"""MockLLMProvider: every analysis task's response schema must come back
valid and genuinely reflect the evidence passed in - never an empty
placeholder unrelated to the input."""

from __future__ import annotations

import pytest

from app.analysis.llm.mock import MockLLMProvider
from app.prompts.alert_summarization import AlertSummary
from app.prompts.base import build_user_prompt
from app.prompts.change_significance import ChangeSignificanceAssessment
from app.prompts.extraction import ExtractionResult
from app.prompts.relevance import RelevanceAssessment
from app.prompts.report_generation import ReportContent
from app.prompts.synthesis import SynthesisResult

provider = MockLLMProvider()


@pytest.mark.asyncio
async def test_relevance_matches_focus_keyword():
    prompt = build_user_prompt(
        "assess",
        {
            "monitoring_focus": "electric vehicle manufacturing expansion",
            "document_text": "The company announced a new manufacturing facility for electric vehicles.",
        },
    )
    result = await provider.generate_structured("system", prompt, RelevanceAssessment)
    assert result.is_relevant is True
    assert result.matches_focus is True


@pytest.mark.asyncio
async def test_extraction_pulls_first_sentence_as_signal():
    prompt = build_user_prompt(
        "extract", {"document_text": "Acme Corp opened a new facility. It will employ 200 people."}
    )
    result = await provider.generate_structured("system", prompt, ExtractionResult)
    assert len(result.signals) == 1
    assert "Acme Corp opened a new facility" in result.signals[0].description


@pytest.mark.asyncio
async def test_synthesis_requires_multiple_signals():
    prompt_one = build_user_prompt("synthesize", {"signals": [{"description": "a"}]})
    result_one = await provider.generate_structured("system", prompt_one, SynthesisResult)
    assert result_one.conclusions == []

    prompt_two = build_user_prompt(
        "synthesize", {"signals": [{"description": "a"}, {"description": "b"}]}
    )
    result_two = await provider.generate_structured("system", prompt_two, SynthesisResult)
    assert len(result_two.conclusions) == 1
    assert result_two.conclusions[0].source_count == 2


@pytest.mark.asyncio
async def test_report_reflects_evidence_counts():
    prompt = build_user_prompt(
        "report",
        {
            "company_profile": {"name": "Acme Corp"},
            "source_documents": [
                {
                    "id": "d1",
                    "title": "Job posting",
                    "url": "https://x.com/1",
                    "source_type": "job_posting",
                    "retrieved_date": "2026-01-01",
                }
            ],
            "detected_changes": [
                {
                    "id": "c1",
                    "summary": "New job posting detected",
                    "change_type": "new_document",
                    "severity": "medium",
                    "confidence_score": 0.6,
                    "created_at": "2026-01-01",
                }
            ],
            "sources_that_failed_to_collect": ["Broken Source"],
        },
    )
    result = await provider.generate_structured("system", prompt, ReportContent)
    assert "Acme Corp" in result.executive_summary
    assert "1" in result.executive_summary  # document/change counts mentioned
    assert len(result.recent_developments) == 1
    assert len(result.hiring_signals) == 1
    assert any("Broken Source" in u for u in result.unknowns_and_missing_data)


@pytest.mark.asyncio
async def test_report_grounds_overview_in_discovered_profile_even_with_no_documents():
    """The bug this covers: a report generated before any monitoring run had
    collected evidence used to come back near-empty even though the company's
    discovered profile (from onboarding) had real data. That profile data
    must now ground company_overview/market_positioning."""
    prompt = build_user_prompt(
        "report",
        {
            "company_profile": {
                "name": "Stripe",
                "description": "Stripe builds economic infrastructure for the internet.",
                "industry": "Financial infrastructure",
                "headquarters": "South San Francisco, California",
                "competitors": ["PayPal"],
                "aliases": [],
            },
            "source_documents": [],
            "detected_changes": [],
            "sources_that_failed_to_collect": [],
        },
    )
    result = await provider.generate_structured("system", prompt, ReportContent)
    assert "economic infrastructure" in result.company_overview
    assert "South San Francisco" in result.company_overview
    assert "PayPal" in result.market_positioning


@pytest.mark.asyncio
async def test_change_significance_reflects_deterministic_severity():
    prompt = build_user_prompt(
        "assess",
        {
            "change_type": "leadership_change",
            "deterministic_severity": "high",
            "deterministic_confidence": 0.8,
        },
    )
    result = await provider.generate_structured("system", prompt, ChangeSignificanceAssessment)
    assert result.is_meaningful is True
    assert result.should_notify is True
    assert "high" in result.why_it_matters


@pytest.mark.asyncio
async def test_alert_summary_title_under_100_chars():
    prompt = build_user_prompt(
        "summarize",
        {
            "company_name": "Acme Corp",
            "change_type": "price_change",
            "severity": "medium",
            "confidence": 0.6,
        },
    )
    result = await provider.generate_structured("system", prompt, AlertSummary)
    assert len(result.title) <= 100
    assert "Acme Corp" in result.title


@pytest.mark.asyncio
async def test_generate_text_does_not_crash():
    prompt = build_user_prompt("do something", {"a": 1})
    text = await provider.generate_text("system", prompt)
    assert isinstance(text, str)
    assert len(text) > 0