core-runtime 0.1.1__py3-none-any.whl

shared/infrastructure/intent/operational_config.py

@@ -0,0 +1,1343 @@
+# src/shared/infrastructure/intent/operational_config.py
+"""
+Loader for .intent/enforcement/config/operational_config.yaml (ADR-040).
+
+Every runtime-behavior knob that previously lived as a literal in src/ is
+declared in operational_config.yaml; this module reads the file via
+IntentRepository and returns a tree of frozen dataclasses callers can
+consume. On any failure — missing file, parse error, malformed section —
+the loader logs a warning and falls back to defaults that match the
+shipped YAML values. It never raises, so the daemon can boot and the
+remediation loop can keep running while a governor reviews the failure.
+
+Pattern mirrors src/will/workers/circuit_breaker.py: one frozen dataclass
+per top-level YAML section, plus a single public load_*_config function.
+The workers section is nested — each sub-key becomes its own Worker*Config
+collected into WorkersConfig.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from typing import Any
+
+from shared.logger import getLogger
+
+
+logger = getLogger(__name__)
+
+
+# ---------------------------------------------------------------------------
+# Coercion helpers
+# ---------------------------------------------------------------------------
+
+
+def _section(raw: dict[str, Any] | None, name: str) -> dict[str, Any]:
+    if not raw:
+        return {}
+    sub = raw.get(name)
+    if sub is None:
+        return {}
+    if not isinstance(sub, dict):
+        logger.warning(
+            "operational_config: section %r is not a mapping — using fallback defaults.",
+            name,
+        )
+        return {}
+    return sub
+
+
+def _get_int(sec: dict[str, Any], key: str, default: int) -> int:
+    val = sec.get(key)
+    if val is None:
+        return default
+    try:
+        return int(val)
+    except (TypeError, ValueError):
+        logger.warning(
+            "operational_config: %s should be int, got %r — using fallback %d.",
+            key,
+            val,
+            default,
+        )
+        return default
+
+
+def _get_float(sec: dict[str, Any], key: str, default: float) -> float:
+    val = sec.get(key)
+    if val is None:
+        return default
+    try:
+        return float(val)
+    except (TypeError, ValueError):
+        logger.warning(
+            "operational_config: %s should be float, got %r — using fallback %r.",
+            key,
+            val,
+            default,
+        )
+        return default
+
+
+# ID: 4d2a8e4c-1a8b-4b6e-9c2a-3f0e7a8b5d11
+def _get_bool(sec: dict[str, Any], key: str, default: bool) -> bool:
+    val = sec.get(key)
+    if val is None:
+        return default
+    if isinstance(val, bool):
+        return val
+    logger.warning(
+        "operational_config: %s should be bool, got %r — using fallback %r.",
+        key,
+        val,
+        default,
+    )
+    return default
+
+
+# ID: 9b4f3e2c-7a8d-4e1f-b3c5-2d6e9f0a1b8c
+def _get_str_tuple(
+    sec: dict[str, Any], key: str, default: tuple[str, ...]
+) -> tuple[str, ...]:
+    val = sec.get(key)
+    if val is None:
+        return default
+    if isinstance(val, list) and all(isinstance(x, str) for x in val):
+        return tuple(val)
+    logger.warning(
+        "operational_config: %s should be a list of strings, got %r — using fallback %r.",
+        key,
+        val,
+        default,
+    )
+    return default
+
+
+# ---------------------------------------------------------------------------
+# Section dataclasses
+# ---------------------------------------------------------------------------
+
+
+@dataclass(frozen=True)
+# ID: d077c23b-94b9-48eb-966c-92ade3d25d5c
+class LLMConfig:
+    default_max_tokens: int = 4096
+    default_max_length: int = 4096
+    http_timeout_sec: int = 60
+    request_timeout_sec: int = 300
+    provider_timeout_sec: int = 180
+
+
+@dataclass(frozen=True)
+# ID: 570aadb4-d174-4abc-bb31-dbb578c80401
+class EmbeddingConfig:
+    chunk_size: int = 512
+    chunk_overlap: int = 50
+    max_chars: int = 20000
+    provider_request_timeout_sec: float = 120.0
+    provider_connect_timeout_sec: float = 10.0
+    utils_request_timeout_sec: float = 30.0
+
+
+@dataclass(frozen=True)
+# ID: 4b29ce9f-d734-44ff-9bc3-f9b956878f4a
+class ChunkingConfig:
+    max_chunk_chars: int = 1500
+
+
+@dataclass(frozen=True)
+# ID: dee98ce9-d8e4-4ad7-bc41-f0c9a2291389
+class BlackboardConfig:
+    """Blackboard hygiene + ADR-082 writer-as-sensor retention knobs.
+
+    sla_default_seconds — fallback SLA tier when entry_type is unmapped.
+
+    ADR-082 retention policy — two sweep mechanisms hosted in
+    BlackboardShopManager:
+
+    - Mechanism 1 (terminal telemetry): hard DELETE rows whose subject
+      starts with one of telemetry_subject_prefixes, status is already
+      terminal, and created_at is older than telemetry_ttl_days.
+    - Mechanism 2 (DELEGATE OPEN findings): transition status open →
+      resolved for rows whose subject is in delegate_finding_subjects
+      and created_at is older than delegate_finding_ttl_days. The
+      writer's run.complete report preserves the event payload, so
+      auto-resolution is information-preserving.
+
+    sweep_batch_max — row cap per sweep run (rails per
+    feedback_destructive_autonomous_needs_rails_first).
+    """
+
+    sla_default_seconds: int = 3600
+    telemetry_ttl_days: int = 7
+    telemetry_subject_prefixes: tuple[str, ...] = ("loop_hold.sample::",)
+    delegate_finding_ttl_days: int = 7
+    delegate_finding_subjects: tuple[str, ...] = (
+        "coherence.violation_executor.blast_bound",
+        "coherence.repo_artifacts.drift",
+    )
+    sweep_batch_max: int = 500
+
+
+@dataclass(frozen=True)
+# ID: eeee047a-d8a0-46d7-858a-1c6bb47079b5
+class HealthLogConfig:
+    stale_threshold_seconds: int = 3600
+
+
+@dataclass(frozen=True)
+# ID: d6b4d3bb-cb67-4a62-90a7-1a6ca68e50e0
+class DaemonConfig:
+    one_shot_interval_sec: int = 300
+    # ADR-081 Step 0 — loop-hold instrumentation. Defaults are the OFF state
+    # so a YAML revert returns to clean baseline. Step 3a-telemetry subscribes
+    # a structured handler to the slow-callback warnings these settings emit.
+    slow_callback_duration_sec: float = 0.1
+    set_debug: bool = False
+
+
+@dataclass(frozen=True)
+# ID: 7e8f9a0b-1c2d-3e4f-5a6b-7c8d9e0f1a2b
+class WorkerClassificationConfig:
+    """ADR-081 D7 — gates for the runtime.worker_process_classification rule.
+
+    The rule consumes loop_hold.sample blackboard entries (Step 3a-telemetry)
+    and fires advisory findings when observed loop-hold contradicts the
+    declared requires_dedicated_process state on a worker. The three gates
+    here are the thresholds D7 specifies; values are tunable via
+    .intent/enforcement/config/operational_config.yaml without code change.
+    """
+
+    loop_hold_escalation_sec: float = 5.0
+    loop_hold_deescalation_sec: float = 1.0
+    cycle_window: int = 5
+
+
+@dataclass(frozen=True)
+# ID: 546c1ad9-c7b5-43a7-af49-d2083f7fe865
+class ProposalsConfig:
+    list_limit: int = 100
+    pending_limit: int = 50
+
+
+@dataclass(frozen=True)
+# ID: 367521a6-1521-4fb1-89f1-f1ac6c01a1a1
+class ConsequenceLogConfig:
+    default_lookback_seconds: int = 3600
+
+
+@dataclass(frozen=True)
+# ID: 8b3c6f4e-2a91-43d7-b582-7e1d4a9c0f6b
+class AuditConfig:
+    """ADR-044 — incremental llm_gate verdict cache knobs.
+
+    - llm_gate_verdict_cache_ttl_days: rows older than this are swept at
+      the start of each audit run. TTL is hygiene, not correctness —
+      content + rule hash mismatches are the correctness mechanism.
+    - llm_gate_cache_staleness_threshold_seconds: when the crawler's
+      repo_artifacts.last_crawled_at is older than this, the engine
+      recomputes file_content_hash inline rather than trusting the
+      stored value. Bounds the window in which a stale crawler hash
+      could produce an incorrect cache hit.
+    """
+
+    llm_gate_verdict_cache_ttl_days: int = 30
+    llm_gate_cache_staleness_threshold_seconds: int = 3600
+
+
+@dataclass(frozen=True)
+# ID: 3101bd97-4957-40cc-85f5-96fa0e874cd4
+class CoverageConfig:
+    gap_threshold_pct: float = 75.0
+    warn_pct: int = 75
+    low_bucket_pct: int = 50
+    gap_list_limit: int = 10
+    collect_timeout_sec: int = 120
+    full_run_timeout_sec: int = 300
+    quality_min_pct: int = 75
+    watcher_required_pct: float = 75.0
+    watcher_rescan_hours: int = 24
+    batch_remediation_threshold_pct: float = 75.0
+    batch_remediation_partial_success_pct: int = 50
+    single_file_target_pct: float = 75.0
+
+
+@dataclass(frozen=True)
+# ID: c1625d9c-2e4a-44a4-a009-55e2a75e07fe
+class WorkflowGateConfig:
+    linter_timeout_sec: float = 30.0
+    quality_timeout_sec: float = 60.0
+    ruff_format_timeout_sec: float = 60.0
+    import_timeout_sec: float = 60.0
+
+
+@dataclass(frozen=True)
+# ID: e37c8ce2-6a3a-45d7-a63e-f9b3ec810b14
+class ContextConfig:
+    score_target_file: int = 100
+    score_target_path: int = 80
+    score_target_symbol: int = 120
+    score_has_content: int = 30
+    vector_top_k: int = 10
+    db_provider_max_items: int = 100
+    token_estimate_overhead: int = 300
+    cache_ttl_hours: int = 24
+    db_recent_packets_limit: int = 10
+
+
+@dataclass(frozen=True)
+# ID: a77084a2-9655-4884-a17a-95d0d3f16d91
+class HealthConfig:
+    worker_alive_threshold_sec: int = 600
+    worker_warn_threshold_sec: int = 3600
+    long_lookback_hours: int = 24
+    medium_lookback_minutes: int = 60
+    short_lookback_minutes: int = 30
+    recent_lookback_minutes: int = 10
+
+
+@dataclass(frozen=True)
+# ID: a199bb79-5c14-408c-a116-256f72347d9f
+class TestingConfig:
+    pytest_collection_timeout_sec: int = 30
+    pytest_execution_timeout_sec: int = 300
+    sandbox_timeout_sec: int = 30
+    executor_timeout_sec: int = 60
+    metrics_timeout_sec: int = 30
+    simple_gen_timeout_sec: float = 20.0
+    context_aware_gen_timeout_sec: float = 15.0
+    runtime_validator_timeout_sec: int = 60
+    max_failures: int = 10
+    snippet_max_lines: int = 20
+
+
+@dataclass(frozen=True)
+# ID: 399b9dd7-d2cf-4be5-9721-b8e020e51250
+class StrategySelectorConfig:
+    min_recommended_score: int = 15
+    score_role_preferred: int = 25
+    score_role_discouraged: int = 25
+    score_generalized_match: int = 16
+    score_rule_preferred: int = 12
+    score_rule_discouraged: int = 12
+    score_size_bonus: int = 10
+    score_size_penalty: int = 10
+    score_strong_split: int = 14
+    score_cluster_count_bonus: int = 8
+    score_cluster_count_penalty: int = 8
+    score_constraint_role_bonus: int = 9
+    score_constraint_role_penalty: int = 8
+    score_conservatism_bias_bonus: int = 8
+    score_conservatism_structural_bonus: int = 5
+    large_file_lines: int = 400
+    small_file_lines: int = 200
+    strong_split_lines: int = 450
+
+
+@dataclass(frozen=True)
+# ID: cea759ac-b5e2-4204-adb3-c0c5c6c1bab1
+class AnalyzersConfig:
+    file_complexity_high_threshold: int = 15
+    class_methods_high_threshold: int = 10
+    function_body_high_threshold: int = 25
+    function_body_low_threshold: int = 10
+    max_file_lines: int = 400
+    max_function_lines: int = 50
+    max_module_lines: int = 400
+
+
+@dataclass(frozen=True)
+# ID: 046dc166-28df-4900-9c59-0dcf5eece8df
+class RefactorConfig:
+    responsibilities_threshold: int = 20
+    cohesion_threshold: int = 12
+    coupling_threshold: int = 10
+    loc_threshold: int = 400
+
+
+@dataclass(frozen=True)
+# ID: d2c55fd2-9508-4863-a1d1-ed5c171914e5
+class ClarityConfig:
+    structural_complexity: int = 20
+    structural_lines: int = 300
+    logic_simplification_threshold: int = 10
+
+
+@dataclass(frozen=True)
+# ID: ca897e8c-2d2a-4384-9c0e-e50a7a150e63
+class ComplexityConfig:
+    god_method_threshold: int = 30
+    extraction_threshold: int = 15
+
+
+@dataclass(frozen=True)
+# ID: 9f5cb6d0-a2e8-4e7e-850f-5d52cacad63a
+class ParsingConfig:
+    min_block_len: int = 10
+    score_test_fn: int = 1000
+    score_test_class: int = 1000
+    score_import: int = 100
+    score_pytest: int = 500
+
+
+@dataclass(frozen=True)
+# ID: 531bbd4c-d181-44d7-a951-2d1ed6f3f092
+class ExecutionConfig:
+    task_timeout_sec: int = 300
+    workflow_timeout_minutes: int = 30
+    orchestrator_max_steps: int = 10
+
+
+@dataclass(frozen=True)
+# ID: 4f332080-09c6-4ee6-b93b-41ab22095174
+class ActionConfig:
+    max_data_size_bytes: int = 5_242_880
+
+
+@dataclass(frozen=True)
+# ID: 4ed8b90f-159e-4147-af1d-d5a01cbc4ef7
+class PromptPipelineConfig:
+    max_file_size_bytes: int = 1_048_576
+
+
+@dataclass(frozen=True)
+# ID: d586577a-437e-4c26-8dd1-ba5283b8a6f0
+class ValidatorConfig:
+    lru_cache_size: int = 1024
+    lru_cache_size_small: int = 512
+
+
+@dataclass(frozen=True)
+# ID: 15bf16dc-c670-4742-a338-39a134a8d744
+class MemoryConfig:
+    recency_days: int = 30
+    episode_retention_days: int = 30
+    reflection_retention_days: int = 90
+
+
+@dataclass(frozen=True)
+# ID: 972301d5-6370-47c1-a9b2-1668aa89b332
+class RepositoriesConfig:
+    decision_trace_default_limit: int = 10
+    decision_trace_max_limit: int = 100
+    decision_trace_retention_days: int = 30
+    refusal_default_limit: int = 20
+    refusal_by_type_limit: int = 50
+    symbol_definition_default_limit: int = 500
+    project_def_max_limit: int = 500
+    project_def_default_limit: int = 100
+
+
+@dataclass(frozen=True)
+# ID: c21addc5-f8d0-41b5-a716-8984e28da0d8
+class GitConfig:
+    recent_commits_n: int = 10
+    changed_files_log_n: int = 20
+
+
+@dataclass(frozen=True)
+# ID: 51388277-8f97-4655-b5a2-a704716ef4fe
+class VectorsConfig:
+    index_batch_size: int = 10
+    scan_limit: int = 10000
+    report_preview_count: int = 10
+    policy_vectorizer_batch_size: int = 10
+
+
+@dataclass(frozen=True)
+# ID: 2d33c98b-1ed4-499b-a821-4764ed72a4ec
+class SyncConfig:
+    artifact_embed_batch_size: int = 10
+    policy_index_batch_size: int = 10
+    pattern_index_batch_size: int = 10
+    specs_index_batch_size: int = 10
+
+
+@dataclass(frozen=True)
+# ID: 87095648-7a0f-4b32-b430-eb54cba9aa39
+class AuthorityPackageConfig:
+    search_limit: int = 10
+
+
+@dataclass(frozen=True)
+# ID: eb02fecc-3b0e-4b78-8aef-48eac9572b85
+class StrategicAuditorConfig:
+    sample_limit: int = 100
+    commit_lookback: int = 15
+    compact_max_chars: int = 1500
+
+
+@dataclass(frozen=True)
+# ID: 1db478c8-329c-401f-864c-c7a3982f22f4
+class MiscConfig:
+    linelength_max_line_chars: int = 100
+    enrichment_description_max_chars: int = 500
+    conversation_max_content_chars: int = 2000
+    code_snippet_context_lines: int = 20
+    embedding_search_default_limit: int = 10
+    retriever_search_limit: int = 10
+    retriever_context_lines: int = 20
+    file_navigator_read_max_lines: int = 200
+    file_navigator_max_read_bytes: int = 1_048_576
+    knowledge_consolidation_max_lines: int = 10
+    context_export_http_timeout_sec: int = 10
+    metadata_max_comment_length: int = 120
+    census_hotspot_limit: int = 10
+    clustering_default_n_clusters: int = 15
+    legacy_scan_display_limit: int = 10
+    limb_status_recent_limit: int = 15
+    refusal_inspect_default_limit: int = 20
+    refusal_inspect_by_type_limit: int = 20
+    capability_tagging_default_llm_confidence: float = 0.70
+    knowledge_min_occurrences: int = 3
+    knowledge_max_lines: int = 10
+    context_aware_test_context_lines: int = 40
+    perf_overhead_warning_pct: int = 50
+    perf_overhead_error_pct: int = 100
+    context_search_display_limit: int = 20
+    proposals_display_limit: int = 20
+
+
+# ---------------------------------------------------------------------------
+# Workers section — one dataclass per worker sub-key
+# ---------------------------------------------------------------------------
+
+
+@dataclass(frozen=True)
+# ID: 2b9a3a7b-f6c2-404b-8b3d-05ecf70b618d
+class WorkerCallSiteRewriterConfig:
+    claim_limit: int = 50
+
+
+@dataclass(frozen=True)
+# ID: cd230702-b80c-41a2-8d52-a0297aa989c0
+class WorkerDocWriterConfig:
+    batch_size: int = 25
+
+
+@dataclass(frozen=True)
+# ID: d627b73c-8a7c-4260-adff-c7a2d274d12e
+class WorkerDocWorkerConfig:
+    batch_size: int = 50
+
+
+@dataclass(frozen=True)
+# ID: 7cd69bc3-41f2-48b7-b330-5d170cf86860
+class WorkerPromptArtifactWriterConfig:
+    claim_limit: int = 25
+
+
+@dataclass(frozen=True)
+# ID: 123297dc-8907-4baf-9602-f33f79002b48
+class WorkerPromptExtractorConfig:
+    claim_limit: int = 25
+    context_lines: int = 40
+
+
+@dataclass(frozen=True)
+# ID: 593b96fb-329b-46e4-b490-3029df029e47
+class WorkerCapabilityTaggerConfig:
+    batch_size: int = 20
+
+
+@dataclass(frozen=True)
+# ID: 76350f2c-892a-41ef-8562-1bee90a5e69b
+class WorkerProposalConsumerConfig:
+    claim_limit: int = 5
+
+
+@dataclass(frozen=True)
+# ID: 47236015-9d93-4f93-af91-1334353de07f
+class WorkerViolationExecutorConfig:
+    claim_limit: int = 50
+
+
+@dataclass(frozen=True)
+# ID: accbcdf4-23be-44bd-b0c0-4212dfe9fd4e
+class WorkerViolationRemediatorConfig:
+    claim_limit: int = 50
+    scan_limit: int = 200
+    ceremony_timeout_sec: int = 30
+    semantic_examples_limit: int = 3
+    min_role_confidence: float = 0.55
+
+
+@dataclass(frozen=True)
+# ID: 4f4f42b4-c4eb-4d05-b97a-5dcd97dbca5c
+class WorkerShopConfig:
+    glide_off_multiplier: float = 0.10
+    fallback_threshold_sec: int = 600
+    findings_scan_limit: int = 200
+
+
+@dataclass(frozen=True)
+# ID: db4c3515-9de4-44ee-9bb5-283ef53b87b6
+class WorkerProposalPipelineShopConfig:
+    stuck_approved_sla_sec: int = 1800
+    stuck_executing_sla_sec: int = 900
+    repeated_failure_threshold: int = 3
+    repeated_failure_lookback_sec: int = 86400
+    findings_scan_limit: int = 200
+
+
+@dataclass(frozen=True)
+# ID: 9447633e-e6c3-49e5-8a2b-34570dfb0d74
+class WorkerTestRemediatorConfig:
+    scan_limit: int = 200
+
+
+@dataclass(frozen=True)
+# ID: 87dc2e43-4fa4-4f06-9729-5863e4e970bf
+class WorkerTestRunnerSensorConfig:
+    scan_limit: int = 50
+
+
+@dataclass(frozen=True)
+# ID: 382dfb51-5dde-4934-8ede-f89042f2c3df
+class WorkerIntentInspectorConfig:
+    alignment_batch: int = 20
+
+
+@dataclass(frozen=True)
+# ID: b4073ed0-e290-4ce0-95d3-14f11647e428
+class WorkerObserverConfig:
+    stale_threshold_seconds: int = 3600
+
+
+@dataclass(frozen=True)
+# ID: 0e3397a8-6a9a-49c6-a253-0829101e8a4e
+class WorkerCoherenceSensorConfig:
+    lookback_seconds: int = 7200
+
+
+@dataclass(frozen=True)
+# ID: e1b5e532-edb4-405f-b09f-57339041baf9
+class WorkersConfig:
+    call_site_rewriter: WorkerCallSiteRewriterConfig = field(
+        default_factory=WorkerCallSiteRewriterConfig
+    )
+    doc_writer: WorkerDocWriterConfig = field(default_factory=WorkerDocWriterConfig)
+    doc_worker: WorkerDocWorkerConfig = field(default_factory=WorkerDocWorkerConfig)
+    prompt_artifact_writer: WorkerPromptArtifactWriterConfig = field(
+        default_factory=WorkerPromptArtifactWriterConfig
+    )
+    prompt_extractor: WorkerPromptExtractorConfig = field(
+        default_factory=WorkerPromptExtractorConfig
+    )
+    capability_tagger: WorkerCapabilityTaggerConfig = field(
+        default_factory=WorkerCapabilityTaggerConfig
+    )
+    proposal_consumer: WorkerProposalConsumerConfig = field(
+        default_factory=WorkerProposalConsumerConfig
+    )
+    violation_executor: WorkerViolationExecutorConfig = field(
+        default_factory=WorkerViolationExecutorConfig
+    )
+    violation_remediator: WorkerViolationRemediatorConfig = field(
+        default_factory=WorkerViolationRemediatorConfig
+    )
+    worker_shop: WorkerShopConfig = field(default_factory=WorkerShopConfig)
+    proposal_pipeline_shop: WorkerProposalPipelineShopConfig = field(
+        default_factory=WorkerProposalPipelineShopConfig
+    )
+    test_remediator: WorkerTestRemediatorConfig = field(
+        default_factory=WorkerTestRemediatorConfig
+    )
+    test_runner_sensor: WorkerTestRunnerSensorConfig = field(
+        default_factory=WorkerTestRunnerSensorConfig
+    )
+    intent_inspector: WorkerIntentInspectorConfig = field(
+        default_factory=WorkerIntentInspectorConfig
+    )
+    observer: WorkerObserverConfig = field(default_factory=WorkerObserverConfig)
+    coherence_sensor: WorkerCoherenceSensorConfig = field(
+        default_factory=WorkerCoherenceSensorConfig
+    )
+
+
+# ---------------------------------------------------------------------------
+# Top-level config
+# ---------------------------------------------------------------------------
+
+
+@dataclass(frozen=True)
+# ID: 973dd73d-45af-4909-a275-c4fd44af56fb
+class OperationalConfig:
+    llm: LLMConfig = field(default_factory=LLMConfig)
+    embedding: EmbeddingConfig = field(default_factory=EmbeddingConfig)
+    chunking: ChunkingConfig = field(default_factory=ChunkingConfig)
+    workers: WorkersConfig = field(default_factory=WorkersConfig)
+    blackboard: BlackboardConfig = field(default_factory=BlackboardConfig)
+    health_log: HealthLogConfig = field(default_factory=HealthLogConfig)
+    daemon: DaemonConfig = field(default_factory=DaemonConfig)
+    worker_classification: WorkerClassificationConfig = field(
+        default_factory=WorkerClassificationConfig
+    )
+    proposals: ProposalsConfig = field(default_factory=ProposalsConfig)
+    consequence_log: ConsequenceLogConfig = field(default_factory=ConsequenceLogConfig)
+    audit: AuditConfig = field(default_factory=AuditConfig)
+    coverage: CoverageConfig = field(default_factory=CoverageConfig)
+    workflow_gate: WorkflowGateConfig = field(default_factory=WorkflowGateConfig)
+    context: ContextConfig = field(default_factory=ContextConfig)
+    health: HealthConfig = field(default_factory=HealthConfig)
+    testing: TestingConfig = field(default_factory=TestingConfig)
+    strategy_selector: StrategySelectorConfig = field(
+        default_factory=StrategySelectorConfig
+    )
+    analyzers: AnalyzersConfig = field(default_factory=AnalyzersConfig)
+    refactor: RefactorConfig = field(default_factory=RefactorConfig)
+    clarity: ClarityConfig = field(default_factory=ClarityConfig)
+    complexity: ComplexityConfig = field(default_factory=ComplexityConfig)
+    parsing: ParsingConfig = field(default_factory=ParsingConfig)
+    execution: ExecutionConfig = field(default_factory=ExecutionConfig)
+    action: ActionConfig = field(default_factory=ActionConfig)
+    prompt_pipeline: PromptPipelineConfig = field(default_factory=PromptPipelineConfig)
+    validator: ValidatorConfig = field(default_factory=ValidatorConfig)
+    memory: MemoryConfig = field(default_factory=MemoryConfig)
+    repositories: RepositoriesConfig = field(default_factory=RepositoriesConfig)
+    git: GitConfig = field(default_factory=GitConfig)
+    vectors: VectorsConfig = field(default_factory=VectorsConfig)
+    sync: SyncConfig = field(default_factory=SyncConfig)
+    authority_package: AuthorityPackageConfig = field(
+        default_factory=AuthorityPackageConfig
+    )
+    strategic_auditor: StrategicAuditorConfig = field(
+        default_factory=StrategicAuditorConfig
+    )
+    misc: MiscConfig = field(default_factory=MiscConfig)
+
+
+# ---------------------------------------------------------------------------
+# Per-section factories
+# ---------------------------------------------------------------------------
+
+
+def _load_llm(raw: dict[str, Any]) -> LLMConfig:
+    sec = _section(raw, "llm")
+    return LLMConfig(
+        default_max_tokens=_get_int(sec, "default_max_tokens", 4096),
+        default_max_length=_get_int(sec, "default_max_length", 4096),
+        http_timeout_sec=_get_int(sec, "http_timeout_sec", 60),
+        request_timeout_sec=_get_int(sec, "request_timeout_sec", 300),
+        provider_timeout_sec=_get_int(sec, "provider_timeout_sec", 180),
+    )
+
+
+def _load_embedding(raw: dict[str, Any]) -> EmbeddingConfig:
+    sec = _section(raw, "embedding")
+    return EmbeddingConfig(
+        chunk_size=_get_int(sec, "chunk_size", 512),
+        chunk_overlap=_get_int(sec, "chunk_overlap", 50),
+        max_chars=_get_int(sec, "max_chars", 20000),
+        provider_request_timeout_sec=_get_float(
+            sec, "provider_request_timeout_sec", 120.0
+        ),
+        provider_connect_timeout_sec=_get_float(
+            sec, "provider_connect_timeout_sec", 10.0
+        ),
+        utils_request_timeout_sec=_get_float(sec, "utils_request_timeout_sec", 30.0),
+    )
+
+
+def _load_chunking(raw: dict[str, Any]) -> ChunkingConfig:
+    sec = _section(raw, "chunking")
+    return ChunkingConfig(max_chunk_chars=_get_int(sec, "max_chunk_chars", 1500))
+
+
+def _load_blackboard(raw: dict[str, Any]) -> BlackboardConfig:
+    sec = _section(raw, "blackboard")
+    return BlackboardConfig(
+        sla_default_seconds=_get_int(sec, "sla_default_seconds", 3600),
+        telemetry_ttl_days=_get_int(sec, "telemetry_ttl_days", 7),
+        telemetry_subject_prefixes=_get_str_tuple(
+            sec, "telemetry_subject_prefixes", ("loop_hold.sample::",)
+        ),
+        delegate_finding_ttl_days=_get_int(sec, "delegate_finding_ttl_days", 7),
+        delegate_finding_subjects=_get_str_tuple(
+            sec,
+            "delegate_finding_subjects",
+            (
+                "coherence.violation_executor.blast_bound",
+                "coherence.repo_artifacts.drift",
+            ),
+        ),
+        sweep_batch_max=_get_int(sec, "sweep_batch_max", 500),
+    )
+
+
+def _load_health_log(raw: dict[str, Any]) -> HealthLogConfig:
+    sec = _section(raw, "health_log")
+    return HealthLogConfig(
+        stale_threshold_seconds=_get_int(sec, "stale_threshold_seconds", 3600)
+    )
+
+
+def _load_daemon(raw: dict[str, Any]) -> DaemonConfig:
+    sec = _section(raw, "daemon")
+    return DaemonConfig(
+        one_shot_interval_sec=_get_int(sec, "one_shot_interval_sec", 300),
+        slow_callback_duration_sec=_get_float(sec, "slow_callback_duration_sec", 0.1),
+        set_debug=_get_bool(sec, "set_debug", False),
+    )
+
+
+# ID: 6a7b8c9d-0e1f-2a3b-4c5d-6e7f8a9b0c1d
+def _load_worker_classification(raw: dict[str, Any]) -> WorkerClassificationConfig:
+    sec = _section(raw, "worker_classification")
+    return WorkerClassificationConfig(
+        loop_hold_escalation_sec=_get_float(sec, "loop_hold_escalation_sec", 5.0),
+        loop_hold_deescalation_sec=_get_float(sec, "loop_hold_deescalation_sec", 1.0),
+        cycle_window=_get_int(sec, "cycle_window", 5),
+    )
+
+
+def _load_proposals(raw: dict[str, Any]) -> ProposalsConfig:
+    sec = _section(raw, "proposals")
+    return ProposalsConfig(
+        list_limit=_get_int(sec, "list_limit", 100),
+        pending_limit=_get_int(sec, "pending_limit", 50),
+    )
+
+
+def _load_consequence_log(raw: dict[str, Any]) -> ConsequenceLogConfig:
+    sec = _section(raw, "consequence_log")
+    return ConsequenceLogConfig(
+        default_lookback_seconds=_get_int(sec, "default_lookback_seconds", 3600)
+    )
+
+
+def _load_audit(raw: dict[str, Any]) -> AuditConfig:
+    sec = _section(raw, "audit")
+    return AuditConfig(
+        llm_gate_verdict_cache_ttl_days=_get_int(
+            sec, "llm_gate_verdict_cache_ttl_days", 30
+        ),
+        llm_gate_cache_staleness_threshold_seconds=_get_int(
+            sec, "llm_gate_cache_staleness_threshold_seconds", 3600
+        ),
+    )
+
+
+def _load_coverage(raw: dict[str, Any]) -> CoverageConfig:
+    sec = _section(raw, "coverage")
+    return CoverageConfig(
+        gap_threshold_pct=_get_float(sec, "gap_threshold_pct", 75.0),
+        warn_pct=_get_int(sec, "warn_pct", 75),
+        low_bucket_pct=_get_int(sec, "low_bucket_pct", 50),
+        gap_list_limit=_get_int(sec, "gap_list_limit", 10),
+        collect_timeout_sec=_get_int(sec, "collect_timeout_sec", 120),
+        full_run_timeout_sec=_get_int(sec, "full_run_timeout_sec", 300),
+        quality_min_pct=_get_int(sec, "quality_min_pct", 75),
+        watcher_required_pct=_get_float(sec, "watcher_required_pct", 75.0),
+        watcher_rescan_hours=_get_int(sec, "watcher_rescan_hours", 24),
+        batch_remediation_threshold_pct=_get_float(
+            sec, "batch_remediation_threshold_pct", 75.0
+        ),
+        batch_remediation_partial_success_pct=_get_int(
+            sec, "batch_remediation_partial_success_pct", 50
+        ),
+        single_file_target_pct=_get_float(sec, "single_file_target_pct", 75.0),
+    )
+
+
+def _load_workflow_gate(raw: dict[str, Any]) -> WorkflowGateConfig:
+    sec = _section(raw, "workflow_gate")
+    return WorkflowGateConfig(
+        linter_timeout_sec=_get_float(sec, "linter_timeout_sec", 30.0),
+        quality_timeout_sec=_get_float(sec, "quality_timeout_sec", 60.0),
+        ruff_format_timeout_sec=_get_float(sec, "ruff_format_timeout_sec", 60.0),
+        import_timeout_sec=_get_float(sec, "import_timeout_sec", 60.0),
+    )
+
+
+def _load_context(raw: dict[str, Any]) -> ContextConfig:
+    sec = _section(raw, "context")
+    return ContextConfig(
+        score_target_file=_get_int(sec, "score_target_file", 100),
+        score_target_path=_get_int(sec, "score_target_path", 80),
+        score_target_symbol=_get_int(sec, "score_target_symbol", 120),
+        score_has_content=_get_int(sec, "score_has_content", 30),
+        vector_top_k=_get_int(sec, "vector_top_k", 10),
+        db_provider_max_items=_get_int(sec, "db_provider_max_items", 100),
+        token_estimate_overhead=_get_int(sec, "token_estimate_overhead", 300),
+        cache_ttl_hours=_get_int(sec, "cache_ttl_hours", 24),
+        db_recent_packets_limit=_get_int(sec, "db_recent_packets_limit", 10),
+    )
+
+
+def _load_health(raw: dict[str, Any]) -> HealthConfig:
+    sec = _section(raw, "health")
+    return HealthConfig(
+        worker_alive_threshold_sec=_get_int(sec, "worker_alive_threshold_sec", 600),
+        worker_warn_threshold_sec=_get_int(sec, "worker_warn_threshold_sec", 3600),
+        long_lookback_hours=_get_int(sec, "long_lookback_hours", 24),
+        medium_lookback_minutes=_get_int(sec, "medium_lookback_minutes", 60),
+        short_lookback_minutes=_get_int(sec, "short_lookback_minutes", 30),
+        recent_lookback_minutes=_get_int(sec, "recent_lookback_minutes", 10),
+    )
+
+
+def _load_testing(raw: dict[str, Any]) -> TestingConfig:
+    sec = _section(raw, "testing")
+    return TestingConfig(
+        pytest_collection_timeout_sec=_get_int(
+            sec, "pytest_collection_timeout_sec", 30
+        ),
+        pytest_execution_timeout_sec=_get_int(sec, "pytest_execution_timeout_sec", 300),
+        sandbox_timeout_sec=_get_int(sec, "sandbox_timeout_sec", 30),
+        executor_timeout_sec=_get_int(sec, "executor_timeout_sec", 60),
+        metrics_timeout_sec=_get_int(sec, "metrics_timeout_sec", 30),
+        simple_gen_timeout_sec=_get_float(sec, "simple_gen_timeout_sec", 20.0),
+        context_aware_gen_timeout_sec=_get_float(
+            sec, "context_aware_gen_timeout_sec", 15.0
+        ),
+        runtime_validator_timeout_sec=_get_int(
+            sec, "runtime_validator_timeout_sec", 60
+        ),
+        max_failures=_get_int(sec, "max_failures", 10),
+        snippet_max_lines=_get_int(sec, "snippet_max_lines", 20),
+    )
+
+
+def _load_strategy_selector(raw: dict[str, Any]) -> StrategySelectorConfig:
+    sec = _section(raw, "strategy_selector")
+    return StrategySelectorConfig(
+        min_recommended_score=_get_int(sec, "min_recommended_score", 15),
+        score_role_preferred=_get_int(sec, "score_role_preferred", 25),
+        score_role_discouraged=_get_int(sec, "score_role_discouraged", 25),
+        score_generalized_match=_get_int(sec, "score_generalized_match", 16),
+        score_rule_preferred=_get_int(sec, "score_rule_preferred", 12),
+        score_rule_discouraged=_get_int(sec, "score_rule_discouraged", 12),
+        score_size_bonus=_get_int(sec, "score_size_bonus", 10),
+        score_size_penalty=_get_int(sec, "score_size_penalty", 10),
+        score_strong_split=_get_int(sec, "score_strong_split", 14),
+        score_cluster_count_bonus=_get_int(sec, "score_cluster_count_bonus", 8),
+        score_cluster_count_penalty=_get_int(sec, "score_cluster_count_penalty", 8),
+        score_constraint_role_bonus=_get_int(sec, "score_constraint_role_bonus", 9),
+        score_constraint_role_penalty=_get_int(sec, "score_constraint_role_penalty", 8),
+        score_conservatism_bias_bonus=_get_int(sec, "score_conservatism_bias_bonus", 8),
+        score_conservatism_structural_bonus=_get_int(
+            sec, "score_conservatism_structural_bonus", 5
+        ),
+        large_file_lines=_get_int(sec, "large_file_lines", 400),
+        small_file_lines=_get_int(sec, "small_file_lines", 200),
+        strong_split_lines=_get_int(sec, "strong_split_lines", 450),
+    )
+
+
+def _load_analyzers(raw: dict[str, Any]) -> AnalyzersConfig:
+    sec = _section(raw, "analyzers")
+    return AnalyzersConfig(
+        file_complexity_high_threshold=_get_int(
+            sec, "file_complexity_high_threshold", 15
+        ),
+        class_methods_high_threshold=_get_int(sec, "class_methods_high_threshold", 10),
+        function_body_high_threshold=_get_int(sec, "function_body_high_threshold", 25),
+        function_body_low_threshold=_get_int(sec, "function_body_low_threshold", 10),
+        max_file_lines=_get_int(sec, "max_file_lines", 400),
+        max_function_lines=_get_int(sec, "max_function_lines", 50),
+        max_module_lines=_get_int(sec, "max_module_lines", 400),
+    )
+
+
+def _load_refactor(raw: dict[str, Any]) -> RefactorConfig:
+    sec = _section(raw, "refactor")
+    return RefactorConfig(
+        responsibilities_threshold=_get_int(sec, "responsibilities_threshold", 20),
+        cohesion_threshold=_get_int(sec, "cohesion_threshold", 12),
+        coupling_threshold=_get_int(sec, "coupling_threshold", 10),
+        loc_threshold=_get_int(sec, "loc_threshold", 400),
+    )
+
+
+def _load_clarity(raw: dict[str, Any]) -> ClarityConfig:
+    sec = _section(raw, "clarity")
+    return ClarityConfig(
+        structural_complexity=_get_int(sec, "structural_complexity", 20),
+        structural_lines=_get_int(sec, "structural_lines", 300),
+        logic_simplification_threshold=_get_int(
+            sec, "logic_simplification_threshold", 10
+        ),
+    )
+
+
+def _load_complexity(raw: dict[str, Any]) -> ComplexityConfig:
+    sec = _section(raw, "complexity")
+    return ComplexityConfig(
+        god_method_threshold=_get_int(sec, "god_method_threshold", 30),
+        extraction_threshold=_get_int(sec, "extraction_threshold", 15),
+    )
+
+
+def _load_parsing(raw: dict[str, Any]) -> ParsingConfig:
+    sec = _section(raw, "parsing")
+    return ParsingConfig(
+        min_block_len=_get_int(sec, "min_block_len", 10),
+        score_test_fn=_get_int(sec, "score_test_fn", 1000),
+        score_test_class=_get_int(sec, "score_test_class", 1000),
+        score_import=_get_int(sec, "score_import", 100),
+        score_pytest=_get_int(sec, "score_pytest", 500),
+    )
+
+
+def _load_execution(raw: dict[str, Any]) -> ExecutionConfig:
+    sec = _section(raw, "execution")
+    return ExecutionConfig(
+        task_timeout_sec=_get_int(sec, "task_timeout_sec", 300),
+        workflow_timeout_minutes=_get_int(sec, "workflow_timeout_minutes", 30),
+        orchestrator_max_steps=_get_int(sec, "orchestrator_max_steps", 10),
+    )
+
+
+def _load_action(raw: dict[str, Any]) -> ActionConfig:
+    sec = _section(raw, "action")
+    return ActionConfig(
+        max_data_size_bytes=_get_int(sec, "max_data_size_bytes", 5_242_880)
+    )
+
+
+def _load_prompt_pipeline(raw: dict[str, Any]) -> PromptPipelineConfig:
+    sec = _section(raw, "prompt_pipeline")
+    return PromptPipelineConfig(
+        max_file_size_bytes=_get_int(sec, "max_file_size_bytes", 1_048_576)
+    )
+
+
+def _load_validator(raw: dict[str, Any]) -> ValidatorConfig:
+    sec = _section(raw, "validator")
+    return ValidatorConfig(
+        lru_cache_size=_get_int(sec, "lru_cache_size", 1024),
+        lru_cache_size_small=_get_int(sec, "lru_cache_size_small", 512),
+    )
+
+
+def _load_memory(raw: dict[str, Any]) -> MemoryConfig:
+    sec = _section(raw, "memory")
+    return MemoryConfig(
+        recency_days=_get_int(sec, "recency_days", 30),
+        episode_retention_days=_get_int(sec, "episode_retention_days", 30),
+        reflection_retention_days=_get_int(sec, "reflection_retention_days", 90),
+    )
+
+
+def _load_repositories(raw: dict[str, Any]) -> RepositoriesConfig:
+    sec = _section(raw, "repositories")
+    return RepositoriesConfig(
+        decision_trace_default_limit=_get_int(sec, "decision_trace_default_limit", 10),
+        decision_trace_max_limit=_get_int(sec, "decision_trace_max_limit", 100),
+        decision_trace_retention_days=_get_int(
+            sec, "decision_trace_retention_days", 30
+        ),
+        refusal_default_limit=_get_int(sec, "refusal_default_limit", 20),
+        refusal_by_type_limit=_get_int(sec, "refusal_by_type_limit", 50),
+        symbol_definition_default_limit=_get_int(
+            sec, "symbol_definition_default_limit", 500
+        ),
+        project_def_max_limit=_get_int(sec, "project_def_max_limit", 500),
+        project_def_default_limit=_get_int(sec, "project_def_default_limit", 100),
+    )
+
+
+def _load_git(raw: dict[str, Any]) -> GitConfig:
+    sec = _section(raw, "git")
+    return GitConfig(
+        recent_commits_n=_get_int(sec, "recent_commits_n", 10),
+        changed_files_log_n=_get_int(sec, "changed_files_log_n", 20),
+    )
+
+
+def _load_vectors(raw: dict[str, Any]) -> VectorsConfig:
+    sec = _section(raw, "vectors")
+    return VectorsConfig(
+        index_batch_size=_get_int(sec, "index_batch_size", 10),
+        scan_limit=_get_int(sec, "scan_limit", 10000),
+        report_preview_count=_get_int(sec, "report_preview_count", 10),
+        policy_vectorizer_batch_size=_get_int(sec, "policy_vectorizer_batch_size", 10),
+    )
+
+
+def _load_sync(raw: dict[str, Any]) -> SyncConfig:
+    sec = _section(raw, "sync")
+    return SyncConfig(
+        artifact_embed_batch_size=_get_int(sec, "artifact_embed_batch_size", 10),
+        policy_index_batch_size=_get_int(sec, "policy_index_batch_size", 10),
+        pattern_index_batch_size=_get_int(sec, "pattern_index_batch_size", 10),
+        specs_index_batch_size=_get_int(sec, "specs_index_batch_size", 10),
+    )
+
+
+def _load_authority_package(raw: dict[str, Any]) -> AuthorityPackageConfig:
+    sec = _section(raw, "authority_package")
+    return AuthorityPackageConfig(search_limit=_get_int(sec, "search_limit", 10))
+
+
+def _load_strategic_auditor(raw: dict[str, Any]) -> StrategicAuditorConfig:
+    sec = _section(raw, "strategic_auditor")
+    return StrategicAuditorConfig(
+        sample_limit=_get_int(sec, "sample_limit", 100),
+        commit_lookback=_get_int(sec, "commit_lookback", 15),
+        compact_max_chars=_get_int(sec, "compact_max_chars", 1500),
+    )
+
+
+def _load_misc(raw: dict[str, Any]) -> MiscConfig:
+    sec = _section(raw, "misc")
+    return MiscConfig(
+        linelength_max_line_chars=_get_int(sec, "linelength_max_line_chars", 100),
+        enrichment_description_max_chars=_get_int(
+            sec, "enrichment_description_max_chars", 500
+        ),
+        conversation_max_content_chars=_get_int(
+            sec, "conversation_max_content_chars", 2000
+        ),
+        code_snippet_context_lines=_get_int(sec, "code_snippet_context_lines", 20),
+        embedding_search_default_limit=_get_int(
+            sec, "embedding_search_default_limit", 10
+        ),
+        retriever_search_limit=_get_int(sec, "retriever_search_limit", 10),
+        retriever_context_lines=_get_int(sec, "retriever_context_lines", 20),
+        file_navigator_read_max_lines=_get_int(
+            sec, "file_navigator_read_max_lines", 200
+        ),
+        file_navigator_max_read_bytes=_get_int(
+            sec, "file_navigator_max_read_bytes", 1_048_576
+        ),
+        knowledge_consolidation_max_lines=_get_int(
+            sec, "knowledge_consolidation_max_lines", 10
+        ),
+        context_export_http_timeout_sec=_get_int(
+            sec, "context_export_http_timeout_sec", 10
+        ),
+        metadata_max_comment_length=_get_int(sec, "metadata_max_comment_length", 120),
+        census_hotspot_limit=_get_int(sec, "census_hotspot_limit", 10),
+        clustering_default_n_clusters=_get_int(
+            sec, "clustering_default_n_clusters", 15
+        ),
+        legacy_scan_display_limit=_get_int(sec, "legacy_scan_display_limit", 10),
+        limb_status_recent_limit=_get_int(sec, "limb_status_recent_limit", 15),
+        refusal_inspect_default_limit=_get_int(
+            sec, "refusal_inspect_default_limit", 20
+        ),
+        refusal_inspect_by_type_limit=_get_int(
+            sec, "refusal_inspect_by_type_limit", 20
+        ),
+        capability_tagging_default_llm_confidence=_get_float(
+            sec, "capability_tagging_default_llm_confidence", 0.70
+        ),
+        knowledge_min_occurrences=_get_int(sec, "knowledge_min_occurrences", 3),
+        knowledge_max_lines=_get_int(sec, "knowledge_max_lines", 10),
+        context_aware_test_context_lines=_get_int(
+            sec, "context_aware_test_context_lines", 40
+        ),
+        perf_overhead_warning_pct=_get_int(sec, "perf_overhead_warning_pct", 50),
+        perf_overhead_error_pct=_get_int(sec, "perf_overhead_error_pct", 100),
+        context_search_display_limit=_get_int(sec, "context_search_display_limit", 20),
+        proposals_display_limit=_get_int(sec, "proposals_display_limit", 20),
+    )
+
+
+def _load_workers(raw: dict[str, Any]) -> WorkersConfig:
+    workers = _section(raw, "workers")
+    return WorkersConfig(
+        call_site_rewriter=WorkerCallSiteRewriterConfig(
+            claim_limit=_get_int(
+                _section(workers, "call_site_rewriter"), "claim_limit", 50
+            )
+        ),
+        doc_writer=WorkerDocWriterConfig(
+            batch_size=_get_int(_section(workers, "doc_writer"), "batch_size", 25)
+        ),
+        doc_worker=WorkerDocWorkerConfig(
+            batch_size=_get_int(_section(workers, "doc_worker"), "batch_size", 50)
+        ),
+        prompt_artifact_writer=WorkerPromptArtifactWriterConfig(
+            claim_limit=_get_int(
+                _section(workers, "prompt_artifact_writer"), "claim_limit", 25
+            )
+        ),
+        prompt_extractor=WorkerPromptExtractorConfig(
+            claim_limit=_get_int(
+                _section(workers, "prompt_extractor"), "claim_limit", 25
+            ),
+            context_lines=_get_int(
+                _section(workers, "prompt_extractor"), "context_lines", 40
+            ),
+        ),
+        capability_tagger=WorkerCapabilityTaggerConfig(
+            batch_size=_get_int(
+                _section(workers, "capability_tagger"), "batch_size", 20
+            )
+        ),
+        proposal_consumer=WorkerProposalConsumerConfig(
+            claim_limit=_get_int(
+                _section(workers, "proposal_consumer"), "claim_limit", 5
+            )
+        ),
+        violation_executor=WorkerViolationExecutorConfig(
+            claim_limit=_get_int(
+                _section(workers, "violation_executor"), "claim_limit", 50
+            )
+        ),
+        violation_remediator=WorkerViolationRemediatorConfig(
+            claim_limit=_get_int(
+                _section(workers, "violation_remediator"), "claim_limit", 50
+            ),
+            scan_limit=_get_int(
+                _section(workers, "violation_remediator"), "scan_limit", 200
+            ),
+            ceremony_timeout_sec=_get_int(
+                _section(workers, "violation_remediator"), "ceremony_timeout_sec", 30
+            ),
+            semantic_examples_limit=_get_int(
+                _section(workers, "violation_remediator"),
+                "semantic_examples_limit",
+                3,
+            ),
+            min_role_confidence=_get_float(
+                _section(workers, "violation_remediator"),
+                "min_role_confidence",
+                0.55,
+            ),
+        ),
+        worker_shop=WorkerShopConfig(
+            glide_off_multiplier=_get_float(
+                _section(workers, "worker_shop"), "glide_off_multiplier", 0.10
+            ),
+            fallback_threshold_sec=_get_int(
+                _section(workers, "worker_shop"), "fallback_threshold_sec", 600
+            ),
+            findings_scan_limit=_get_int(
+                _section(workers, "worker_shop"), "findings_scan_limit", 200
+            ),
+        ),
+        proposal_pipeline_shop=WorkerProposalPipelineShopConfig(
+            stuck_approved_sla_sec=_get_int(
+                _section(workers, "proposal_pipeline_shop"),
+                "stuck_approved_sla_sec",
+                1800,
+            ),
+            stuck_executing_sla_sec=_get_int(
+                _section(workers, "proposal_pipeline_shop"),
+                "stuck_executing_sla_sec",
+                900,
+            ),
+            repeated_failure_threshold=_get_int(
+                _section(workers, "proposal_pipeline_shop"),
+                "repeated_failure_threshold",
+                3,
+            ),
+            repeated_failure_lookback_sec=_get_int(
+                _section(workers, "proposal_pipeline_shop"),
+                "repeated_failure_lookback_sec",
+                86400,
+            ),
+            findings_scan_limit=_get_int(
+                _section(workers, "proposal_pipeline_shop"),
+                "findings_scan_limit",
+                200,
+            ),
+        ),
+        test_remediator=WorkerTestRemediatorConfig(
+            scan_limit=_get_int(_section(workers, "test_remediator"), "scan_limit", 200)
+        ),
+        test_runner_sensor=WorkerTestRunnerSensorConfig(
+            scan_limit=_get_int(
+                _section(workers, "test_runner_sensor"), "scan_limit", 50
+            )
+        ),
+        intent_inspector=WorkerIntentInspectorConfig(
+            alignment_batch=_get_int(
+                _section(workers, "intent_inspector"), "alignment_batch", 20
+            )
+        ),
+        observer=WorkerObserverConfig(
+            stale_threshold_seconds=_get_int(
+                _section(workers, "observer"), "stale_threshold_seconds", 3600
+            )
+        ),
+        coherence_sensor=WorkerCoherenceSensorConfig(
+            lookback_seconds=_get_int(
+                _section(workers, "coherence_sensor"), "lookback_seconds", 7200
+            )
+        ),
+    )
+
+
+# ---------------------------------------------------------------------------
+# Public loader
+# ---------------------------------------------------------------------------
+
+
+# ID: f5b20a98-a3b5-4e8d-a19e-f310639cd2a9
+def load_operational_config() -> OperationalConfig:
+    """
+    Load .intent/enforcement/config/operational_config.yaml via IntentRepository.
+
+    Returns an OperationalConfig with every section populated from the
+    YAML file. On any failure — missing file, parse error, non-mapping
+    section, type-coercion error on a single field — the loader logs a
+    warning at WARNING level and substitutes the matching fallback
+    default. It never raises, so callers can use the result unconditionally.
+
+    The drift window is bounded by IntentRepository.reload() on the next
+    audit-sensor cycle (ADR-039).
+    """
+    raw: dict[str, Any] = {}
+    try:
+        from shared.infrastructure.intent.intent_repository import (
+            get_intent_repository,
+        )
+
+        repo = get_intent_repository()
+        config_path = repo.resolve_rel("enforcement/config/operational_config.yaml")
+        loaded = repo.load_document(config_path)
+        if isinstance(loaded, dict):
+            raw = loaded
+        else:
+            logger.warning(
+                "operational_config: operational_config.yaml did not parse as a "
+                "dict — using fallback defaults."
+            )
+    except Exception as exc:
+        logger.warning(
+            "operational_config: could not load .intent/enforcement/config/"
+            "operational_config.yaml (%s) — using fallback defaults.",
+            exc,
+        )
+
+    return OperationalConfig(
+        llm=_load_llm(raw),
+        embedding=_load_embedding(raw),
+        chunking=_load_chunking(raw),
+        workers=_load_workers(raw),
+        blackboard=_load_blackboard(raw),
+        health_log=_load_health_log(raw),
+        daemon=_load_daemon(raw),
+        worker_classification=_load_worker_classification(raw),
+        proposals=_load_proposals(raw),
+        consequence_log=_load_consequence_log(raw),
+        audit=_load_audit(raw),
+        coverage=_load_coverage(raw),
+        workflow_gate=_load_workflow_gate(raw),
+        context=_load_context(raw),
+        health=_load_health(raw),
+        testing=_load_testing(raw),
+        strategy_selector=_load_strategy_selector(raw),
+        analyzers=_load_analyzers(raw),
+        refactor=_load_refactor(raw),
+        clarity=_load_clarity(raw),
+        complexity=_load_complexity(raw),
+        parsing=_load_parsing(raw),
+        execution=_load_execution(raw),
+        action=_load_action(raw),
+        prompt_pipeline=_load_prompt_pipeline(raw),
+        validator=_load_validator(raw),
+        memory=_load_memory(raw),
+        repositories=_load_repositories(raw),
+        git=_load_git(raw),
+        vectors=_load_vectors(raw),
+        sync=_load_sync(raw),
+        authority_package=_load_authority_package(raw),
+        strategic_auditor=_load_strategic_auditor(raw),
+        misc=_load_misc(raw),
+    )