationError, match="current-work"):
adapter(wrong, None, _context(adapter))
with pytest.raises(PlannerValidationError, match="fields"):
adapter(unknown, None, _context(adapter))
assert source.observations == []
def test_decision_surrogate_never_contains_raw_port_data() -> None:
candidate = _candidate("skill:local", "load")
adapter, _source, _facts, _materials, _installs = _adapter((candidate,))
decision = adapter(_observation(), None, _context(adapter))
encoded = decision.to_json()
assert "raw_prompt" not in encoded
assert "filesystem_path" not in encoded
assert "material body" not in encoded
def test_planning_context_digest_binds_the_complete_planning_environment() -> None:
candidate = _candidate("skill:local", "load")
adapter, source, facts, materials, installs = _adapter((candidate,))
expected = hashlib.sha256(
json.dumps(
{
"benefit_facts_snapshot_digest": facts.benefit_facts_snapshot_digest,
"calibration_digest": adapter.planner.policy.calibration_digest,
"catalog_namespace_digest": _digest("catalog-namespace"),
"catalog_retrieval_snapshot_digest": source.catalog_snapshot_digest,
"installation_snapshot_digest": installs.installation_snapshot_digest,
"material_snapshot_digest": materials.material_snapshot_digest,
"planner_version": "planner-v3",
"policy_digest": adapter.planner.policy.policy_digest,
"schema": "ctx.planning-environment-v1",
},
ensure_ascii=True,
allow_nan=False,
separators=(",", ":"),
sort_keys=True,
).encode()
).hexdigest()
assert adapter.catalog_snapshot_digest == expected
assert adapter.catalog_retrieval_snapshot_digest == source.catalog_snapshot_digest
assert adapter.catalog_snapshot_digest != source.catalog_snapshot_digest
with pytest.raises(PlannerValidationError, match="planning environment"):
adapter(
_observation(),
None,
PlanningContext(
planner_version="planner-v3",
catalog_snapshot_digest=source.catalog_snapshot_digest,
),
)
def test_typed_install_bundles_are_supported_uniformly_across_installable_kinds() -> None:
candidates = (
_candidate("skill:remote", "install"),
_candidate("agent:remote", "install"),
_candidate("mcp-server:remote", "install"),
)
adapter, _source, _facts, _materials, _installs = _adapter(candidates)
rows = adapter(_observation(), None, _context(adapter)).to_dict()["value"]["capabilities"]
assert [row["capability_id"] for row in rows] == [
"agent:remote",
"mcp-server:remote",
"skill:remote",
]
assert {row["authority"]["type"] for row in rows} == {"install"}
def test_invalid_install_authority_is_local_to_one_candidate() -> None:
invalid = _candidate("agent:remote", "install")
valid = _candidate("skill:local", "load")
adapter, _source, _facts, _materials, installs = _adapter(
(invalid, valid),
installs={invalid.capability_id: None},
)
rows = adapter(_observation(), None, _context(adapter)).to_dict()["value"]["capabilities"]
assert [row["capability_id"] for row in rows] == [valid.capability_id]
assert installs.calls == [invalid.capability_id, invalid.capability_id]
def test_malformed_install_bundle_is_local_to_one_candidate() -> None:
invalid = _candidate("agent:remote", "install")
valid = _candidate("skill:local", "load")
adapter, _source, _facts, _materials, installs = _adapter(
(invalid, valid),
installs={invalid.capability_id: cast(InstallPlanningBundle, object())},
)
rows = adapter(_observation(), None, _context(adapter)).to_dict()["value"]["capabilities"]
assert [row["capability_id"] for row in rows] == [valid.capability_id]
assert rows[0]["authority"]["type"] == "load"
assert installs.calls == [invalid.capability_id, invalid.capability_id]
def test_direct_load_authority_valid_to_unavailable_is_global_across_observations() -> None:
candidate = _candidate("skill:local", "load")
adapter, source, facts, materials, installs = _adapter((candidate,))
class _UnavailableMaterials(_Materials):
unavailable = False
def load_bundle(
self,
presentation: CapabilityCandidate,
) -> CatalogLoadPlanningBundle | None:
if self.unavailable:
raise CandidateAuthorityUnavailable("injected candidate-local failure")
return super().load_bundle(presentation)
changing = _UnavailableMaterials(
dict(materials.values),
material_snapshot_digest=materials.material_snapshot_digest,
)
adapter = AuthenticatedReplayDecisionPlannerV3(
planner=adapter.planner,
source=source,
benefit_facts_port=facts,
material_port=changing,
install_bundle_port=installs,
planner_version="planner-v3",
catalog_namespace_digest=_digest("catalog-namespace"),
)
adapter(_observation(), None, _context(adapter))
changing.unavailable = True
with pytest.raises(PlannerValidationError, match="load authority output drifted"):
adapter(
_observation(active=("skill:unrelated",)),
None,
_context(adapter),
)
def test_direct_install_authority_valid_to_unavailable_is_global_across_observations() -> None:
candidate = _candidate("skill:remote", "install")
adapter, source, facts, materials, installs = _adapter((candidate,))
class _UnavailableInstalls(_Installs):
unavailable = False
def describe_bundle(
self,
capability_id: str,
kind: str,
) -> InstallPlanningBundle | None:
if self.unavailable:
raise CandidateAuthorityUnavailable("injected candidate-local failure")
return super().describe_bundle(capability_id, kind)
changing = _UnavailableInstalls(
dict(installs.values),
installation_snapshot_digest=installs.installation_snapshot_digest,
)
adapter = AuthenticatedReplayDecisionPlannerV3(
planner=adapter.planner,
source=source,
benefit_facts_port=facts,
material_port=materials,
install_bundle_port=changing,
planner_version="planner-v3",
catalog_namespace_digest=_digest("catalog-namespace"),
)
adapter(_observation(), None, _context(adapter))
changing.unavailable = True
with pytest.raises(PlannerValidationError, match="install authority output drifted"):
adapter(
_observation(active=("skill:unrelated",)),
None,
_context(adapter),
)
def test_load_authority_rejects_a_different_retrieval_presentation() -> None:
candidate = _candidate("skill:local", "load")
substituted = _candidate("skill:local", "load", source_salt="substituted")
adapter, _source, _facts, materials, _installs = _adapter(
(candidate,),
materials={candidate.capability_id: _catalog_load_bundle(substituted)},
)
value = adapter(_observation(), None, _context(adapter)).to_dict()["value"]
assert value["status"] == "abstained"
assert value["capabilities"] == []
assert materials.calls == [candidate.capability_id, candidate.capability_id]
def test_typed_load_bundle_cannot_mix_cross_snapshot_material() -> None:
invalid = _candidate("skill:cross-snapshot", "load")
valid = _candidate("skill:valid-load", "load")
cross_snapshot = _load_descriptor(
invalid.capability_id,
provenance_salt="other-material-snapshot",
)
adapter, _source, _facts, _materials, _installs = _adapter(
(invalid, valid),
materials={
invalid.capability_id: _catalog_load_bundle(
invalid,
descriptor=cross_snapshot,
),
valid.capability_id: _catalog_load_bundle(valid),
},
)
rows = adapter(_observation(), None, _context(adapter)).to_dict()["value"]["capabilities"]
assert [row["capability_id"] for row in rows] == [valid.capability_id]
assert rows[0]["authority"]["type"] == "load"
def test_end_validation_catches_same_digest_output_mutation_on_first_call() -> None:
candidate = _candidate("skill:local", "load")
adapter, source, facts, materials, installs = _adapter((candidate,))
first = materials.values[candidate.capability_id]
replacement = _catalog_load_bundle(
candidate,
descriptor=_load_descriptor(candidate.capability_id, material_salt="replacement"),
)
class _MutatingMaterials(_Materials):
def load_bundle(
self,
presentation: CapabilityCandidate,
) -> CatalogLoadPlanningBundle | None:
value = super().load_bundle(presentation)
if len(self.calls) == 1:
self.values[presentation.capability_id] = replacement
return value
mutating = _MutatingMaterials(
{candidate.capability_id: cast(CatalogLoadPlanningBundle, first)},
material_snapshot_digest=materials.material_snapshot_digest,
)
adapter = AuthenticatedReplayDecisionPlannerV3(
planner=adapter.planner,
source=source,
benefit_facts_port=facts,
material_port=mutating,
install_bundle_port=installs,
planner_version="planner-v3",
catalog_namespace_digest=_digest("catalog-namespace"),
)
with pytest.raises(PlannerValidationError, match="output drifted"):
adapter(_observation(), None, _context(adapter))
def test_port_cannot_hide_concurrent_digest_drift_during_a_read() -> None:
candidate = _candidate("skill:local", "load")
adapter, source, facts, materials, installs = _adapter((candidate,))
class _DriftingMaterials(_Materials):
def load_bundle(
self,
presentation: CapabilityCandidate,
) -> CatalogLoadPlanningBundle | None:
value = super().load_bundle(presentation)
self.material_snapshot_digest = _digest("concurrent-drift")
return value
drifting = _DriftingMaterials(
dict(materials.values),
material_snapshot_digest=materials.material_snapshot_digest,
)
adapter = AuthenticatedReplayDecisionPlannerV3(
planner=adapter.planner,
source=source,
benefit_facts_port=facts,
material_port=drifting,
install_bundle_port=installs,
planner_version="planner-v3",
catalog_namespace_digest=_digest("catalog-namespace"),
)
with pytest.raises(PlannerValidationError, match="material snapshot drifted"):
adapter(_observation(), None, _context(adapter))
def test_snapshot_output_memory_is_bounded_and_never_evicts_drift_evidence(
monkeypatch: pytest.MonkeyPatch,
) -> None:
monkeypatch.setattr(planning_v3_module, "_MAX_PINNED_SNAPSHOT_OUTPUTS", 1)
adapter, _source, _facts, _materials, _installs = _adapter(())
adapter(_observation(requested_limit=5), None, _context(adapter))
with pytest.raises(PlannerValidationError, match="bound is exhausted"):
adapter(_observation(requested_limit=4), None, _context(adapter))
|