from __future__ import annotations

from src.services.agent_nemotron_smoke_gate import (
    evaluate_nemotron_contract_tuned_smoke_gate,
)


def test_smoke_gate_blocks_latency_even_when_runner_is_valid():
    report = evaluate_nemotron_contract_tuned_smoke_gate(
        runner_report={
            "valid": True,
            "candidate_variant_id": "nemo_nemotron_fabric_contract_tuned_v1",
            "requests": 5,
            "results": 5,
            "external_error_records": 0,
            "fallback_used_records": 0,
            "trace_incomplete_records": 0,
            "retry_used_records": 1,
            "avg_latency_ms": 200000,
            "p95_latency_ms": 374591.0851,
            "model": "nvidia/nemotron-3-super-120b-a12b",
        }
    ).to_dict()

    assert report["approved_for_full_replay"] is False
    assert report["decision"] == "blocked"
    assert report["gates"]["runner_valid"] is True
    assert report["gates"]["latency_budget_met"] is False
    assert report["failures"] == ["latency_budget_exceeded"]
    assert report["runner_summary"]["retry_used_records"] == 1


def test_smoke_gate_approves_clean_fast_smoke():
    report = evaluate_nemotron_contract_tuned_smoke_gate(
        runner_report={
            "valid": True,
            "candidate_variant_id": "nemo_nemotron_fabric_contract_tuned_v1",
            "requests": 5,
            "results": 5,
            "external_error_records": 0,
            "fallback_used_records": 0,
            "trace_incomplete_records": 0,
            "retry_used_records": 0,
            "avg_latency_ms": 20000,
            "p95_latency_ms": 44000,
            "model": "nvidia/nemotron-3-super-120b-a12b",
        }
    ).to_dict()

    assert report["approved_for_full_replay"] is True
    assert report["decision"] == "approved_for_full_replay"
    assert report["gates"]["latency_budget_met"] is True