generated from coulomb/repo-seed
Add acceptance boundary regression coverage
This commit is contained in:
136
tests/test_acceptance_boundary.py
Normal file
136
tests/test_acceptance_boundary.py
Normal file
@@ -0,0 +1,136 @@
|
||||
from repo_registry.acceptance import AgenticReviewDecision
|
||||
from repo_registry.core.service import RegistryService
|
||||
from repo_registry.repo_ingestion.git import GitIngestionService
|
||||
from repo_registry.storage.sqlite import RegistryStore
|
||||
|
||||
|
||||
class BoundaryApprovingReviewer:
|
||||
reviewer_id = "boundary-agent"
|
||||
policy_version = "agentic-review-policy/test"
|
||||
|
||||
def review(self, request):
|
||||
return [
|
||||
AgenticReviewDecision(
|
||||
action="approve",
|
||||
target_type="candidate_graph",
|
||||
target_id=request.candidate_graph.analysis_run.id,
|
||||
rationale="README and source refs support the generated API capability.",
|
||||
criterion_ids=["RREG-QC-004"],
|
||||
evidence_refs=["README.md", "app.py"],
|
||||
)
|
||||
]
|
||||
|
||||
|
||||
def make_service(tmp_path, *, reviewer=None):
|
||||
store = RegistryStore(tmp_path / "registry.sqlite3")
|
||||
store.initialize()
|
||||
return RegistryService(
|
||||
store,
|
||||
ingestion=GitIngestionService(tmp_path / "checkouts"),
|
||||
agentic_reviewer=reviewer,
|
||||
)
|
||||
|
||||
|
||||
def write_api_repo(tmp_path):
|
||||
source = tmp_path / "api-repo"
|
||||
source.mkdir()
|
||||
(source / "README.md").write_text("# API Repo\nReports health.\n", encoding="utf-8")
|
||||
(source / "app.py").write_text('@app.get("/health")\ndef health():\n return {}\n', encoding="utf-8")
|
||||
return source
|
||||
|
||||
|
||||
def write_provider_repo(tmp_path):
|
||||
source = tmp_path / "provider-repo"
|
||||
source.mkdir()
|
||||
(source / "README.md").write_text("# Provider Repo\n", encoding="utf-8")
|
||||
(source / "providers.py").write_text(
|
||||
"provider_registry = {'openrouter': OpenRouterAdapter}\n",
|
||||
encoding="utf-8",
|
||||
)
|
||||
return source
|
||||
|
||||
|
||||
def test_deterministic_analysis_leaves_candidates_pending(tmp_path):
|
||||
service = make_service(tmp_path)
|
||||
repository = service.register_repository(
|
||||
name="API Repo",
|
||||
url=str(write_api_repo(tmp_path)),
|
||||
)
|
||||
|
||||
summary = service.analyze_repository(repository.id, use_llm_assistance=False)
|
||||
|
||||
graph = service.candidate_graph(repository.id, summary.analysis_run.id)
|
||||
assert service.ability_map(repository.id).abilities == []
|
||||
assert {
|
||||
capability.status
|
||||
for ability in graph.abilities
|
||||
for capability in ability.capabilities
|
||||
} == {"candidate"}
|
||||
|
||||
|
||||
def test_deterministic_gates_flag_provider_regression_without_approval(tmp_path):
|
||||
service = make_service(tmp_path)
|
||||
repository = service.register_repository(
|
||||
name="Provider Repo",
|
||||
url=str(write_provider_repo(tmp_path)),
|
||||
)
|
||||
|
||||
summary = service.analyze_repository(repository.id, use_llm_assistance=False)
|
||||
|
||||
graph = service.candidate_graph(repository.id, summary.analysis_run.id)
|
||||
decisions = service.list_review_decisions(repository.id, summary.analysis_run.id)
|
||||
assert service.ability_map(repository.id).abilities == []
|
||||
assert graph.abilities[0].capabilities[0].status == "candidate"
|
||||
assert any(
|
||||
decision.action == "quality_gate_evaluation"
|
||||
and "RREG-QC-002" in decision.criterion_ids
|
||||
for decision in decisions
|
||||
)
|
||||
|
||||
|
||||
def test_agentic_review_is_only_automated_approval_path(tmp_path):
|
||||
service = make_service(tmp_path, reviewer=BoundaryApprovingReviewer())
|
||||
repository = service.register_repository(
|
||||
name="Agent Approved Repo",
|
||||
url=str(write_api_repo(tmp_path)),
|
||||
)
|
||||
|
||||
summary = service.analyze_repository(
|
||||
repository.id,
|
||||
use_llm_assistance=False,
|
||||
agentic_review=True,
|
||||
)
|
||||
|
||||
decisions = service.list_review_decisions(repository.id, summary.analysis_run.id)
|
||||
assert service.ability_map(repository.id).abilities
|
||||
assert any(
|
||||
decision.action == "agentic_approve_candidate_graph"
|
||||
and decision.reviewer_type == "agent"
|
||||
and decision.rationale
|
||||
and decision.criteria_version == "repo-scoping-quality-criteria/v1"
|
||||
and decision.evidence_refs == ["README.md", "app.py"]
|
||||
for decision in decisions
|
||||
)
|
||||
|
||||
|
||||
def test_manual_approval_path_still_works(tmp_path):
|
||||
service = make_service(tmp_path)
|
||||
repository = service.register_repository(
|
||||
name="Manual Review Repo",
|
||||
url=str(write_api_repo(tmp_path)),
|
||||
)
|
||||
summary = service.analyze_repository(repository.id, use_llm_assistance=False)
|
||||
|
||||
service.approve_candidate_graph(
|
||||
repository.id,
|
||||
summary.analysis_run.id,
|
||||
notes="Manual curator approval.",
|
||||
)
|
||||
|
||||
decisions = service.list_review_decisions(repository.id, summary.analysis_run.id)
|
||||
assert service.ability_map(repository.id).abilities
|
||||
assert any(
|
||||
decision.action == "approve_candidate_graph"
|
||||
and decision.reviewer_type == "human"
|
||||
for decision in decisions
|
||||
)
|
||||
@@ -230,7 +230,7 @@ versioned in `docs/quality-criteria` and linked through workplan notes.
|
||||
|
||||
```task
|
||||
id: RREG-WP-0014-T07
|
||||
status: todo
|
||||
status: done
|
||||
priority: high
|
||||
state_hub_task_id: "37a22c89-ded5-42dd-aaa9-ece79477fcff"
|
||||
```
|
||||
@@ -246,6 +246,14 @@ Acceptance criteria:
|
||||
deterministic rules.
|
||||
- Existing manual review and approval paths keep working.
|
||||
|
||||
Implementation note 2026-05-15: added
|
||||
`tests/test_acceptance_boundary.py` as an end-to-end regression boundary for
|
||||
analysis and review. The suite proves deterministic analysis leaves generated
|
||||
characteristics pending, deterministic quality gates can flag the provider
|
||||
routing self-scan regression without approval, configured agentic review is the
|
||||
only automated approval path and must leave rationale/criteria/evidence audit
|
||||
metadata, and manual approval still produces human review decisions.
|
||||
|
||||
## T08: Migration And Compatibility Plan
|
||||
|
||||
```task
|
||||
|
||||
Reference in New Issue
Block a user