okstra 0.234.1 → 0.236.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/cli-registry.mjs +7 -0
- package/dist/cli-registry.mjs.map +1 -1
- package/dist/commands/implementation-report/implementation-report.d.mts +1 -0
- package/dist/commands/implementation-report/implementation-report.mjs +73 -0
- package/dist/commands/implementation-report/implementation-report.mjs.map +1 -0
- package/dist/commands/trace/comparison.d.mts +39 -0
- package/dist/commands/trace/comparison.mjs +138 -0
- package/dist/commands/trace/comparison.mjs.map +1 -0
- package/dist/commands/trace/document.d.mts +8 -0
- package/dist/commands/trace/document.mjs +37 -0
- package/dist/commands/trace/document.mjs.map +1 -0
- package/dist/commands/trace/line-counts.d.mts +15 -0
- package/dist/commands/trace/line-counts.mjs +107 -0
- package/dist/commands/trace/line-counts.mjs.map +1 -0
- package/dist/commands/trace/model.d.mts +2 -1
- package/dist/commands/trace/model.mjs +13 -1
- package/dist/commands/trace/model.mjs.map +1 -1
- package/dist/commands/trace/progression.mjs +7 -3
- package/dist/commands/trace/progression.mjs.map +1 -1
- package/dist/commands/trace/task-link.d.mts +3 -0
- package/dist/commands/trace/task-link.mjs.map +1 -1
- package/dist/commands/trace/trace.mjs +64 -44
- package/dist/commands/trace/trace.mjs.map +1 -1
- package/dist/lib/host-registry-client.d.mts +9 -2
- package/dist/lib/host-registry-client.mjs +29 -3
- package/dist/lib/host-registry-client.mjs.map +1 -1
- package/dist/lib/python-helper.d.mts +1 -1
- package/dist/lib/python-helper.mjs +11 -5
- package/dist/lib/python-helper.mjs.map +1 -1
- package/dist/lib/runtime-payload.mjs +1 -0
- package/dist/lib/runtime-payload.mjs.map +1 -1
- package/dist/lib/types.d.mts +1 -0
- package/docs/architecture/storage-model.md +50 -0
- package/docs/architecture.md +1 -1
- package/docs/cli.md +37 -15
- package/docs/project-structure-overview.md +6 -1
- package/package.json +1 -1
- package/runtime/BUILD.json +2 -2
- package/runtime/agents/operations/trace-tests-review.json +6 -0
- package/runtime/bin/lib/okstra/usage.sh +8 -8
- package/runtime/bin/okstra-qwen-exec.sh +5 -0
- package/runtime/prompts/duties/test-reviewer.json +32 -0
- package/runtime/prompts/lead/convergence.md +4 -1
- package/runtime/prompts/lead/okstra-lead-contract.md +3 -3
- package/runtime/prompts/lead/report-writer.md +13 -1
- package/runtime/prompts/profiles/_common-contract.md +10 -1
- package/runtime/prompts/wizard/prompts.ko.json +16 -4
- package/runtime/python/okstra_ctl/adapters/providers/antigravity/adapter.py +10 -2
- package/runtime/python/okstra_ctl/adapters/providers/claude/adapter.py +16 -1
- package/runtime/python/okstra_ctl/adapters/providers/codex/adapter.py +20 -5
- package/runtime/python/okstra_ctl/adapters/providers/kimi/adapter.py +17 -9
- package/runtime/python/okstra_ctl/adapters/providers/qwen/adapter.py +88 -0
- package/runtime/python/okstra_ctl/adapters/providers/qwen/manifest.json +1 -0
- package/runtime/python/okstra_ctl/adapters/providers/qwen/settings.json +12 -0
- package/runtime/python/okstra_ctl/adapters/providers/zai/adapter.py +8 -2
- package/runtime/python/okstra_ctl/agent/invocation.py +109 -56
- package/runtime/python/okstra_ctl/agent/prompt_cli/materialize.py +17 -0
- package/runtime/python/okstra_ctl/attempt_evidence.py +5 -2
- package/runtime/python/okstra_ctl/business_flow/contracts.py +11 -0
- package/runtime/python/okstra_ctl/business_flow/engine.py +90 -32
- package/runtime/python/okstra_ctl/business_flow/hooks.py +4 -2
- package/runtime/python/okstra_ctl/business_flow/source.py +39 -7
- package/runtime/python/okstra_ctl/business_flow/store.py +94 -28
- package/runtime/python/okstra_ctl/code_review_census.py +38 -0
- package/runtime/python/okstra_ctl/code_review_coverage.py +15 -4
- package/runtime/python/okstra_ctl/code_review_prepare.py +15 -3
- package/runtime/python/okstra_ctl/code_review_scope.py +3 -2
- package/runtime/python/okstra_ctl/code_review_target.py +16 -6
- package/runtime/python/okstra_ctl/conformance_evidence.py +44 -21
- package/runtime/python/okstra_ctl/consumers.py +6 -1
- package/runtime/python/okstra_ctl/convergence.py +95 -5
- package/runtime/python/okstra_ctl/convergence_critic_verify_prompt.py +11 -10
- package/runtime/python/okstra_ctl/convergence_engine.py +30 -15
- package/runtime/python/okstra_ctl/convergence_reverify_prompt.py +9 -2
- package/runtime/python/okstra_ctl/coverage_census.py +75 -12
- package/runtime/python/okstra_ctl/design_review.py +6 -26
- package/runtime/python/okstra_ctl/design_review_evidence.py +6 -26
- package/runtime/python/okstra_ctl/dispatch_core.py +9 -2
- package/runtime/python/okstra_ctl/dispatch_state.py +5 -0
- package/runtime/python/okstra_ctl/domain/role.py +1 -0
- package/runtime/python/okstra_ctl/entrypoints/hosts.py +1 -1
- package/runtime/python/okstra_ctl/execution_manifest.py +8 -0
- package/runtime/python/okstra_ctl/implementation_comparison/__init__.py +12 -0
- package/runtime/python/okstra_ctl/implementation_comparison/__main__.py +129 -0
- package/runtime/python/okstra_ctl/implementation_comparison/assembly.py +340 -0
- package/runtime/python/okstra_ctl/implementation_comparison/baseline.py +254 -0
- package/runtime/python/okstra_ctl/implementation_comparison/business_reuse.py +118 -0
- package/runtime/python/okstra_ctl/implementation_comparison/commit_source.py +113 -0
- package/runtime/python/okstra_ctl/implementation_comparison/html.py +49 -0
- package/runtime/python/okstra_ctl/implementation_comparison/inputs.py +348 -0
- package/runtime/python/okstra_ctl/implementation_comparison/model.py +47 -0
- package/runtime/python/okstra_ctl/implementation_comparison/preparation.py +195 -0
- package/runtime/python/okstra_ctl/implementation_comparison/publication.py +317 -0
- package/runtime/python/okstra_ctl/implementation_comparison/run_evidence.py +208 -0
- package/runtime/python/okstra_ctl/implementation_comparison/run_integration.py +198 -0
- package/runtime/python/okstra_ctl/implementation_comparison/source.py +342 -0
- package/runtime/python/okstra_ctl/implementation_comparison/target.py +144 -0
- package/runtime/python/okstra_ctl/implementation_comparison/task_assembly.py +333 -0
- package/runtime/python/okstra_ctl/implementation_comparison/task_selection.py +309 -0
- package/runtime/python/okstra_ctl/implementation_generation_evidence.py +134 -0
- package/runtime/python/okstra_ctl/implementation_outcome.py +2 -2
- package/runtime/python/okstra_ctl/initial_prompt_materialization.py +1 -0
- package/runtime/python/okstra_ctl/json_boundary.py +28 -8
- package/runtime/python/okstra_ctl/lead_progress.py +1 -1
- package/runtime/python/okstra_ctl/model_discovery.py +1 -1
- package/runtime/python/okstra_ctl/models.py +2 -2
- package/runtime/python/okstra_ctl/paths.py +7 -4
- package/runtime/python/okstra_ctl/phases/discussion/profile.md +2 -2
- package/runtime/python/okstra_ctl/phases/error_analysis/profile.md +1 -1
- package/runtime/python/okstra_ctl/phases/implementation_option_selection/profile.md +1 -1
- package/runtime/python/okstra_ctl/phases/implementation_planning/authoring.py +2 -1
- package/runtime/python/okstra_ctl/phases/implementation_planning/instructions/plan-body-verification.md +1 -1
- package/runtime/python/okstra_ctl/phases/implementation_planning/profile.md +3 -3
- package/runtime/python/okstra_ctl/phases/requirements_discovery/profile.md +5 -5
- package/runtime/python/okstra_ctl/plan_run_root.py +8 -4
- package/runtime/python/okstra_ctl/qa_procedures.py +33 -2
- package/runtime/python/okstra_ctl/registry/provider_registry.py +1 -1
- package/runtime/python/okstra_ctl/render.py +4 -0
- package/runtime/python/okstra_ctl/render_final_report.py +1 -1
- package/runtime/python/okstra_ctl/report_execution_evidence.py +88 -15
- package/runtime/python/okstra_ctl/report_finalize.py +18 -4
- package/runtime/python/okstra_ctl/report_html/render.py +46 -7
- package/runtime/python/okstra_ctl/report_html/validation.py +32 -0
- package/runtime/python/okstra_ctl/report_synthesis_packet.py +37 -3
- package/runtime/python/okstra_ctl/report_view_artifacts.py +12 -0
- package/runtime/python/okstra_ctl/review_verdicts.py +9 -0
- package/runtime/python/okstra_ctl/run.py +17 -1
- package/runtime/python/okstra_ctl/run_context.py +18 -3
- package/runtime/python/okstra_ctl/schema_excerpt.py +14 -9
- package/runtime/python/okstra_ctl/stage_done.py +11 -3
- package/runtime/python/okstra_ctl/stage_ledger.py +2 -2
- package/runtime/python/okstra_ctl/stage_map.py +3 -3
- package/runtime/python/okstra_ctl/trace_review.py +1 -0
- package/runtime/python/okstra_ctl/trace_tasks.py +22 -7
- package/runtime/python/okstra_ctl/verdict_blocks.py +7 -2
- package/runtime/python/okstra_ctl/verification_dispatch.py +32 -0
- package/runtime/python/okstra_ctl/verification_evidence.py +23 -3
- package/runtime/python/okstra_ctl/verification_generation.py +3 -2
- package/runtime/python/okstra_ctl/verification_inputs.py +3 -2
- package/runtime/python/okstra_ctl/verification_invocation.py +7 -3
- package/runtime/python/okstra_ctl/wizard/engine.py +28 -1
- package/runtime/python/okstra_ctl/wizard/ids.py +8 -0
- package/runtime/python/okstra_ctl/wizard/picker_navigation.py +64 -2
- package/runtime/python/okstra_ctl/wizard/roles.py +194 -45
- package/runtime/python/okstra_ctl/wizard/sources.py +10 -1
- package/runtime/python/okstra_ctl/wizard/state.py +67 -20
- package/runtime/python/okstra_ctl/wizard/steps_roles.py +6 -3
- package/runtime/python/okstra_ctl/wizard/worker_settings.py +129 -25
- package/runtime/python/okstra_ctl/worker_audit_ledger.py +44 -8
- package/runtime/python/okstra_ctl/worker_prompt_body.py +26 -2
- package/runtime/python/okstra_ctl/worker_prompt_headers.py +12 -9
- package/runtime/python/okstra_token_usage/claude.py +5 -1
- package/runtime/python/okstra_token_usage/collect.py +3 -3
- package/runtime/python/okstra_token_usage/pricing.py +1 -0
- package/runtime/schemas/report-synthesis-packet-v1.0.schema.json +17 -0
- package/runtime/skills/_fragments/wizard-relay-protocol.md +1 -1
- package/runtime/skills/okstra-code-review/SKILL.md +1 -1
- package/runtime/skills/okstra-code-review/references/census-rules.md +1 -1
- package/runtime/skills/okstra-code-review/references/review-calibration.md +1 -1
- package/runtime/skills/okstra-discussion/SKILL.md +1 -1
- package/runtime/skills/okstra-run/SKILL.md +33 -2
- package/runtime/skills/okstra-tdd/SKILL.md +74 -13
- package/runtime/templates/reports/html/assets/implementation-comparison.css +39 -0
- package/runtime/templates/reports/html/assets/implementation-comparison.js +55 -0
- package/runtime/templates/reports/html/i18n/en.json +113 -0
- package/runtime/templates/reports/html/i18n/ko.json +113 -0
- package/runtime/templates/reports/html/implementation-comparison.template.html +77 -0
- package/runtime/templates/reports/implementation-comparison-notes.md +72 -0
- package/runtime/templates/trace/trace.template.md +2 -0
- package/runtime/templates/worker-prompt-preamble.md +12 -2
- package/runtime/validators/validate-report-views.py +4 -36
|
@@ -93,7 +93,7 @@ _CLI_EPILOG = r"""Usage:
|
|
|
93
93
|
okstra code-review target --task-key <k> --stage <N> [--base <ref>] [--project-root <dir>] [--cwd <dir>] --json
|
|
94
94
|
okstra code-review target --branch <name> --base <ref> [--date <YYYY-MM-DD>] [--project-root <dir>] --json
|
|
95
95
|
okstra code-review check-coverage --head <ref> [--base <ref>] --census <table-path> --completion <completion-path> [--completion ...] [--project-root <dir>]
|
|
96
|
-
okstra code-review census --base <ref> --head <ref> --rules <review-rules.json> --out <artifact-path> [--policy <path> ...] [--run-manifest <path>] [--resource <pack-relative-resource> ...]
|
|
96
|
+
okstra code-review census --base <ref> --head <ref> --rules <review-rules.json> --out <artifact-path> [--policy <path> ...] [--run-manifest <path>] [--resource <pack-relative-resource> ...] [--function-inventory <path>]
|
|
97
97
|
okstra code-review assign --census <review-id>.census.md [--project-root <dir>]
|
|
98
98
|
okstra code-review prepare --review-id <id> --census <id>.census.md --review-policy <path> --base <commit> --head <commit> --slots <slots.json> --brief <path> --host-runtime <host> [--resource <id> ...]
|
|
99
99
|
okstra code-review start --batch <id>.batch.json [--max-calls <n>]
|
|
@@ -118,7 +118,9 @@ base evidence at `--base`, never the working tree, prints covered cells, missing
|
|
|
118
118
|
prints its Review policy path/digest, and selects its shared catalog with the
|
|
119
119
|
captured architecture. --run-manifest reuses frozen implementation rules;
|
|
120
120
|
--policy captures additional applied standards. It exits 1 when changed
|
|
121
|
-
non-Python code still needs function cells
|
|
121
|
+
non-Python code still needs function cells. --function-inventory supplies
|
|
122
|
+
them: a missing file is written as a template, and a filled one lists
|
|
123
|
+
{"functions": [{"start", "end", "name"}], "rationale"} per file. `assign` writes one
|
|
122
124
|
<review-id>.<axis>.census.md per axis that has cells beside the census and
|
|
123
125
|
prints each axis's cell count and path (`-` for an axis without cells); an
|
|
124
126
|
existing assignment is reused only when identical. `prepare` assigns the census,
|
|
@@ -171,6 +173,8 @@ def _build_parser() -> argparse.ArgumentParser:
|
|
|
171
173
|
parser.add_argument("--rules", default="",
|
|
172
174
|
help="census: canonical shared review-rules.json provenance")
|
|
173
175
|
parser.add_argument("--out", default="", help="census: file to write the census to")
|
|
176
|
+
parser.add_argument("--function-inventory", default="",
|
|
177
|
+
help="census: JSON function spans for files whose boundaries it cannot read")
|
|
174
178
|
parser.add_argument("--policy", action="append", default=[],
|
|
175
179
|
help="census: additional policy to freeze; scope: captured policy to fingerprint")
|
|
176
180
|
parser.add_argument("--review-policy", default="", help="scope: shared frozen review policy")
|
|
@@ -430,7 +434,7 @@ def main(argv: list[str] | None = None) -> int:
|
|
|
430
434
|
|
|
431
435
|
|
|
432
436
|
def _census(args: argparse.Namespace, project_root: Path) -> int:
|
|
433
|
-
from .code_review_census import
|
|
437
|
+
from .code_review_census import build_review_census_with_inventory
|
|
434
438
|
from .code_review_coverage import read_review_census
|
|
435
439
|
from .review_policy import render_review_rule_catalog, select_review_policy
|
|
436
440
|
|
|
@@ -450,9 +454,15 @@ def _census(args: argparse.Namespace, project_root: Path) -> int:
|
|
|
450
454
|
policy, tuple(changed), project_paths=paths, manifest_contents=manifests,
|
|
451
455
|
resources=args.resource,
|
|
452
456
|
)
|
|
453
|
-
|
|
454
|
-
|
|
455
|
-
|
|
457
|
+
inventory_path = None
|
|
458
|
+
if args.function_inventory:
|
|
459
|
+
inventory_path = Path(args.function_inventory).absolute()
|
|
460
|
+
# The inventory binds commits, so a later commit cannot reuse spans read
|
|
461
|
+
# from an earlier one under the same ref name.
|
|
462
|
+
result, _ = build_review_census_with_inventory(
|
|
463
|
+
project_root, _resolve_base_commit(project_root, args.base),
|
|
464
|
+
_resolve_base_commit(project_root, args.head), Path(args.rules), changed,
|
|
465
|
+
inventory_path, frozen_rules=render_review_rule_catalog(selected),
|
|
456
466
|
previous=Path(args.previous_census) if args.previous_census else None,
|
|
457
467
|
)
|
|
458
468
|
selection = (f"Review policy digest: {digest}\nReview resources: "
|
|
@@ -9,20 +9,20 @@ from __future__ import annotations
|
|
|
9
9
|
from contextlib import contextmanager
|
|
10
10
|
from dataclasses import dataclass, field
|
|
11
11
|
from pathlib import Path
|
|
12
|
-
from typing import Any, Iterable, Iterator, Mapping
|
|
12
|
+
from typing import Any, Callable, Iterable, Iterator, Mapping, cast
|
|
13
13
|
|
|
14
14
|
from .agent.invocation import AgentInvocationError, reserved_generation_reference
|
|
15
15
|
from .conformance import QaResult, execution_evidence_defect, qa_result_from_execution
|
|
16
16
|
from .execution_identity import ExecutionManifest, Invocation
|
|
17
|
-
from .execution_manifest import read_execution_manifest
|
|
18
|
-
from .json_boundary import JsonBoundaryError, load_owned_object
|
|
17
|
+
from .execution_manifest import read_execution_manifest, execution_manifest_view_from_payload
|
|
18
|
+
from .json_boundary import ArtifactByteReader, JsonBoundaryError, load_owned_object
|
|
19
19
|
from .paths import RunRef
|
|
20
20
|
from .qa_commands import QA_VERIFIER_AUDIENCE_BY_TASK_TYPE
|
|
21
21
|
from .verification_conformance import conformance_stage
|
|
22
22
|
from .verification_dispatch import verification_generation_number
|
|
23
23
|
from .verification_evidence import (
|
|
24
24
|
VerificationExecutionRecord, conformance_check_owner_ref,
|
|
25
|
-
|
|
25
|
+
read_verification_execution_snapshot,
|
|
26
26
|
)
|
|
27
27
|
from .verification_generation import (
|
|
28
28
|
VerificationGenerationRecord, read_verification_generation,
|
|
@@ -89,6 +89,7 @@ class RunVerifiers:
|
|
|
89
89
|
ref: RunRef
|
|
90
90
|
verifiers: dict[str, list[BoundInvocation]]
|
|
91
91
|
latest_attempt: dict[str, int]
|
|
92
|
+
execution: ExecutionManifest | None = None
|
|
92
93
|
|
|
93
94
|
|
|
94
95
|
@dataclass(frozen=True)
|
|
@@ -101,6 +102,7 @@ class CheckRunRecord:
|
|
|
101
102
|
record: VerificationExecutionRecord | None
|
|
102
103
|
output: bytes | None
|
|
103
104
|
defect: str | None
|
|
105
|
+
raw_bytes: bytes | None = None
|
|
104
106
|
|
|
105
107
|
|
|
106
108
|
@contextmanager
|
|
@@ -131,18 +133,21 @@ def _collect(
|
|
|
131
133
|
})
|
|
132
134
|
|
|
133
135
|
|
|
134
|
-
def load_run_verifiers(project: Path, manifest_path: Path) -> RunVerifiers:
|
|
135
|
-
manifest = load_owned_object(manifest_path, artifact="run manifest")
|
|
136
|
+
def load_run_verifiers(project: Path, manifest_path: Path, *, reader: ArtifactByteReader | None = None) -> RunVerifiers:
|
|
137
|
+
manifest = load_owned_object(manifest_path, artifact="run manifest", reader=reader)
|
|
136
138
|
ref = RunRef.from_manifest(manifest_path, manifest)
|
|
137
139
|
audience = QA_VERIFIER_AUDIENCE_BY_TASK_TYPE.get(manifest.get("taskType"))
|
|
138
140
|
if audience is None:
|
|
139
141
|
raise ValueError("verification evidence requires a QA-executing run")
|
|
140
|
-
|
|
142
|
+
if reader is None:
|
|
143
|
+
execution = read_execution_manifest(manifest_path)
|
|
144
|
+
else:
|
|
145
|
+
execution, _ = execution_manifest_view_from_payload(manifest)
|
|
141
146
|
latest_attempt = _latest_attempts(execution)
|
|
142
147
|
return RunVerifiers(
|
|
143
148
|
manifest, ref,
|
|
144
|
-
_bound_verifiers(project, ref, manifest, execution, audience, latest_attempt),
|
|
145
|
-
latest_attempt,
|
|
149
|
+
_bound_verifiers(project, ref, manifest, execution, audience, latest_attempt, reader=reader),
|
|
150
|
+
latest_attempt, execution,
|
|
146
151
|
)
|
|
147
152
|
|
|
148
153
|
|
|
@@ -171,6 +176,7 @@ def _stage_evidence(
|
|
|
171
176
|
def _bound_verifiers(
|
|
172
177
|
project: Path, ref: RunRef, manifest: Mapping[str, Any],
|
|
173
178
|
execution: ExecutionManifest, audience: str, latest_attempt: Mapping[str, int],
|
|
179
|
+
*, reader: ArtifactByteReader | None = None,
|
|
174
180
|
) -> dict[str, list[BoundInvocation]]:
|
|
175
181
|
verifiers: dict[str, list[BoundInvocation]] = {}
|
|
176
182
|
for index, invocation in enumerate(execution.invocations):
|
|
@@ -178,6 +184,7 @@ def _bound_verifiers(
|
|
|
178
184
|
verifiers.setdefault(invocation.role_execution_ref, []).append(_bind(
|
|
179
185
|
project, ref, manifest, invocation, index,
|
|
180
186
|
latest_attempt.get(invocation.invocation_ref, 1),
|
|
187
|
+
reader=reader,
|
|
181
188
|
))
|
|
182
189
|
for bindings in verifiers.values():
|
|
183
190
|
bindings.sort(key=lambda binding: (binding.generation_number, binding.index),
|
|
@@ -203,6 +210,7 @@ def _required_stages(verifiers: Mapping[str, list[BoundInvocation]]) -> frozense
|
|
|
203
210
|
def _bind(
|
|
204
211
|
project: Path, ref: RunRef, manifest: Mapping[str, Any], invocation: Invocation,
|
|
205
212
|
index: int, attempt: int,
|
|
213
|
+
*, reader: ArtifactByteReader | None = None,
|
|
206
214
|
) -> BoundInvocation:
|
|
207
215
|
"""The generation the invocation's metadata binds, which must be the one its
|
|
208
216
|
dispatch kind seals and still have the bytes the metadata recorded."""
|
|
@@ -211,6 +219,7 @@ def _bind(
|
|
|
211
219
|
relative = path.relative_to(project).as_posix()
|
|
212
220
|
reference = reserved_generation_reference(
|
|
213
221
|
project, manifest, invocation.invocation_ref, attempt,
|
|
222
|
+
**({"reader": reader} if reader is not None else {}),
|
|
214
223
|
)
|
|
215
224
|
if reference is None:
|
|
216
225
|
return BoundInvocation(invocation, number, index, None,
|
|
@@ -221,7 +230,7 @@ def _bind(
|
|
|
221
230
|
f"not to {relative}, the generation its dispatch seals"
|
|
222
231
|
)
|
|
223
232
|
try:
|
|
224
|
-
generation = read_verification_generation(path, reference.digest)
|
|
233
|
+
generation = read_verification_generation(path, reference.digest, **({"reader": reader} if reader is not None else {}))
|
|
225
234
|
except ValueError as exc:
|
|
226
235
|
raise ValueError(
|
|
227
236
|
f"generation {relative} of invocation {invocation.invocation_ref}: {exc}"
|
|
@@ -263,40 +272,54 @@ def bound_attempt(
|
|
|
263
272
|
def read_latest_check_run(
|
|
264
273
|
ref: RunRef, role_execution_ref: str, binding: BoundInvocation, attempt: int,
|
|
265
274
|
check_ref: str,
|
|
275
|
+
*, reader: ArtifactByteReader | None = None,
|
|
276
|
+
directory_entries: Callable[[Path], Iterable[Path]] | None = None,
|
|
266
277
|
) -> CheckRunRecord | None:
|
|
267
278
|
"""The latest run of one check under an attempt of a bound invocation; None
|
|
268
279
|
when the attempt never ran the check."""
|
|
269
280
|
invocation_ref = binding.invocation.invocation_ref
|
|
270
|
-
check_run = ref.latest_verification_check_run(invocation_ref, attempt, check_ref
|
|
281
|
+
check_run = ref.latest_verification_check_run(invocation_ref, attempt, check_ref,
|
|
282
|
+
**({"directory_entries": directory_entries} if directory_entries is not None else {}))
|
|
271
283
|
if check_run is None:
|
|
272
284
|
return None
|
|
273
285
|
run_dir = ref.verification_execution_dir(invocation_ref, attempt, check_ref, check_run)
|
|
274
|
-
record, output, defect = _read_bound_record(run_dir, {
|
|
286
|
+
record, output, defect, raw = _read_bound_record(run_dir, {
|
|
275
287
|
"invocationRef": invocation_ref, "roleExecutionRef": role_execution_ref,
|
|
276
288
|
"attempt": attempt, "checkRef": check_ref, "checkRun": check_run,
|
|
277
289
|
"verificationGeneration": binding.generation_ref,
|
|
278
|
-
})
|
|
279
|
-
return CheckRunRecord(check_run, run_dir / "record.json", record, output, defect)
|
|
290
|
+
}, reader=reader)
|
|
291
|
+
return CheckRunRecord(check_run, run_dir / "record.json", record, output, defect, raw)
|
|
280
292
|
|
|
281
293
|
|
|
282
294
|
def _read_bound_record(
|
|
283
295
|
run_dir: Path, expected: Mapping[str, object],
|
|
284
|
-
|
|
296
|
+
*, reader: ArtifactByteReader | None = None,
|
|
297
|
+
) -> tuple[VerificationExecutionRecord | None, bytes | None, str | None, bytes | None]:
|
|
285
298
|
"""One check run's record, its saved stdout, and why the run is no usable
|
|
286
299
|
evidence (None when it is); `expected` are the record fields its location
|
|
287
300
|
and binding determine."""
|
|
301
|
+
raw: bytes | None = None
|
|
302
|
+
|
|
303
|
+
def consumed_record(path: Path) -> bytes:
|
|
304
|
+
nonlocal raw
|
|
305
|
+
body = path.read_bytes() if reader is None else reader(path)
|
|
306
|
+
raw = body
|
|
307
|
+
return body
|
|
308
|
+
|
|
288
309
|
try:
|
|
289
|
-
|
|
310
|
+
stored = read_verification_execution_snapshot(run_dir / "record.json", reader=consumed_record)
|
|
311
|
+
record = cast(VerificationExecutionRecord, stored.value)
|
|
290
312
|
except (JsonBoundaryError, OSError) as exc:
|
|
291
|
-
return None, None, f"record is unreadable ({exc})"
|
|
313
|
+
return None, None, f"record is unreadable ({exc})", raw
|
|
292
314
|
if any(record[name] != value for name, value in expected.items()):
|
|
293
315
|
return record, None, ("record identity, attempt, check run or generation "
|
|
294
|
-
"differs from its location")
|
|
316
|
+
"differs from its location"), raw
|
|
295
317
|
try:
|
|
296
|
-
|
|
318
|
+
output_path = run_dir / "stdout.log"
|
|
319
|
+
output = output_path.read_bytes() if reader is None else reader(output_path)
|
|
297
320
|
except OSError as exc:
|
|
298
|
-
return record, None, f"saved stdout is unreadable ({exc})"
|
|
299
|
-
return record, output, execution_evidence_defect(record, output)
|
|
321
|
+
return record, None, f"saved stdout is unreadable ({exc})", raw
|
|
322
|
+
return record, output, execution_evidence_defect(record, output), raw
|
|
300
323
|
|
|
301
324
|
|
|
302
325
|
def _combine(outcomes: list[_VerifierOutcome]) -> StageEvidence:
|
|
@@ -83,8 +83,13 @@ def latest_done_by_stage(rows: List[Dict[str, Any]]) -> Dict[int, Dict[str, Any]
|
|
|
83
83
|
`started` 를 다시 기록하는 것은 정상 흐름이고(`_equivalent_row_exists`),
|
|
84
84
|
그동안 의존 stage 는 종전대로 마지막 done head 에서 base 를 잡는다.
|
|
85
85
|
철회는 리드의 명시적 판정인 `failed` 하나로만 일어난다.
|
|
86
|
+
|
|
87
|
+
Only a later done row supersedes a failed settlement; started never restores
|
|
88
|
+
completion after that withdrawal.
|
|
86
89
|
"""
|
|
87
|
-
last = last_lifecycle_status_by_stage(
|
|
90
|
+
last = last_lifecycle_status_by_stage([
|
|
91
|
+
row for row in rows if row.get("status") in _LEAD_SETTLED_STATUSES
|
|
92
|
+
])
|
|
88
93
|
out: Dict[int, Dict[str, Any]] = {}
|
|
89
94
|
for r in rows:
|
|
90
95
|
if r.get("status") == "done" and isinstance(r.get("stage"), int):
|
|
@@ -1592,6 +1592,61 @@ def _recorded_results(
|
|
|
1592
1592
|
return result_args
|
|
1593
1593
|
|
|
1594
1594
|
|
|
1595
|
+
# Verdicts that accept the cited evidence. `_grounded_vote` also checks a
|
|
1596
|
+
# burden-not-met refusal, which re-inspected it; counter-evidence stands on what
|
|
1597
|
+
# the verifier found instead.
|
|
1598
|
+
_GROUNDED_VERDICTS = ("agree", "supplement")
|
|
1599
|
+
|
|
1600
|
+
|
|
1601
|
+
def _origin_evidence_paths(plan_path: Path) -> dict[str, frozenset[str]]:
|
|
1602
|
+
"""Each queued finding's cited files, from the working state beside the round plan.
|
|
1603
|
+
|
|
1604
|
+
``plan-round`` writes `convergence-round-<N>-plan-<suffix>.json` next to the
|
|
1605
|
+
`convergence-work-<suffix>.json` it planned from. A plan named otherwise has
|
|
1606
|
+
no working state to read, and its votes are collected as given.
|
|
1607
|
+
"""
|
|
1608
|
+
match = re.fullmatch(r"convergence-round-\d+-plan-(?P<suffix>.+)\.json", Path(plan_path).name)
|
|
1609
|
+
path = Path(plan_path).with_name(f"convergence-work-{match['suffix']}.json") if match else None
|
|
1610
|
+
if path is None or not path.is_file():
|
|
1611
|
+
return {}
|
|
1612
|
+
from .worker_audit_ledger import named_files
|
|
1613
|
+
|
|
1614
|
+
findings = load_owned_json_object(path).get("findings") or []
|
|
1615
|
+
return {
|
|
1616
|
+
str(row["findingId"]): frozenset(named_files(str(row.get("originEvidence") or "")))
|
|
1617
|
+
for row in findings if isinstance(row, Mapping) and row.get("findingId")
|
|
1618
|
+
}
|
|
1619
|
+
|
|
1620
|
+
|
|
1621
|
+
def _grounded_vote(vote: FindingVote, origin: frozenset[str] | None) -> dict[str, Any]:
|
|
1622
|
+
"""The vote as collected; one that opened none of the cited files counts as unverifiable.
|
|
1623
|
+
|
|
1624
|
+
Evidence that names no file (a query result, prose) leaves nothing to compare,
|
|
1625
|
+
so such a vote is collected as given.
|
|
1626
|
+
"""
|
|
1627
|
+
from .worker_audit_ledger import recorded_as_opened
|
|
1628
|
+
|
|
1629
|
+
row = {
|
|
1630
|
+
"verdict": vote.verdict,
|
|
1631
|
+
"disagreeBasis": vote.disagree_basis,
|
|
1632
|
+
"explanation": vote.explanation,
|
|
1633
|
+
}
|
|
1634
|
+
leans_on_origin = vote.verdict in _GROUNDED_VERDICTS or vote.disagree_basis == "burden-not-met"
|
|
1635
|
+
if not origin or not leans_on_origin:
|
|
1636
|
+
return row
|
|
1637
|
+
if any(recorded_as_opened(cited, frozenset(vote.opened)) for cited in origin):
|
|
1638
|
+
return row
|
|
1639
|
+
return {
|
|
1640
|
+
"verdict": "unverifiable",
|
|
1641
|
+
"disagreeBasis": None,
|
|
1642
|
+
"explanation": (
|
|
1643
|
+
"Collected as unverifiable: the Opened line names none of the cited files ("
|
|
1644
|
+
+ ", ".join(sorted(origin)) + "). " + vote.explanation
|
|
1645
|
+
),
|
|
1646
|
+
}
|
|
1647
|
+
|
|
1648
|
+
|
|
1649
|
+
|
|
1595
1650
|
def _collect_results(args: argparse.Namespace) -> tuple[str, Path]:
|
|
1596
1651
|
plan = load_owned_json_object(args.plan)
|
|
1597
1652
|
planned = _planned_finding_ids(plan)
|
|
@@ -1624,14 +1679,13 @@ def _collect_results(args: argparse.Namespace) -> tuple[str, Path]:
|
|
|
1624
1679
|
"close their attempt as ok — a discarded or failed attempt's file is not "
|
|
1625
1680
|
"a vote"
|
|
1626
1681
|
)
|
|
1682
|
+
origins = _origin_evidence_paths(args.plan)
|
|
1627
1683
|
by_finding: dict[str, dict[str, Any]] = {}
|
|
1628
1684
|
for worker, parsed in votes.items():
|
|
1629
1685
|
for finding_id, vote in parsed.items():
|
|
1630
|
-
by_finding.setdefault(finding_id, {})[worker] =
|
|
1631
|
-
|
|
1632
|
-
|
|
1633
|
-
"explanation": vote.explanation,
|
|
1634
|
-
}
|
|
1686
|
+
by_finding.setdefault(finding_id, {})[worker] = _grounded_vote(
|
|
1687
|
+
vote, origins.get(finding_id),
|
|
1688
|
+
)
|
|
1635
1689
|
write_json_atomic(args.output, {
|
|
1636
1690
|
"schemaVersion": "1.0",
|
|
1637
1691
|
"round": plan.get("round"),
|
|
@@ -2123,6 +2177,36 @@ def _read_if_present(path: Path) -> str | None:
|
|
|
2123
2177
|
return path.read_text(encoding="utf-8") if path.is_file() else None
|
|
2124
2178
|
|
|
2125
2179
|
|
|
2180
|
+
def _evidence_ledger(result_path: Path) -> frozenset[str] | None:
|
|
2181
|
+
"""The Evidence read paths of a result's audit sidecar.
|
|
2182
|
+
|
|
2183
|
+
``None`` when the result's prompt asked for no ledger — the rule
|
|
2184
|
+
validate-run applies — so a result of such a prompt is not checked. A prompt
|
|
2185
|
+
that asked for one and a sidecar without rows give an empty ledger.
|
|
2186
|
+
"""
|
|
2187
|
+
from .worker_artifact_paths import WorkerArtifactPathError, audit_sidecar_rel
|
|
2188
|
+
from .worker_audit_ledger import audit_evidence_read_paths, ledger_required
|
|
2189
|
+
|
|
2190
|
+
prefix, separator, rest = result_path.name.partition("-worker-")
|
|
2191
|
+
prompt = result_path.parent.parent / "prompts" / f"{prefix}-worker-prompt-{rest}"
|
|
2192
|
+
if not separator or not ledger_required(prompt):
|
|
2193
|
+
return None
|
|
2194
|
+
try:
|
|
2195
|
+
sidecar = result_path.with_name(audit_sidecar_rel(result_path.name))
|
|
2196
|
+
except WorkerArtifactPathError:
|
|
2197
|
+
return None
|
|
2198
|
+
return frozenset(audit_evidence_read_paths(_read_if_present(sidecar) or ""))
|
|
2199
|
+
|
|
2200
|
+
|
|
2201
|
+
def _joined_ledger(
|
|
2202
|
+
first: frozenset[str] | None, gapfill: frozenset[str] | None,
|
|
2203
|
+
) -> frozenset[str] | None:
|
|
2204
|
+
"""A gap-fill without its own ledger is not checked: it may cite files it opened anew."""
|
|
2205
|
+
if gapfill is None:
|
|
2206
|
+
return None
|
|
2207
|
+
return gapfill | (first or frozenset())
|
|
2208
|
+
|
|
2209
|
+
|
|
2126
2210
|
def _census_audit(args: argparse.Namespace) -> dict[str, Any]:
|
|
2127
2211
|
"""Audit the census verdicts. Unjudged cells are data, never an error."""
|
|
2128
2212
|
authority = validated_run_authority(args.run_manifest)
|
|
@@ -2141,10 +2225,16 @@ def _census_audit(args: argparse.Namespace) -> dict[str, Any]:
|
|
|
2141
2225
|
raise ConvergenceContractError(
|
|
2142
2226
|
f"--gapfill names {unrostered}, which no --result names"
|
|
2143
2227
|
)
|
|
2228
|
+
first_ledgers = {worker: _evidence_ledger(path) for worker, path in result_paths.items()}
|
|
2144
2229
|
audit = audit_census(
|
|
2145
2230
|
census,
|
|
2146
2231
|
{worker: _read_if_present(path) for worker, path in result_paths.items()},
|
|
2147
2232
|
{worker: _read_if_present(path) for worker, path in gapfill_paths.items()},
|
|
2233
|
+
opened=first_ledgers,
|
|
2234
|
+
gapfill_opened={
|
|
2235
|
+
worker: _joined_ledger(first_ledgers.get(worker), _evidence_ledger(path))
|
|
2236
|
+
for worker, path in gapfill_paths.items()
|
|
2237
|
+
},
|
|
2148
2238
|
)
|
|
2149
2239
|
for row in audit["workers"]:
|
|
2150
2240
|
row["resultPath"] = relative_to_project_root(
|
|
@@ -9,7 +9,7 @@
|
|
|
9
9
|
|
|
10
10
|
이 모듈이 그 지시문을 만든다. 입력은 리드가 `apply-critic-gaps` 에 줄 것과 같은
|
|
11
11
|
coverage 배치(투표 전, `gaps[]` 만 채운 상태)와 Round 0 그룹(분석 로스터)이다.
|
|
12
|
-
배정은 엔진과 같은 규칙(`critic_gap_assignees
|
|
12
|
+
배정은 엔진과 같은 규칙(`critic_gap_assignees`: all analysers up to two, otherwise round-robin)이라
|
|
13
13
|
`apply-critic-gaps` 의 커버리지 검사와 어긋나지 않는다. gap 마다 critic 결과
|
|
14
14
|
파일의 `### [<gapId>]` 절과 critic 감사 사이드카를 실어, 검증자가 리드의 전사가
|
|
15
15
|
아니라 critic 이 실제로 인용한 것을 판단하게 한다. 응답 형식은 번호 라운드와
|
|
@@ -131,10 +131,11 @@ def critic_verify_gaps(
|
|
|
131
131
|
) -> list[CriticGap]:
|
|
132
132
|
"""이 워커에게 배정된 gap 을 배치 순서대로.
|
|
133
133
|
|
|
134
|
-
|
|
135
|
-
|
|
136
|
-
|
|
137
|
-
|
|
134
|
+
The assignment is `critic_gap_assignees`: every analyser verifies every gap
|
|
135
|
+
on a roster of `FULL_GAP_VERIFICATION_ROSTER` or fewer, otherwise gap *i*
|
|
136
|
+
goes to `roster[i % len(roster)]`. The roster is the Round 0 groups'
|
|
137
|
+
analysis workers in order, the order `apply-critic-gaps` records as
|
|
138
|
+
`analyserRoster`.
|
|
138
139
|
"""
|
|
139
140
|
gaps_value = batch.get("gaps")
|
|
140
141
|
if not isinstance(gaps_value, list) or not gaps_value:
|
|
@@ -153,17 +154,17 @@ def critic_verify_gaps(
|
|
|
153
154
|
gap for gap in gaps_value
|
|
154
155
|
if isinstance(gap, Mapping) and not gap.get("duplicateOf")
|
|
155
156
|
]
|
|
156
|
-
|
|
157
|
+
verifiers_by_gap = critic_gap_assignees(roster, verifiable)
|
|
157
158
|
assigned: list[CriticGap] = []
|
|
158
159
|
seen: set[str] = set()
|
|
159
|
-
for gap,
|
|
160
|
+
for gap, verifiers in zip(verifiable, verifiers_by_gap):
|
|
160
161
|
gap_id = _nonempty_string(gap.get("gapId"))
|
|
161
162
|
if not gap_id:
|
|
162
163
|
raise CriticVerifyPromptError("coverage batch gap has no gapId")
|
|
163
164
|
if gap_id in seen:
|
|
164
165
|
raise CriticVerifyPromptError(f"duplicate critic gapId: {gap_id}")
|
|
165
166
|
seen.add(gap_id)
|
|
166
|
-
if
|
|
167
|
+
if worker_id not in verifiers:
|
|
167
168
|
continue
|
|
168
169
|
ticket_ids = gap.get("ticketIds")
|
|
169
170
|
assigned.append(CriticGap(
|
|
@@ -177,8 +178,8 @@ def critic_verify_gaps(
|
|
|
177
178
|
))
|
|
178
179
|
if not assigned:
|
|
179
180
|
raise CriticVerifyPromptError(
|
|
180
|
-
f"
|
|
181
|
-
|
|
181
|
+
f"no gap is assigned to `{worker_id}`; verifiers in batch order: "
|
|
182
|
+
+ (", ".join("+".join(verifiers) for verifiers in verifiers_by_gap) or "none")
|
|
182
183
|
)
|
|
183
184
|
return assigned
|
|
184
185
|
|
|
@@ -1136,26 +1136,34 @@ def _parse_critic_dispatches(
|
|
|
1136
1136
|
return dispatches, completed
|
|
1137
1137
|
|
|
1138
1138
|
|
|
1139
|
+
# With this many analysers or fewer, every gap goes to all of them: one
|
|
1140
|
+
# verifier per gap lets a single shallow pass decide what merges into the report.
|
|
1141
|
+
FULL_GAP_VERIFICATION_ROSTER = 2
|
|
1142
|
+
|
|
1143
|
+
|
|
1139
1144
|
def critic_gap_assignees(
|
|
1140
1145
|
roster_order: list[str],
|
|
1141
1146
|
gaps: list[Mapping[str, Any]],
|
|
1142
|
-
) -> list[str]:
|
|
1143
|
-
"""
|
|
1147
|
+
) -> list[tuple[str, ...]]:
|
|
1148
|
+
"""The verifiers of each gap, in batch order.
|
|
1144
1149
|
|
|
1145
|
-
|
|
1146
|
-
|
|
1147
|
-
|
|
1148
|
-
|
|
1150
|
+
A roster of `FULL_GAP_VERIFICATION_ROSTER` or fewer verifies every gap in
|
|
1151
|
+
full. A larger roster is walked round-robin, one verifier per gap: a
|
|
1152
|
+
rejection costs the same as an acceptance in a stretch recorded as
|
|
1153
|
+
`rounds: []`, so each analyser reads only its share.
|
|
1149
1154
|
|
|
1150
|
-
|
|
1155
|
+
Declared duplicates (`duplicateOf`) are not assigned; they are not verified.
|
|
1151
1156
|
"""
|
|
1152
1157
|
verifiable = [gap for gap in gaps if not gap.get("duplicateOf")]
|
|
1153
1158
|
if not roster_order:
|
|
1154
1159
|
return []
|
|
1155
|
-
|
|
1156
|
-
|
|
1157
|
-
|
|
1158
|
-
|
|
1160
|
+
if len(roster_order) <= FULL_GAP_VERIFICATION_ROSTER:
|
|
1161
|
+
return [tuple(roster_order) for _ in verifiable]
|
|
1162
|
+
return [(worker,) for worker in _round_robin_assignees(roster_order, len(verifiable))]
|
|
1163
|
+
|
|
1164
|
+
|
|
1165
|
+
def _round_robin_assignees(roster_order: list[str], count: int) -> list[str]:
|
|
1166
|
+
return [roster_order[index % len(roster_order)] for index in range(count)]
|
|
1159
1167
|
|
|
1160
1168
|
|
|
1161
1169
|
def _critic_gap_coverage_errors(
|
|
@@ -1174,13 +1182,20 @@ def _critic_gap_coverage_errors(
|
|
|
1174
1182
|
계약 위반이 아니라 더 본 것이고, 그 사이 방치된 gap 은 표가 없어
|
|
1175
1183
|
`unverifiedGaps` 로 스스로 드러난다.
|
|
1176
1184
|
"""
|
|
1177
|
-
assignees =
|
|
1178
|
-
|
|
1185
|
+
assignees = {
|
|
1186
|
+
worker for verifiers in critic_gap_assignees(roster_order, gaps) for worker in verifiers
|
|
1187
|
+
}
|
|
1188
|
+
# A run finished under the earlier one-verifier rule dispatched the round-robin set.
|
|
1189
|
+
earlier = set(_round_robin_assignees(
|
|
1190
|
+
roster_order, len([gap for gap in gaps if not gap.get("duplicateOf")]),
|
|
1191
|
+
)) if roster_order else set()
|
|
1192
|
+
if dispatched in (set(roster_order), assignees, earlier):
|
|
1179
1193
|
return []
|
|
1180
1194
|
missing = sorted(assignees - dispatched)
|
|
1181
1195
|
return [
|
|
1182
|
-
"critic dispatches must cover every gap
|
|
1183
|
-
"
|
|
1196
|
+
"critic dispatches must cover every gap's assigned verifiers (all analysers "
|
|
1197
|
+
f"up to {FULL_GAP_VERIFICATION_ROSTER}, otherwise round-robin) or account "
|
|
1198
|
+
"for every non-critic analyser exactly "
|
|
1184
1199
|
f"once; missing: {', '.join(missing) or 'none'}"
|
|
1185
1200
|
]
|
|
1186
1201
|
|
|
@@ -94,7 +94,14 @@ ADVERSARIAL_RESPONSE = """### <finding-id>
|
|
|
94
94
|
**Basis** (only if REFUTED): counter-evidence | burden-not-met
|
|
95
95
|
**Explanation** (required for every verdict, SURVIVES included): <2-3 sentences; for
|
|
96
96
|
SURVIVES say what you attacked and why the attack failed; for counter-evidence include
|
|
97
|
-
the file:line you found>
|
|
97
|
+
the file:line you found>
|
|
98
|
+
**Opened**: <every file you opened for this finding, as path:line, comma-separated>"""
|
|
99
|
+
|
|
100
|
+
# `okstra convergence collect-results` enforces this for round votes
|
|
101
|
+
# (`convergence._grounded_vote`); critic and acceptance votes are not collected there.
|
|
102
|
+
OPENED_RULE = """A SURVIVES, SURVIVES-WITH-CAVEAT or burden-not-met verdict whose `**Opened**` line
|
|
103
|
+
names none of the files the finding cites is collected as UNVERIFIABLE: a verdict
|
|
104
|
+
on evidence you did not open is not a vote."""
|
|
98
105
|
|
|
99
106
|
def _nonempty_string(value: Any) -> str:
|
|
100
107
|
return value if isinstance(value, str) and value.strip() else ""
|
|
@@ -222,7 +229,7 @@ def reverify_prompt_body(
|
|
|
222
229
|
"collector also reads `**Verdict:** …` and `- Verdict: …` as the same field.\n\n"
|
|
223
230
|
)
|
|
224
231
|
rows.append(ADVERSARIAL_RESPONSE.replace("<finding-id>", findings[0].finding_id))
|
|
225
|
-
rows.append("\n")
|
|
232
|
+
rows.append("\n\n" + OPENED_RULE + "\n")
|
|
226
233
|
if len(findings) > 1:
|
|
227
234
|
rows.append(f"\n### {findings[1].finding_id}\n**Verdict**: ...\n")
|
|
228
235
|
return "".join(rows)
|