stonks-oracle/tests/test_suppression.py

"""Tests for recommendation suppression logic (data quality checks).

Requirements: 7.4
"""
from datetime import datetime, timedelta, timezone

from services.recommendation.suppression import (
    DataQualityContext,
    SuppressionConfig,
    SuppressionReason,
    build_quality_context_from_summary,
    evaluate_suppression,
)
from services.shared.schemas import TrendDirection, TrendSummary, TrendWindow

NOW = datetime(2026, 4, 11, 12, 0, 0, tzinfo=timezone.utc)


def _make_summary(**overrides) -> TrendSummary:
    defaults = dict(
        entity_type="company",
        entity_id="AAPL",
        window=TrendWindow.SEVEN_DAY,
        trend_direction=TrendDirection.BULLISH,
        trend_strength=0.5,
        confidence=0.65,
        top_supporting_evidence=["doc1", "doc2", "doc3"],
        top_opposing_evidence=[],
        dominant_catalysts=["earnings"],
        material_risks=["regulatory scrutiny"],
        contradiction_score=0.1,
        generated_at=NOW,
    )
    defaults.update(overrides)
    return TrendSummary(**defaults)


def _make_quality_ctx(**overrides) -> DataQualityContext:
    defaults = dict(
        total_documents=5,
        valid_documents=4,
        failed_documents=1,
        avg_extraction_confidence=0.7,
        newest_evidence_at=NOW - timedelta(hours=6),
        source_types={"news_api", "filings_api"},
    )
    defaults.update(overrides)
    return DataQualityContext(**defaults)


# ---------------------------------------------------------------------------
# No suppression for good quality data
# ---------------------------------------------------------------------------


def test_no_suppression_good_quality():
    summary = _make_summary()
    ctx = _make_quality_ctx()
    result = evaluate_suppression(summary, ctx, reference_time=NOW)
    assert result.suppressed is False
    assert result.reasons == []
    assert result.data_quality_score > 0.3


# ---------------------------------------------------------------------------
# Suppression triggers
# ---------------------------------------------------------------------------


def test_suppressed_low_extraction_confidence():
    summary = _make_summary()
    ctx = _make_quality_ctx(avg_extraction_confidence=0.2)
    result = evaluate_suppression(summary, ctx, reference_time=NOW)
    assert result.suppressed is True
    assert SuppressionReason.LOW_DATA_CONFIDENCE in result.reasons


def test_suppressed_stale_evidence():
    summary = _make_summary()
    ctx = _make_quality_ctx(newest_evidence_at=NOW - timedelta(days=10))
    result = evaluate_suppression(summary, ctx, reference_time=NOW)
    assert result.suppressed is True
    assert SuppressionReason.STALE_EVIDENCE in result.reasons


def test_suppressed_high_failure_rate():
    summary = _make_summary()
    ctx = _make_quality_ctx(total_documents=10, failed_documents=6, valid_documents=4)
    result = evaluate_suppression(summary, ctx, reference_time=NOW)
    assert result.suppressed is True
    assert SuppressionReason.HIGH_EXTRACTION_FAILURE_RATE in result.reasons


def test_suppressed_insufficient_valid_documents():
    summary = _make_summary(
        top_supporting_evidence=["doc1"],
        top_opposing_evidence=[],
    )
    ctx = _make_quality_ctx(total_documents=1, valid_documents=1, failed_documents=0)
    result = evaluate_suppression(summary, ctx, reference_time=NOW)
    assert result.suppressed is True
    assert SuppressionReason.INSUFFICIENT_VALID_DOCUMENTS in result.reasons


def test_suppressed_low_source_diversity():
    """When min_source_types > available source types, suppression fires."""
    summary = _make_summary()
    ctx = _make_quality_ctx(source_types=set())
    config = SuppressionConfig(min_source_types=2)
    result = evaluate_suppression(summary, ctx, config=config, reference_time=NOW)
    assert result.suppressed is True
    assert SuppressionReason.LOW_SOURCE_DIVERSITY in result.reasons


# ---------------------------------------------------------------------------
# Fallback to summary-based context
# ---------------------------------------------------------------------------


def test_fallback_context_from_summary():
    summary = _make_summary(confidence=0.7)
    ctx = build_quality_context_from_summary(summary)
    assert ctx.total_documents == 3  # 3 supporting + 0 opposing
    assert ctx.valid_documents == 3
    assert ctx.avg_extraction_confidence == 0.7


def test_no_suppression_with_summary_fallback():
    """When no quality context is provided, summary-based fallback is used."""
    summary = _make_summary(confidence=0.7)
    # Default config has min_source_types=1. The fallback now returns
    # source_types={"unknown"} so LOW_SOURCE_DIVERSITY does not fire.
    result = evaluate_suppression(summary, config=SuppressionConfig(), reference_time=NOW)
    assert result.suppressed is False


# ---------------------------------------------------------------------------
# Data quality score
# ---------------------------------------------------------------------------


def test_quality_score_high_for_good_data():
    summary = _make_summary()
    ctx = _make_quality_ctx(
        avg_extraction_confidence=0.85,
        newest_evidence_at=NOW - timedelta(hours=1),
        total_documents=10,
        valid_documents=10,
        failed_documents=0,
    )
    result = evaluate_suppression(summary, ctx, reference_time=NOW)
    assert result.data_quality_score > 0.7


def test_quality_score_low_for_bad_data():
    summary = _make_summary()
    ctx = _make_quality_ctx(
        avg_extraction_confidence=0.1,
        newest_evidence_at=NOW - timedelta(days=14),
        total_documents=3,
        valid_documents=1,
        failed_documents=2,
    )
    result = evaluate_suppression(summary, ctx, reference_time=NOW)
    assert result.data_quality_score < 0.3


# ---------------------------------------------------------------------------
# Custom config
# ---------------------------------------------------------------------------


def test_custom_config_stricter_thresholds():
    summary = _make_summary()
    ctx = _make_quality_ctx(avg_extraction_confidence=0.5)
    strict = SuppressionConfig(min_avg_extraction_confidence=0.6)
    result = evaluate_suppression(summary, ctx, config=strict, reference_time=NOW)
    assert result.suppressed is True
    assert SuppressionReason.LOW_DATA_CONFIDENCE in result.reasons


def test_custom_config_relaxed_thresholds():
    summary = _make_summary()
    ctx = _make_quality_ctx(avg_extraction_confidence=0.3)
    relaxed = SuppressionConfig(min_avg_extraction_confidence=0.2)
    result = evaluate_suppression(summary, ctx, config=relaxed, reference_time=NOW)
    assert SuppressionReason.LOW_DATA_CONFIDENCE not in result.reasons