import copy
import json

from tools.rag.build_chunks import TOKEN_COUNTER
from tools.rag.llm_context import compact_rag_payload


def response(text='Unikátny dôkaz o práci osoby. '):
    return {'status': 'ok', 'query': 'Práca?', 'engine': 'test', 'strategies': ['test'],
        'complete': None, 'truncated': False, 'truncation_reasons': [], 'index_version': {'version': 'test'},
        'instructions': ['Cituj iba dodané zdroje.'], 'answer_format': {},
        'sources': [{'source_id': 'S1', 'document_path': 'pages/eva/README.md', 'title': 'Eva',
                     'source_url': 'https://example.test/eva', 'text': text,
                     'query_evidence': {'huge_diagnostic': text * 10},
                     'retrieval': {'rank': 1}}], 'context': text}


def test_compact_payload_contains_evidence_once_and_preserves_sources():
    original = response()
    before = copy.deepcopy(original)
    compact = compact_rag_payload(original)
    assert original == before
    assert json.dumps(compact, ensure_ascii=False).count(original['sources'][0]['text']) == 1
    assert 'text' not in compact['sources'][0] and 'query_evidence' not in compact['sources'][0]
    assert compact['sources'][0]['source_url'] == 'https://example.test/eva'
    assert compact['index_version'] == original['index_version']
    assert compact_rag_payload(compact) == compact


def test_entire_tool_json_is_bounded_and_marks_omitted_evidence():
    original = response('Veľmi dlhý dôkaz 🧑🏽‍💻. ' * 2000)
    original['sources'].append({**original['sources'][0], 'source_id': 'S2', 'text': 'Ďalší dôkaz'})
    compact = compact_rag_payload(original, max_tokens=700)
    used = TOKEN_COUNTER.count(json.dumps(compact, ensure_ascii=False, separators=(',', ':')))
    assert used <= 700 and compact['context_budget']['used_tokens'] == used
    assert compact['complete'] is False and compact['truncated'] is True
    assert 'llm_context_budget' in compact['truncation_reasons']
    assert 'KONIEC ZDROJA S1' in compact['context']
    assert compact['source_count'] == 1 and '[OBSAH ZDROJA OREZANÝ]' in compact['context']
    assert json.loads(compact['context'].splitlines()[1])['truncated'] is True


def test_diagnostics_are_explicit_and_never_forwarded_by_projection():
    verbose = compact_rag_payload(response(), include_diagnostics=True)
    assert verbose['diagnostics']['sources'][0]['query_evidence']
    projected = compact_rag_payload(verbose)
    assert 'diagnostics' not in projected
    assert 'huge_diagnostic' not in json.dumps(projected)


def test_http_serialization_stays_within_budget_and_diagnostics_are_opt_in(monkeypatch):
    from fastapi.testclient import TestClient
    import app.main as main
    import app.routes as routes
    original = response('Dlhý dôkaz 🐈 ' * 1800)
    monkeypatch.setattr(main, 'embed_query', lambda _: None)
    monkeypatch.setattr(routes, 'build_rag_context', lambda *a, **kw: original)
    monkeypatch.setenv('RAG_LLM_MAX_TOKENS', '700')
    with TestClient(main.app) as client:
        headers = {'X-API-Key': 'a' * 64}
        compact = client.post('/rag', json={'query': 'Práca?'}, headers=headers)
        assert compact.status_code == 200
        assert TOKEN_COUNTER.count(compact.text) <= 700
        assert 'diagnostics' not in compact.json()
        assert 'text' not in compact.json()['sources'][0]
        verbose = client.post('/rag', json={'query': 'Práca?', 'include_diagnostics': True}, headers=headers)
        assert verbose.status_code == 200 and verbose.json()['diagnostics'] == original