fixed a bug:translate maybe timeout

sentence_api/tests/test_ai_tutor.py

@@ -1,6 +1,10 @@
 import pytest
 
-from sentence_api.ai_tutor import AITutor, AITutorError
+import json
+import re
+import threading
+
+from sentence_api.ai_tutor import AITutor, AITutorError, _parse_translation_payload
 
 
 class FakeResponse:
@@ -73,15 +77,13 @@ def test_ai_tutor_rejects_unexpected_response(monkeypatch):
         tutor.explain("Hello.")
 
 
-from sentence_api.ai_tutor import _parse_translation_payload
-
 
 def test_ai_tutor_translates_full_document(monkeypatch):
     requests_payloads = []
 
     def fake_post(endpoint, headers, json, timeout):
         requests_payloads.append((endpoint, headers, json, timeout))
-        return FakeResponse(body={"choices": [{"message": {"content": '["你好。", "见到你很高兴。"]'}}]})
+        return FakeResponse(body={"choices": [{"message": {"content": '{"0":"你好。","1":"见到你很高兴。"}'}}]})
 
     monkeypatch.setattr("sentence_api.ai_tutor.requests.post", fake_post)
     tutor = AITutor("https://ai.test/v1", "secret", "test-model", 1.5)
@@ -94,11 +96,87 @@ def test_ai_tutor_translates_full_document(monkeypatch):
     assert translated == {0: "你好。", 1: "见到你很高兴。"}
     _endpoint, _headers, payload, timeout = requests_payloads[0]
     user_prompt = payload["messages"][1]["content"]
-    assert "Hello." in user_prompt
-    assert "Nice to meet you." in user_prompt
-    assert timeout == 90.0
+    assert "[0] Hello." in user_prompt
+    assert "[1] Nice to meet you." in user_prompt
+    assert timeout == 45.0
+
+
+def _translation_response_from_prompt(payload):
+    user_prompt = payload["messages"][1]["content"]
+    section = user_prompt.split("待翻译句子：", 1)[1]
+    indexes = [int(match) for match in re.findall(r"^\[(\d+)\]", section, re.M)]
+    body = {str(index): f"中文{index}" for index in indexes}
+    return FakeResponse(
+        body={"choices": [{"message": {"content": json.dumps(body, ensure_ascii=False)}}]}
+    )
+
+
+def test_ai_tutor_translates_long_input_in_batches(monkeypatch):
+    requests_payloads = []
+    lock = threading.Lock()
+
+    def fake_post(endpoint, headers, json, timeout):
+        with lock:
+            requests_payloads.append(json)
+        return _translation_response_from_prompt(json)
+
+    monkeypatch.setattr("sentence_api.ai_tutor.requests.post", fake_post)
+    tutor = AITutor("https://ai.test/v1", "secret", "test-model", 1)
+    sentences = [(index, f"This is sentence {index} with enough text.") for index in range(30)]
+
+    translated = tutor.translate_sentences(sentences, course_title="Long talk")
+
+    assert translated == {index: f"中文{index}" for index in range(30)}
+    assert len(requests_payloads) >= 3
+    first_prompt = requests_payloads[0]["messages"][1]["content"]
+    assert "前文" in first_prompt or "后文" in first_prompt or "待翻译句子：" in first_prompt
+
+
+def test_ai_tutor_skips_cached_sentences_and_persists_windows(monkeypatch):
+    persisted = []
+
+    def fake_post(endpoint, headers, json, timeout):
+        return _translation_response_from_prompt(json)
+
+    monkeypatch.setattr("sentence_api.ai_tutor.requests.post", fake_post)
+    tutor = AITutor("https://ai.test/v1", "secret", "test-model", 1)
+    translated = tutor.translate_sentences(
+        [(0, "Hello."), (1, "Thanks."), (2, "Bye.")],
+        existing={0: "你好。"},
+        persist=persisted.append,
+        course_title="Greetings",
+    )
+
+    assert translated[0] == "你好。"
+    assert translated[1] == "中文1"
+    assert translated[2] == "中文2"
+    assert persisted
+    assert 0 not in persisted[0]
+
+
+def test_ai_tutor_splits_invalid_windows_and_retries(monkeypatch):
+    calls = {"count": 0}
+
+    def fake_post(endpoint, headers, json, timeout):
+        calls["count"] += 1
+        if calls["count"] == 1:
+            return FakeResponse(body={"choices": [{"message": {"content": "not-json"}}]})
+        return _translation_response_from_prompt(json)
+
+    monkeypatch.setattr("sentence_api.ai_tutor.requests.post", fake_post)
+    tutor = AITutor("https://ai.test/v1", "secret", "test-model", 1)
+    sentences = [(index, f"This is sentence {index} with enough text.") for index in range(12)]
+
+    translated = tutor.translate_sentences(sentences)
+
+    assert translated == {index: f"中文{index}" for index in range(12)}
+    assert calls["count"] >= 3
 
 
 def test_parse_translation_payload_accepts_fenced_json_and_numbered_lines():
-    assert _parse_translation_payload('```json\n["一", "二"]\n```', 2) == ["一", "二"]
-    assert _parse_translation_payload("1. 一\n2. 二", 2) == ["一", "二"]
+    assert _parse_translation_payload('```json\n["一", "二"]\n```', 2) == {0: "一", 1: "二"}
+    assert _parse_translation_payload("1. 一\n2. 二", 2) == {0: "一", 1: "二"}
+    assert _parse_translation_payload('prefix {"12":"甲","13":"乙"} suffix', [12, 13]) == {
+        12: "甲",
+        13: "乙",
+    }

sentence_api/tests/test_api.py

@@ -268,6 +268,53 @@ def test_subtitle_translation_uses_full_document_and_is_cached(tmp_path, monkeyp
     assert requested[0][0] == [(0, "Hello."), (1, "Nice to meet you.")]
 
 
+def test_subtitle_translation_is_shared_with_other_users(tmp_path, monkeypatch):
+    client = make_client(tmp_path)
+    _ready_video(client.app.state.video_repository)
+    requested = []
+
+    def translate_sentences(sentences, **kwargs):
+        requested.append((sentences, kwargs))
+        return {index: f"中文{index}" for index, _text in sentences}
+
+    class StubAITutor:
+        pass
+
+    stub = StubAITutor()
+    stub.translate_sentences = translate_sentences
+    monkeypatch.setattr(client.app.state, "ai_tutor", stub)
+
+    first = client.post(f"/api/v1/videos/{VIDEO_HASH}/subtitle-translation")
+    assert first.status_code == 200, first.text
+
+    stored = client.get(f"/api/v1/videos/{VIDEO_HASH}/subtitle-translation")
+    assert stored.status_code == 200
+    assert stored.json() == first.json()
+
+    other = TestClient(client.app)
+    signup = other.post(
+        "/api/v1/auth/register",
+        json={"username": "viewer2", "password": "password-123"},
+    )
+    assert signup.status_code == 201, signup.text
+    other.headers.update({"Authorization": f"Bearer {signup.json()['token']}"})
+
+    shared = other.get(f"/api/v1/videos/{VIDEO_HASH}/subtitle-translation")
+    assert shared.status_code == 200
+    assert shared.json()["translations"] == [
+        {"index": 0, "zh_text": "中文0"},
+        {"index": 1, "zh_text": "中文1"},
+    ]
+    boundaries = other.get(f"/api/v1/videos/{VIDEO_HASH}/sentence-boundaries")
+    assert boundaries.status_code == 200
+    assert [item["zh_text"] for item in boundaries.json()["sentences"]] == ["中文0", "中文1"]
+
+    reused = other.post(f"/api/v1/videos/{VIDEO_HASH}/subtitle-translation")
+    assert reused.status_code == 200
+    assert reused.json() == first.json()
+    assert len(requested) == 1
+
+
 def test_oss_video_url_is_used_for_catalog_and_content(tmp_path):
     index_path = tmp_path / "boundaries.json"
     index_path.write_text("{}", encoding="utf-8")