from __future__ import annotations

import pytest

from backend.app.core.config import settings
from backend.app.services.campaign_vectorizer import FEATURE_KEYS
from backend.app.services.blueprint_engine import build_blueprint, score_campaign_for_segment
from backend.app.services.campaign_vectorizer import extract_campaign_features
from backend.app.services.gmm_engine import run_gmm_clustering
from backend.app.services.rag_optimizer import optimize_campaign


@pytest.fixture(autouse=True)
def _force_heuristic_mode(monkeypatch: pytest.MonkeyPatch) -> None:
    monkeypatch.setattr(settings, "ollama_api_key", "")


def test_heuristic_vectorizer_changes_with_campaign_text() -> None:
    evidence_text = (
        "Phase 3 randomized trial data demonstrated improved progression-free survival, "
        "safety, and endpoint performance in eligible oncology patients."
    )
    access_text = (
        "Digital omnichannel launch with budget-friendly access messaging, rapid adoption, "
        "and broad patient reach across regional accounts."
    )

    evidence_features = extract_campaign_features(evidence_text)["normalized_features"]
    access_features = extract_campaign_features(access_text)["normalized_features"]

    total_delta = sum(abs(evidence_features[key] - access_features[key]) for key in evidence_features)
    assert total_delta > 0.5
    assert evidence_features["trial_phase_relevance"] > access_features["trial_phase_relevance"]
    assert access_features["channel_preference"] > evidence_features["channel_preference"]


def test_optimize_campaign_returns_offline_improvements() -> None:
    campaign_text = "Broad brand campaign with limited clinical detail and minimal access context."

    optimized = optimize_campaign(campaign_text, target_cluster=2)

    assert optimized["original_text"] == campaign_text
    assert optimized["optimized_text"] != campaign_text
    assert optimized["improvements"]
    assert optimized["alignment_score"] is not None


def test_build_blueprint_returns_complete_payload() -> None:
    blueprint = build_blueprint(
        2,
        campaign_text="Budget-sensitive launch planning for a digital-first oncology message.",
        core_message="Refined positioning for the segment.",
    )

    assert blueprint["segmentId"] == "2"
    assert blueprint["segmentName"]
    assert blueprint["coreClinicalMessage"] == "Refined positioning for the segment."
    assert len(blueprint["optimalChannels"]) == 4
    assert len(blueprint["ragInjections"]) == 3
    assert len(blueprint["tacticalSteps"]) == 3
    assert len(blueprint["modifications"]) == 3
    assert blueprint["gapSummary"]["largestGap"] is not None


def test_score_campaign_for_segment_accepts_string_cluster_ids() -> None:
    result = score_campaign_for_segment(
        "c2",
        "Randomized study evidence with endpoint detail for specialist stakeholders.",
    )

    assert result["cluster_id"] == 2
    assert result["status"] in {"approved", "rejected"}
    assert len(result["alternatives"]) == 3
    assert result["feature_gaps"]


def test_score_campaign_uses_fitted_gmm_artifacts_when_supplied() -> None:
    records = [
        {
            key: min(0.95, max(0.05, ((idx + offset) % 9) / 8.0))
            for offset, key in enumerate(FEATURE_KEYS)
        }
        for idx in range(24)
    ]
    artifacts = run_gmm_clustering(records, min_k=2, feature_names=list(FEATURE_KEYS))

    result = score_campaign_for_segment(
        0,
        "Phase 3 oncology launch with digital KOL education and access proof.",
        gmm_artifacts=artifacts,
    )

    assert result["statistical_artifacts"] == "fitted_gmm_covariance"
    assert result["statistical_artifacts_reason"] == ""
    assert len(result["heatmap"]["ranking"]) == artifacts["k"]