Files
CenaForge/clipper/tests/test_english_source.py
MateuusandClaude Opus 5.5 cb8450399d Clipador: Equipe de IA (papéis com IA própria), juiz de contexto e linha do tempo JSON
- roles.py: varredura, analista, editor e juiz, cada um com provedor, modelo, raciocínio
  (Codex), liga/desliga (o editor não desliga) e instruções do administrador somadas às
  padrão; gravado em settings `judge.roles`, com migração automática da configuração antiga
  (provedor/modelo único vira o de todos; o modelo do analista é respeitado). A reserva vale
  por papel.
- Juiz de contexto (só texto): vê só o que vai no corte (+20 s antes/15 s depois marcados
  out=1), escreve o que um espectador frio entende e responde sim/não; context_judge.decide
  decide pelo código: aprova, ajusta o trecho e confere uma vez de novo, frase de contexto
  na tela (vira o texto de abertura do projeto) ou descarta com motivo em português
  (forçável). Corte manual só recebe aviso.
- timeline.py: linha do tempo JSON curta (fala com falante, chat, rosto, volume e a frase
  de tela que a varredura/analista escrevem ao ver os quadros, em scene_notes); mosaico 4x2
  (media.mosaic) no lugar dos quadros soltos; perfil do jogo enxuto. O formato antigo fica
  atrás da opção timelineJson.
- Medição (F0): ai_calls por chamada (papel, IA, tamanho do prompt/imagens, tokens quando a
  CLI informa: codex --json, claude e gemini), avaliação do dono (owner_label: bom/ruim/sem
  contexto) e GET /v1/judge/report (taxa de aprovação, decisões do juiz, concordância).
- Banco v17 (v16 fica reservada às expressões, como marcador sem efeito).
- Testes: test_roles.py e test_context_judge.py; os testes bloqueiam qualquer CLI de IA real.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-18 04:36:25 -03:00

172 lines
9.7 KiB
Python

"""Fonte em inglês (18/11/2026, doc/clipador/IDIOMAS.md): juízes em inglês com o português do painel na mesma
resposta, risada/gíria do chat em inglês, fala do corte traduzida por frase e legenda em português de desenvolvimento
(desligada por padrão)."""
import json
from pathlib import Path
from app import editor, editor_render, layout
from app.judge import (ANALYSIS_SCHEMA_EN, JUDGE_SYSTEM_EN, SCAN_SCHEMA_EN, VERDICT_SCHEMA_EN, Judge, Verdict,
analysis_block, build_prompt, lang_key)
from app.scoring import REACTION_PATTERNS, ChatMessage, Word
from app.subtitles import phrases
from .test_editor import validate
from .test_render_and_duplicates import add_clip
from .test_sources_and_api import env, seed_session # noqa: F401 - fixture reaproveitada
class EnglishJudge(Judge):
"""Juiz falso: registra o idioma pedido e devolve o bloco `pt`, como os schemas em inglês exigem."""
def __init__(self):
super().__init__("codex", "fake", Path("."))
self.seen: dict = {}
self.translated: list[list[str]] = []
def analyze(self, prompt, images, lang="pt"):
self.seen["analyze"] = (lang, prompt)
return {"situation": "jogada", "what_happened": "He robbed the bank.", "screen": "", "speech": "",
"chat_meaning": "", "confidence": 0.9, "action_start_s": 0, "reaction_end_s": 0, "hook": "NO WAY",
"pt": {"what_happened": "Ele roubou o banco.", "screen": "", "speech": "", "chat_meaning": "", "hook": "NÃO ACREDITO"}}
def judge(self, prompt, start_s, end_s, images=None, lang="pt", **_limits):
self.seen["judge"] = (lang, prompt)
return Verdict(True, 8.0, start_s, end_s, "He robbed the bank solo", "Insane solo heist.", "great play", "fake",
("heist",), pt={"title": "Ele roubou o banco sozinho", "caption": "Assalto solo insano.",
"reason": "jogada ótima"})
def translate_lines(self, lines):
self.translated.append(list(lines))
return [f"PT: {text}" for text in lines]
def _english_session(engine) -> int:
sid = seed_session(engine)
engine.db.run("UPDATE sessions SET language = 'en-US' WHERE id = ?", sid)
words = [[281.0, 281.3, "No"], [281.3, 281.6, "way"], [281.6, 282.0, "bro."], [284.0, 284.4, "Let's"], [284.4, 284.9, "go!"]]
engine.db.run("UPDATE segments SET words_json = '[]' WHERE session_id = ?", sid)
engine.db.run("UPDATE segments SET words_json = ? WHERE session_id = ? AND idx = 9", json.dumps(words), sid)
return sid
def test_lang_key():
assert lang_key("en-US") == "en" and lang_key("en") == "en"
assert lang_key("pt-BR") == "pt" and lang_key("es-419") == "pt" and lang_key("") == "pt" and lang_key(None) == "pt"
def test_english_schemas_require_the_portuguese_block():
assert ANALYSIS_SCHEMA_EN["required"][-1] == "pt" and VERDICT_SCHEMA_EN["required"][-1] == "pt"
assert "summary_pt" in SCAN_SCHEMA_EN["required"]
assert "pt" in SCAN_SCHEMA_EN["properties"]["moments"]["items"]["required"]
# Os CLIs exigem todas as propriedades em `required` (strict).
for schema in (ANALYSIS_SCHEMA_EN, VERDICT_SCHEMA_EN, SCAN_SCHEMA_EN):
assert set(schema["properties"]) == set(schema["required"])
assert "LMAO" in JUDGE_SYSTEM_EN and "💀" in JUDGE_SYSTEM_EN
def test_english_prompt_labels_and_analysis_block_without_pt():
prompt = build_prompt("Tfue", "twitch", 100, 160, 150, [Word(140, 141, "bro")], [ChatMessage(151, "LMAO")], {},
"GTA 6 Waiting Room", lang="en")
assert "Channel: Tfue" in prompt and "Transcript:" in prompt and "Chat sample:" in prompt
assert "Transcrição" not in prompt and "Canal:" not in prompt
block = analysis_block({"situation": "jogada", "pt": {"what_happened": "x"}}, "en")
assert block.startswith("ANALYST's analysis") and '"pt"' not in block
def test_english_chat_laughter_and_slang_count_as_reaction():
def tags(text):
return {tag for tag, (pattern, _) in REACTION_PATTERNS.items() if pattern.search(text)}
assert "risada" in tags("LMAO") and "risada" in tags("lol") and "risada" in tags("💀💀") and "risada" in tags("im dead")
assert "risada" in tags("😭😭") and "risada" not in tags("lollipop")
assert "susto" in tags("no way") and "susto" in tags("bruhhh") and "susto" in tags("nahhh")
assert "hype" in tags("LETS GOOO") and "hype" in tags("sheesh") and "hype" in tags("clutch")
assert "clip" in tags("clip that")
def test_phrases_split_on_sentence_end_and_pause():
words = [Word(0, 0.3, "No"), Word(0.3, 0.6, "way"), Word(0.6, 0.9, "bro."), Word(0.9, 1.2, "This"), Word(1.2, 1.5, "is"),
Word(1.5, 1.8, "crazy"), Word(1.8, 2.1, "man."), Word(3.5, 3.8, "Okay")]
assert [t for _, _, t in phrases(words)] == ["No way bro. This is crazy man.", "Okay"]
assert phrases([]) == []
def test_english_session_judges_in_english_and_stores_portuguese(env): # noqa: F811
client, engine = env
engine.judge = EnglishJudge()
sid = _english_session(engine)
add_clip(engine, sid, 1, "candidate", 285, 270, 340)
engine.judge_clip(engine.db.get("SELECT * FROM clips WHERE id = 1"))
assert engine.judge.seen["analyze"][0] == "en" and engine.judge.seen["judge"][0] == "en"
# Linha do tempo JSON (padrão desde a Equipe de IA, EQUIPE-IA.md): legenda das chaves e a fala dentro dela.
editor_prompt = engine.judge.seen["judge"][1]
assert "Timeline (JSON Lines" in editor_prompt and "ANALYST's analysis" in editor_prompt
assert '"sp":"No way bro.' in editor_prompt
clip = client.get("/v1/clips/1", headers=_auth()).json()
assert clip["title"] == "He robbed the bank solo" # o vídeo/publicação: inglês
assert clip["judge"]["pt"]["title"] == "Ele roubou o banco sozinho" # o painel: português
assert clip["judge"]["analysis"]["pt"]["what_happened"] == "Ele roubou o banco."
lines = clip["translation"]["lines"]
assert [x["text"] for x in lines] == ["No way bro.", "Let's go!"]
assert lines[0]["pt"] == "PT: No way bro." and clip["translation"]["language"] == "pt-BR"
# Tradução que já cobre o trecho é reaproveitada (sem chamar a IA de novo).
engine.translate_clip(1)
assert len(engine.judge.translated) == 1
def test_portuguese_session_has_no_translation(env): # noqa: F811
client, engine = env
engine.judge = EnglishJudge()
sid = seed_session(engine)
add_clip(engine, sid, 1, "candidate", 300, 280, 350)
engine.intel.opts["timelineJson"] = False # formato antigo (transcrição + chat) continua atrás da opção
engine.judge_clip(engine.db.get("SELECT * FROM clips WHERE id = 1"))
assert engine.judge.seen["judge"][0] == "pt" and "Transcrição:" in engine.judge.seen["judge"][1]
assert client.get("/v1/clips/1", headers=_auth()).json()["translation"] is None
assert client.post("/v1/clips/1/translate", headers=_auth()).status_code == 409
def test_dev_translation_line_only_when_enabled(tmp_path):
clip = {"session_id": 1, "start_s": 10.0, "end_s": 75.0, "peak_s": 50.0, "title": "He robbed the bank", "source": "chat"}
words = [Word(20, 20.4, "No"), Word(20.4, 20.8, "way"), Word(20.8, 21.2, "bro.")]
lines = [{"s": 20.0, "e": 21.2, "text": "No way bro.", "pt": "Não acredito, mano."}]
for enabled in (False, True):
cfg = layout.merge({"caption": {"translation": enabled}})
p = validate(editor.auto_project(clip, None, words, [], 100.0, cfg), tmp_path)
ass = editor_render.build_project_ass(p, editor.effective_settings(cfg, p), "twitch.tv/tfue", translation=lines)
has = "Não acredito, mano." in ass and "Style: Trans," in ass
assert has is enabled
# Sem tradução disponível nada muda, mesmo ligado.
cfg = layout.merge({"caption": {"translation": True}})
p = validate(editor.auto_project(clip, None, words, [], 100.0, cfg), tmp_path)
assert "Style: Trans," not in editor_render.build_project_ass(p, editor.effective_settings(cfg, p), "x")
assert layout.DEFAULTS["caption"]["translation"] is False
assert layout.validate({"caption": {"translation": True}})["caption"]["translation"] is True
def _auth():
from .test_sources_and_api import auth
return auth()
def test_voice_language_maps_audience_codes_to_whisper_codes():
from app.worker import voice_language
assert voice_language("en-US") == "en" and voice_language("es-419") == "es" and voice_language("pt-BR") == "pt-BR"
assert voice_language("en-GB") == "en" and voice_language("") == "pt-BR"
def test_dev_layers_never_reach_the_final_render(tmp_path):
"""Contrato `dev` (18/11/2026): camadas de desenvolvimento só no vídeo de revisão; `dev=False` = vídeo final limpo."""
assert {k: layout.DEFAULTS["dev"][k] for k in ("translation", "dub")} == {"translation": False, "dub": False}
assert layout.validate({"dev": {"translation": True, "dub": True, "x": 1}})["dev"] == {"translation": True, "dub": True}
cfg = layout.merge({"dev": {"translation": True, "dub": True}})
assert layout.dev_flags(cfg) == {"translation": True, "dub": True}
assert layout.dev_flags(cfg, dev=False) == {"translation": False, "dub": False}
assert layout.dev_flags(layout.merge({"caption": {"translation": True}}))["translation"] is True # legado
clip = {"session_id": 1, "start_s": 10.0, "end_s": 75.0, "peak_s": 50.0, "title": "x", "source": "chat"}
words = [Word(20, 20.4, "No"), Word(20.4, 20.8, "way")]
lines = [{"s": 20.0, "e": 20.8, "text": "No way", "pt": "Não acredito"}]
p = validate(editor.auto_project(clip, None, words, [], 100.0, cfg), tmp_path)
eff = editor.effective_settings(cfg, p)
assert "Não acredito" in editor_render.build_project_ass(p, eff, "x", translation=lines)
assert "Não acredito" not in editor_render.build_project_ass(p, eff, "x", translation=lines, dev=False)