tests/test_comprehensive.py

"""Comprehensive tests for every core module.

Runs with:  python -m tests.test_comprehensive
"""
import os
import sys
import tempfile

# Ensure stdout handles Unicode on Windows.
if sys.stdout.encoding and sys.stdout.encoding.lower() not in ("utf-8", "utf8"):
    sys.stdout.reconfigure(encoding="utf-8", errors="replace")

sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))

from core.config import Settings, get_settings, PROVIDER_ENDPOINTS, DEFAULT_MODELS
from core.schemas import (CrawledPage, BrandBrain, Brief, FactPack,
                           GuardrailReport, Variant, GenerationResult)
from core.llm import LLMClient, _extract_json, LLMError
from core import mock_backend
from core import crawler
from core import brand_brain as bb_mod
from core import brief as brief_mod
from core import fact_pack as fp_mod
from core import writers as writers_mod
from core import guardrails
from core import ranker
from core import ppt as ppt_mod
from core.pipeline import ContentAgent

passed = 0
failed = 0


def check(name: str, condition: bool, detail: str = ""):
    global passed, failed
    if condition:
        passed += 1
        print(f"  ✅ {name}")
    else:
        failed += 1
        print(f"  ❌ {name}  — {detail}")


# ──────────────────────── SCHEMAS ────────────────────────
def test_schemas():
    print("\n── schemas ──")

    # CrawledPage
    p = CrawledPage(url="https://x.com", title="Home", headings=["H1"], text="body")
    check("CrawledPage fields", p.url == "https://x.com" and p.title == "Home")

    # BrandBrain
    bb = BrandBrain(product_name="Test")
    d = bb.to_dict()
    check("BrandBrain.to_dict", isinstance(d, dict) and d["product_name"] == "Test")
    check("BrandBrain defaults", bb.value_props == [] and bb.forbidden_claims == [])

    # Brief
    b = Brief(raw_input="note")
    check("Brief defaults", b.channels == ["linkedin", "instagram", "whatsapp"])
    check("Brief.to_dict", "raw_input" in b.to_dict())

    # FactPack.allowed_text
    fp = FactPack(approved_facts=["fact1"], approved_proof=["proof1"], user_update="update")
    check("FactPack.allowed_text", "fact1" in fp.allowed_text() and "update" in fp.allowed_text())

    fp_empty = FactPack()
    check("FactPack empty allowed_text", fp_empty.allowed_text() == "")

    # GuardrailReport.all_flags
    gr = GuardrailReport(claim_flags=["c1"], lint_flags=["l1"], policy_flags=["p1"])
    check("GuardrailReport.all_flags", len(gr.all_flags()) == 3)

    gr_clean = GuardrailReport()
    check("GuardrailReport clean passed", gr_clean.passed is True and gr_clean.all_flags() == [])

    # Variant
    v = Variant(channel="linkedin", text="hello")
    check("Variant defaults", v.score == 0.0 and v.guardrails.passed)

    # GenerationResult
    result = GenerationResult(brief=b, fact_pack=fp)
    check("GenerationResult", result.variants_by_channel == {} and result.brand_brain is None)


# ──────────────────────── CONFIG ────────────────────────
def test_config():
    print("\n── config ──")

    s = Settings()
    check("Settings defaults", s.provider == "groq")
    check("Settings.resolved_model", s.resolved_model() == DEFAULT_MODELS["groq"])
    check("Settings.endpoint", s.endpoint() == PROVIDER_ENDPOINTS["groq"])

    # Mock mode + no key → not live
    s_mock = Settings(mock_mode=True)
    check("mock_mode → not live", not s_mock.is_live())

    # Live mode + key → live
    s_live = Settings(mock_mode=False, api_key="test-key-1234")
    check("live + key → live", s_live.is_live())

    # to_dict masks key
    d = s_live.to_dict()
    check("to_dict masks key", d["api_key"] == "test…")

    # get_settings with overrides
    s2 = get_settings(mock_mode=True, provider="openrouter")
    check("get_settings overrides", s2.mock_mode is True and s2.provider == "openrouter")

    # Unknown provider falls back to groq defaults
    s3 = Settings(provider="unknown")
    check("unknown provider resolved_model fallback", s3.resolved_model() == DEFAULT_MODELS["groq"])
    check("unknown provider endpoint fallback", s3.endpoint() == PROVIDER_ENDPOINTS["groq"])


# ──────────────────────── LLM / JSON EXTRACT ────────────────────────
def test_llm():
    print("\n── llm ──")

    # _extract_json: plain JSON
    check("extract plain JSON", _extract_json('{"a": 1}') == {"a": 1})

    # With code fences
    check("extract fenced JSON", _extract_json('```json\n{"b": 2}\n```') == {"b": 2})

    # With surrounding text
    check("extract embedded JSON",
          _extract_json('Sure! Here is: {"c": 3} hope that helps') == {"c": 3})

    # Array
    check("extract JSON array", _extract_json('[1, 2, 3]') == [1, 2, 3])

    # Bad JSON raises
    try:
        _extract_json("not json at all")
        check("bad JSON raises", False, "should have raised")
    except LLMError:
        check("bad JSON raises", True)

    # LLMClient mock mode
    settings = Settings(mock_mode=True)
    client = LLMClient(settings)
    check("LLMClient.live is False in mock", not client.live)

    data = client.generate_json("suggest_prompts", "sys", "user", {})
    check("LLMClient mock returns data", "suggestions" in data)


# ──────────────────────── MOCK BACKEND ────────────────────────
def test_mock_backend():
    print("\n── mock_backend ──")

    tasks = ["suggest_prompts", "build_brief", "brand_brain", "fact_pack",
             "write_linkedin", "write_instagram", "write_whatsapp", "ppt_outline"]
    for task in tasks:
        ctx = {"product_name": "TestProd", "raw_input": "test", "user_update": "test", "num_variants": 2}
        result = mock_backend.respond(task, ctx)
        check(f"mock {task}", isinstance(result, dict) and len(result) > 0, f"got: {type(result)}")

    # Unknown task → empty dict
    check("mock unknown task", mock_backend.respond("nonexistent", {}) == {})


# ──────────────────────── CRAWLER ────────────────────────
def test_crawler():
    print("\n── crawler ──")

    settings = Settings(mock_mode=True)
    pages = crawler.crawl("https://example.com", settings)
    check("mock crawl returns pages", len(pages) == 3)
    check("mock crawl page has text", bool(pages[0].text))
    check("mock crawl page has URL", "example.com" in pages[0].url)

    # _same_domain
    check("_same_domain match",
          crawler._same_domain("https://a.com/foo", "https://a.com/bar"))
    check("_same_domain mismatch",
          not crawler._same_domain("https://a.com", "https://b.com"))


# ──────────────────────── BRAND BRAIN ────────────────────────
def test_brand_brain():
    print("\n── brand_brain ──")

    settings = Settings(mock_mode=True)
    client = LLMClient(settings)
    pages = crawler._mock_site("https://acme.com")
    bb = bb_mod.build_brand_brain(pages, client)
    check("brand_brain product_name", bool(bb.product_name))
    check("brand_brain value_props", len(bb.value_props) > 0)
    check("brand_brain source_urls", len(bb.source_urls) == len(pages))
    check("brand_brain forbidden_claims", len(bb.forbidden_claims) > 0)

    # _guess_name
    check("_guess_name from URL", bb_mod._guess_name(pages) == "Acme")
    check("_guess_name empty", bb_mod._guess_name([]) == "YourProduct")


# ──────────────────────── BRIEF ────────────────────────
def test_brief():
    print("\n── brief ──")

    settings = Settings(mock_mode=True)
    client = LLMClient(settings)
    bb = BrandBrain(product_name="Test", one_liner="Test product", value_props=["fast"])

    suggestions = brief_mod.suggest_prompts(bb, client)
    check("suggest_prompts returns list", isinstance(suggestions, list) and len(suggestions) >= 1)

    built = brief_mod.build_brief("shipped a feature", bb, client,
                                  channels=["linkedin"], num_variants=2)
    check("build_brief returns brief", "brief" in built and isinstance(built["brief"], Brief))
    check("build_brief suggested_prompt", bool(built["suggested_prompt"]))
    check("build_brief channels override", built["brief"].channels == ["linkedin"])
    check("build_brief num_variants", built["brief"].num_variants == 2)


# ──────────────────────── FACT PACK ────────────────────────
def test_fact_pack():
    print("\n── fact_pack ──")

    settings = Settings(mock_mode=True)
    client = LLMClient(settings)
    bb = BrandBrain(product_name="Test", one_liner="A test product",
                    value_props=["fast"], features=["auto"])
    brief = Brief(raw_input="launched today", objective="launch")

    fp = fp_mod.build_fact_pack(bb, brief, client)
    check("fact_pack approved_facts", len(fp.approved_facts) > 0)
    check("fact_pack cannot_claim", len(fp.cannot_claim) > 0)
    check("fact_pack user_update preserved", fp.user_update == "launched today")


# ──────────────────────── WRITERS ────────────────────────
def test_writers():
    print("\n── writers ──")

    settings = Settings(mock_mode=True)
    client = LLMClient(settings)
    fp = FactPack(approved_facts=["fact1"], approved_proof=["proof1"],
                  user_update="shipped a feature")
    brief = Brief(raw_input="shipped a feature", num_variants=2)

    for channel in ["linkedin", "instagram", "whatsapp"]:
        variants = writers_mod.write_channel(channel, fp, brief, client)
        check(f"write_{channel} returns variants",
              isinstance(variants, list) and len(variants) == 2)
        check(f"write_{channel} variant has text", bool(variants[0].text))
        check(f"write_{channel} variant.channel", variants[0].channel == channel)


# ──────────────────────── GUARDRAILS ────────────────────────
def test_guardrails():
    print("\n── guardrails ──")

    fp = FactPack(approved_facts=["We integrate with Slack"])

    # Clean variant
    clean = Variant(channel="linkedin", text="We integrate with Slack for better collaboration.")
    report = guardrails.evaluate(clean, fp)
    check("clean variant passes", report.passed)

    # Unsupported claims
    bad_claims = Variant(channel="linkedin",
                         text="We have 10000 users and are the #1 fastest guaranteed platform.")
    report2 = guardrails.evaluate(bad_claims, fp)
    check("bad claims flagged", not report2.passed)
    check("claim_flags populated", len(report2.claim_flags) > 0)

    # Specific claim patterns
    for text, expected_claims in [
        ("50% faster", ["50%"]),
        ("guaranteed results", ["guaranteed"]),
        ("world's best", ["world's", "best"]),
        ("HIPAA compliant", ["compliant", "hipaa"]),
        ("$2M ARR", ["arr"]),
    ]:
        v = Variant(channel="linkedin", text=text)
        flags = guardrails.check_claims(v, fp)
        check(f"claim pattern: '{text}'", len(flags) > 0, f"flags={flags}")

    # Lint checks
    long_v = Variant(channel="whatsapp", text="x" * 800)
    lint_flags = guardrails.check_lints(long_v)
    check("lint: too long whatsapp", any("Too long" in f for f in lint_flags))

    hashtag_v = Variant(channel="linkedin", text="hello",
                        hashtags=["#" + str(i) for i in range(20)])
    lint2 = guardrails.check_lints(hashtag_v)
    check("lint: too many hashtags", any("hashtags" in f.lower() for f in lint2))

    empty_v = Variant(channel="linkedin", text="   ")
    lint3 = guardrails.check_lints(empty_v)
    check("lint: empty content", any("Empty" in f for f in lint3))

    # Policy filter
    bad_policy = Variant(channel="linkedin", text="This is a scam product")
    policy_flags = guardrails.check_policy(bad_policy)
    check("policy: banned term", len(policy_flags) > 0)

    clean_policy = Variant(channel="linkedin", text="Great product for teams")
    check("policy: clean text", len(guardrails.check_policy(clean_policy)) == 0)


# ──────────────────────── RANKER ────────────────────────
def test_ranker():
    print("\n── ranker ──")

    fp = FactPack(approved_facts=["Ships integrations", "Built for lean teams"])

    # Passed guardrails + CTA + good length → high score
    good = Variant(channel="linkedin",
                   text="Ships integrations for lean teams. " * 5,
                   cta="Try it now",
                   guardrails=GuardrailReport(passed=True))
    score = ranker.score_variant(good, fp)
    check("good variant score > 3", score > 3.0, f"score={score}")

    # Failed guardrails → low score
    bad = Variant(channel="linkedin",
                  text="This is a game changer. Revolutionary and next level.",
                  guardrails=GuardrailReport(passed=False, claim_flags=["flag1", "flag2"]))
    bad_score = ranker.score_variant(bad, fp)
    check("bad variant score < good", bad_score < score)

    # Generic phrases penalized
    generic = Variant(channel="linkedin",
                      text="This is a game changer and cutting edge next level tool " * 3,
                      guardrails=GuardrailReport(passed=True))
    generic_score = ranker.score_variant(generic, fp)
    non_generic = Variant(channel="linkedin",
                          text="Our integrations connect with Slack and Notion for faster workflows " * 3,
                          guardrails=GuardrailReport(passed=True))
    non_generic_score = ranker.score_variant(non_generic, fp)
    check("generic penalized vs non-generic", generic_score < non_generic_score,
          f"generic={generic_score}, non_generic={non_generic_score}")

    # Ranking sorts correctly
    variants = [bad, good, generic, non_generic]
    ranked = ranker.rank(variants, fp)
    check("rank returns sorted desc", ranked[0].score >= ranked[-1].score)

    # Very short text → low specificity
    short = Variant(channel="linkedin", text="Hello",
                    guardrails=GuardrailReport(passed=True))
    check("short text specificity = 0.3",
          ranker._specificity(short.text) == 0.3)


# ──────────────────────── PPT ────────────────────────
def test_ppt():
    print("\n── ppt ──")

    settings = Settings(mock_mode=True)
    client = LLMClient(settings)
    fp = FactPack(approved_facts=["fact1"], approved_proof=["proof1"],
                  user_update="shipped integrations")
    brief = Brief(raw_input="shipped integrations", objective="daily_update")

    outline = ppt_mod.build_outline(fp, brief, client)
    check("ppt outline has title", "title" in outline)
    check("ppt outline has slides", len(outline.get("slides", [])) > 0)

    out_dir = tempfile.mkdtemp()
    out_path = os.path.join(out_dir, "test_deck.pptx")
    result_path = ppt_mod.render_pptx(outline, out_path)
    check("pptx file created", os.path.exists(result_path))
    check("pptx file non-empty", os.path.getsize(result_path) > 0)

    # Render with empty outline
    empty_outline = {"title": "Empty", "subtitle": "", "slides": []}
    empty_path = os.path.join(out_dir, "empty_deck.pptx")
    ppt_mod.render_pptx(empty_outline, empty_path)
    check("empty outline renders", os.path.exists(empty_path))


# ──────────────────────── PIPELINE (INTEGRATION) ────────────────────────
def test_pipeline():
    print("\n── pipeline (integration) ──")

    settings = get_settings(mock_mode=True)
    agent = ContentAgent(settings)

    # Full flow
    brand = agent.ingest_website("https://testproduct.io")
    check("pipeline: ingest", bool(brand.product_name))

    suggestions = agent.suggest_prompts(brand)
    check("pipeline: suggestions", len(suggestions) >= 1)

    built = agent.build_brief("we added dark mode", brand,
                              channels=["linkedin", "whatsapp"], num_variants=2)
    brief = built["brief"]
    check("pipeline: brief channels", brief.channels == ["linkedin", "whatsapp"])

    result = agent.generate(brief, brand)
    check("pipeline: generate has channels",
          set(result.variants_by_channel.keys()) == {"linkedin", "whatsapp"})
    for ch, variants in result.variants_by_channel.items():
        check(f"pipeline: {ch} variants count", len(variants) == 2)
        check(f"pipeline: {ch} ranked (scores desc)",
              variants[0].score >= variants[-1].score)

    # PPT export
    out = os.path.join(tempfile.mkdtemp(), "pipeline_deck.pptx")
    path = agent.export_ppt(result.fact_pack, brief, out)
    check("pipeline: pptx exists", os.path.exists(path) and os.path.getsize(path) > 0)


# ──────────────────────── RUN ALL ────────────────────────
def main():
    test_schemas()
    test_config()
    test_llm()
    test_mock_backend()
    test_crawler()
    test_brand_brain()
    test_brief()
    test_fact_pack()
    test_writers()
    test_guardrails()
    test_ranker()
    test_ppt()
    test_pipeline()

    print(f"\n{'='*50}")
    print(f"  Results: {passed} passed, {failed} failed")
    print(f"{'='*50}")
    if failed:
        sys.exit(1)
    else:
        print("  ALL COMPREHENSIVE TESTS PASSED ✅")


if __name__ == "__main__":
    main()