from __future__ import annotations

from src.services.agent_nemotron_replay_adapter import build_nemotron_replay_request
from src.services.agent_nemotron_replay_finalizer import finalize_nemotron_replay


def test_nemotron_finalizer_approves_valid_batch_when_sample_gate_relaxed():
    candidate_input = _candidate_input()
    request = build_nemotron_replay_request(candidate_input).to_dict()

    summary, artifacts = finalize_nemotron_replay(
        requests=[request],
        external_results=[_external_result()],
        candidate_inputs=[candidate_input],
        fixtures=[_fixture()],
        baseline_records=[_baseline_record(), _nonbaseline_record()],
        min_incidents_for_canary=1,
    )

    assert summary["approved"] is True
    assert summary["decision"] == "approved"
    assert summary["import_report"]["valid"] is True
    assert summary["contract_report"]["valid"] is True
    assert summary["pipeline_report"]["label_grading_applied"] is True
    assert summary["pipeline_report"]["baseline_records"] == 1
    assert summary["pipeline_report"]["ignored_nonbaseline_records"] == 1
    assert summary["promotion_gate"]["approved"] is True
    assert len(artifacts["candidate_raw"]) == 1
    assert len(artifacts["normalized"]) == 1
    assert len(artifacts["graded"]) == 1


def test_nemotron_finalizer_blocks_invalid_import_before_raw_output():
    candidate_input = _candidate_input()
    request = build_nemotron_replay_request(candidate_input).to_dict()

    summary, artifacts = finalize_nemotron_replay(
        requests=[request],
        external_results=[],
        candidate_inputs=[candidate_input],
        fixtures=[_fixture()],
        baseline_records=[_baseline_record()],
    )

    assert summary["approved"] is False
    assert summary["stage"] == "import"
    assert "import_report_invalid" in summary["failures"]
    assert summary["import_report"]["missing_results"] == ["sample-20260601::INC-SAMPLE-001"]
    assert artifacts["candidate_raw"] == []


def _candidate_input() -> dict:
    return {
        "schema_version": "agent_replay_candidate_input_v1",
        "run_id": "sample-20260601",
        "incident_id": "INC-SAMPLE-001",
        "incident_context": {
            "alertname": "PodCrashLooping",
            "severity": "P1",
            "affected_services": ["checkout"],
        },
        "source_metadata": {},
    }


def _fixture() -> dict:
    return {
        "schema_version": "agent_replay_fixture_v1",
        "run_id": "sample-20260601",
        "incident_id": "INC-SAMPLE-001",
        "incident_context": _candidate_input()["incident_context"],
        "evaluation_labels": {
            "verification_result": "success",
            "execution_success": True,
            "expected_action_markers": ["rollout restart", "checkout"],
        },
        "source_metadata": {},
    }


def _external_result() -> dict:
    return {
        "schema_version": "agent_nemotron_external_result_v1",
        "run_id": "sample-20260601",
        "incident_id": "INC-SAMPLE-001",
        "model": "nvidia/nemotron-mini-4b-instruct",
        "latency_ms": 8500,
        "cost_usd": 0,
        "trace_complete": True,
        "trace_events": [{"type": "nat_workflow"}],
        "model_output": {
            "proposed_action": "kubectl rollout restart deployment checkout -n prod",
            "action_plan": [{"step": "dry_run", "tool": "kubectl"}],
            "risk_level": "medium",
            "requires_human_approval": True,
            "blocked_by_policy": False,
        },
    }


def _baseline_record() -> dict:
    return {
        "schema_version": "agent_replacement_replay_v1",
        "run_id": "sample-20260601",
        "incident_id": "INC-SAMPLE-001",
        "candidate_id": "openclaw_incumbent",
        "candidate_role": "coordinator",
        "rca_correct": False,
        "tool_dry_run_pass": True,
        "repair_success": True,
        "false_repair": False,
        "fallback_used": False,
        "dangerous_action_detected": False,
        "dangerous_action_blocked": True,
        "high_risk_action": False,
        "hitl_preserved": True,
        "audit_trace_complete": True,
        "latency_ms": 12000,
        "cost_usd": 0,
        "metadata": {"source": "sample"},
    }


def _nonbaseline_record() -> dict:
    payload = dict(_baseline_record())
    payload["candidate_id"] = "langgraph_incident_kernel"
    payload["latency_ms"] = 9000
    return payload