143 lines
5.4 KiB
JSON
143 lines
5.4 KiB
JSON
{
|
|
"task_id": "DEV-000071",
|
|
"task_revision": 1,
|
|
"architecture_revision": "DEV-000070-baseline",
|
|
"base_repo_revision": "3aa981bc596575926549f6133a1dad1a863e913f",
|
|
"risk_tier": "E",
|
|
"intent": "Audit the actual production cognitive path, router reality, deterministic boundary, and generalization limits of current Domovoy; extend the reality evaluation harness with a broader natural-language and cognitive-quality corpus without patching product behavior to improve scores.",
|
|
"requirement_ids": [
|
|
"DEV-000070 baseline complete",
|
|
"Production evidence before behavior change",
|
|
"Provider-neutral router reality audit",
|
|
"Deterministic boundary audit",
|
|
"Generalization and holdout baseline",
|
|
"Free-form cognitive behavior measurement",
|
|
"Real production failure evidence inclusion",
|
|
"No behavior patching to improve evaluation scores"
|
|
],
|
|
"allowed_change_scope": {
|
|
"modules": [
|
|
"interaction",
|
|
"situation",
|
|
"work",
|
|
"capability",
|
|
"continuity",
|
|
"epistemics",
|
|
"artifact"
|
|
],
|
|
"paths": [
|
|
"tests/**",
|
|
"cmd/cognitive-benchmark/**",
|
|
"docs/PROJECT_STATUS.md",
|
|
"docs/evidence/**",
|
|
"dev/ACTIVE_TASK.json",
|
|
"dev/tasks/DEV-000071-cognitive-path-reality-generalization.json"
|
|
]
|
|
},
|
|
"forbidden_change_scope": {
|
|
"paths": [
|
|
"architecture/ARCHITECTURE_CONSTITUTION.md",
|
|
"architecture/architecture-registry.yaml",
|
|
"architecture/architecture-registry.json",
|
|
"architecture/adrs/**",
|
|
"contracts/**",
|
|
"modules/**",
|
|
"adapters/**",
|
|
"cmd/core-runtime/**",
|
|
"cmd/creator-web/**",
|
|
"cmd/cognitive-worker/**",
|
|
"cmd/continuity-runtime/**",
|
|
"cmd/commit-runtime/**",
|
|
"modules/authority/**",
|
|
"modules/commit/**",
|
|
"security/**",
|
|
"ci/**",
|
|
".gitea/**",
|
|
"deploy/**",
|
|
"database/migrations/**"
|
|
]
|
|
},
|
|
"allowed_architecture_delta": {
|
|
"new_cross_module_dependencies": 0,
|
|
"new_canonical_owners": 0,
|
|
"new_privileges": 0,
|
|
"new_third_party_dependencies": 0,
|
|
"new_provider_adapters": 0,
|
|
"new_database_owner_roles": 0,
|
|
"new_durable_canonical_types": 0
|
|
},
|
|
"contracts": {
|
|
"consume": [
|
|
"Existing Interaction HumanUtterance / OriginalHumanIntent / UtteranceInterpretation / ProjectReference evidence",
|
|
"Existing Capability cognitive registry / usage observation / implementation evidence",
|
|
"Existing Work Goal / Delegation evidence",
|
|
"Existing Epistemics ExperienceRecord evidence",
|
|
"Existing Creator Web / Cognitive Worker route behavior as audited runtime behavior"
|
|
],
|
|
"may_change": []
|
|
},
|
|
"semantic_owner": "none-evaluation-audit-only",
|
|
"canonical_state": [
|
|
"No new canonical state",
|
|
"Existing production evidence only",
|
|
"Evaluation artifacts under docs/evidence"
|
|
],
|
|
"state_transitions": [
|
|
"production evidence -> audit report",
|
|
"existing suite_v1 preserved unchanged",
|
|
"new generalization corpus -> baseline run against current unchanged Domovoy",
|
|
"operator-readable evidence report updated without changing product runtime behavior"
|
|
],
|
|
"side_effects_and_privileges": [
|
|
"Read-only production audit over existing evidence and health endpoints",
|
|
"No product behavior patches to improve scores",
|
|
"No forced Ollama/GPU invocation",
|
|
"No activation change for bindings",
|
|
"No Docker-test usage"
|
|
],
|
|
"persistence_and_migrations": [
|
|
"No migrations",
|
|
"No new canonical persistence",
|
|
"No cross-owner SQL join dependency in product code",
|
|
"Evaluation remains tooling/artifact evidence"
|
|
],
|
|
"failure_retry_restart": [
|
|
"Poor scores are accepted evidence and must be reported honestly",
|
|
"UNKNOWN must not be converted into zero when observability is insufficient",
|
|
"Only evaluation/observability defects needed for truthful measurement may be fixed"
|
|
],
|
|
"assumptions": [
|
|
"DEV-000070 merged baseline is current main",
|
|
"Existing cognitive usage observations cover gateway invocations",
|
|
"Production audit can rely on existing database evidence and runtime health without new instrumentation"
|
|
],
|
|
"acceptance_scenarios": [
|
|
"real three-turn production failure sequence is audited from actual production evidence where available",
|
|
"actual registered/discovered/eligible cognitive implementations are reported honestly",
|
|
"answer is provided whether /chat has ever invoked Ollama/GPU",
|
|
"deterministic boundary is documented from current runtime behavior",
|
|
"suite_v1 is preserved unchanged",
|
|
"approximately fifty additional generalization/cognitive cases are added",
|
|
"free-form cognition quality, uncertainty, anti-sycophancy and consequence generalization are measured",
|
|
"baseline is run against current unchanged Domovoy"
|
|
],
|
|
"required_tests": [
|
|
"go test ./...",
|
|
"go vet ./...",
|
|
"go run ./cmd/devctl preflight -active dev/ACTIVE_TASK.json",
|
|
"go run ./cmd/devctl check-architecture",
|
|
"go run ./cmd/devctl check-diff -active dev/ACTIVE_TASK.json",
|
|
"tests/cognitive_advisory/cognitive_worker_env_policy_test.ps1",
|
|
"tests/owner_acl/verify.ps1 -DockerContext docker-prod -PostgresContainer assistant-postgres-1",
|
|
"evaluation harness unit/replay tests",
|
|
"baseline suite_v1 plus generalization suite execution",
|
|
"production health"
|
|
],
|
|
"terminal_gap_states": [
|
|
"BLOCKED_BY_ARCHITECTURE_GAP",
|
|
"BLOCKED_BY_SPEC_GAP",
|
|
"BLOCKED_BY_MISSING_CAPABILITY",
|
|
"BLOCKED_BY_CONFLICT"
|
|
]
|
|
}
|