feat(formatter): реализовано форматирование транскрипта (шаг 4)

- Зачем:
  - необходим модуль формирования markdown-файла с таймкодами и метаданными.
- Что:
  - реализованы format_timestamp (MM:SS.ss / HH:MM:SS.ss), format_transcript (шапка по PRD 3.3, авто-часы при >1ч, пустая речь), write_transcript (UTF-8).
  - добавлено 6 тестов в test_formatter.py (22 теста зелёные).
- Проверка:
  - uv run pytest tests/test_formatter.py -v (6 passed).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

docs/plan.md

@@ -187,11 +187,11 @@
 
 > PRD-ссылки: 3.3 (формат выходного файла)
 
-- [ ] `format_timestamp(seconds: float, use_hours: bool = False) -> str`:
+- [x] `format_timestamp(seconds: float, use_hours: bool = False) -> str`:
   - `False` → `"01:23.45"` (MM:SS.ss)
   - `True` → `"01:23:45.67"` (HH:MM:SS.ss)
   - Сотые — всегда 2 знака после точки
-- [ ] `format_transcript(...)`:
+- [x] `format_transcript(...)`:
   - Шапка по шаблону PRD 3.3 (заголовок, метаданные, разделитель)
   - `language_mode`: `"detected"` если CLI получил `--language auto`, `"forced"` если язык задан явно
   - В шапке: `**Язык**: {language} ({language_mode})` → например `ru (detected)` или `en (forced)`
@@ -199,9 +199,9 @@
   - Автоматически `use_hours=True` если `result.duration > 3600`
   - Сегменты: `[MM:SS.ss - MM:SS.ss] текст\n\n`
   - Если `len(result.segments) == 0` → после разделителя: `\n*Речь не обнаружена.*\n` (файл создаётся с полной шапкой метаданных; warning в stderr выводит CLI в шаге 5)
-- [ ] `write_transcript(content: str, output_path: Path)`:
+- [x] `write_transcript(content: str, output_path: Path)`:
   - `open(output_path, "w", encoding="utf-8")`
-- [ ] Тесты в `tests/test_formatter.py`:
+- [x] Тесты в `tests/test_formatter.py`:
   - `test_format_timestamp_minutes` — обычный таймкод
   - `test_format_timestamp_hours` — формат с часами
   - `test_format_transcript_basic` — проверить шапку + пару сегментов

src/local_transcriber/formatter.py

@@ -5,7 +5,28 @@ from .transcriber import TranscribeResult
 
 
 def format_timestamp(seconds: float, use_hours: bool = False) -> str:
-    raise NotImplementedError
+    total_seconds = int(seconds)
+    centiseconds = int(round((seconds - total_seconds) * 100))
+
+    if use_hours:
+        hours = total_seconds // 3600
+        minutes = (total_seconds % 3600) // 60
+        secs = total_seconds % 60
+        return f"{hours:02d}:{minutes:02d}:{secs:02d}.{centiseconds:02d}"
+
+    minutes = total_seconds // 60
+    secs = total_seconds % 60
+    return f"{minutes:02d}:{secs:02d}.{centiseconds:02d}"
+
+
+def _format_duration(seconds: float) -> str:
+    total = int(seconds)
+    h = total // 3600
+    m = (total % 3600) // 60
+    s = total % 60
+    if h > 0:
+        return f"{h:02d}:{m:02d}:{s:02d}"
+    return f"{m:02d}:{s:02d}"
 
 
 def format_transcript(
@@ -16,8 +37,34 @@ def format_transcript(
     language_mode: str,  # "detected" | "forced"
     transcription_date: datetime | None = None,  # None -> datetime.now()
 ) -> str:
-    raise NotImplementedError
+    date = transcription_date or datetime.now()
+    use_hours = result.duration > 3600
+
+    lines: list[str] = []
+    lines.append(f"# Транскрипт: {source_filename}")
+    lines.append("")
+    lines.append(f"- **Дата транскрипции**: {date.strftime('%Y-%m-%d %H:%M:%S')}")
+    lines.append(f"- **Модель**: {model_name}")
+    lines.append(f"- **Язык**: {result.language} ({language_mode})")
+    lines.append(f"- **Длительность**: {_format_duration(result.duration)}")
+    lines.append(f"- **Устройство**: {device_info}")
+    lines.append("")
+    lines.append("---")
+
+    if not result.segments:
+        lines.append("")
+        lines.append("*Речь не обнаружена.*")
+    else:
+        for seg in result.segments:
+            start = format_timestamp(seg.start, use_hours=use_hours)
+            end = format_timestamp(seg.end, use_hours=use_hours)
+            lines.append("")
+            lines.append(f"[{start} - {end}]{seg.text}")
+
+    lines.append("")
+    return "\n".join(lines)
 
 
 def write_transcript(content: str, output_path: Path) -> None:
-    raise NotImplementedError
+    with open(output_path, "w", encoding="utf-8") as f:
+        f.write(content)

tests/test_formatter.py

@@ -0,0 +1,108 @@
+from datetime import datetime
+from pathlib import Path
+
+from local_transcriber.formatter import (
+    format_timestamp,
+    format_transcript,
+    write_transcript,
+)
+from local_transcriber.transcriber import Segment, TranscribeResult
+
+
+def test_format_timestamp_minutes():
+    assert format_timestamp(0.0) == "00:00.00"
+    assert format_timestamp(83.45) == "01:23.45"
+    assert format_timestamp(9.1) == "00:09.10"
+    assert format_timestamp(599.99) == "09:59.99"
+
+
+def test_format_timestamp_hours():
+    assert format_timestamp(3723.45, use_hours=True) == "01:02:03.45"
+    assert format_timestamp(0.0, use_hours=True) == "00:00:00.00"
+    assert format_timestamp(7261.0, use_hours=True) == "02:01:01.00"
+
+
+def test_format_transcript_basic():
+    result = TranscribeResult(
+        segments=[
+            Segment(start=0.0, end=4.82, text=" Добрый день, коллеги."),
+            Segment(start=4.82, end=9.15, text=" Первый вопрос."),
+        ],
+        language="ru",
+        language_probability=0.97,
+        duration=120.0,
+        device_used="cuda",
+    )
+    content = format_transcript(
+        result,
+        source_filename="meeting.mp4",
+        model_name="large-v3",
+        device_info="CUDA (NVIDIA GeForce RTX 3060)",
+        language_mode="detected",
+        transcription_date=datetime(2026, 3, 17, 14, 30, 5),
+    )
+
+    assert "# Транскрипт: meeting.mp4" in content
+    assert "**Дата транскрипции**: 2026-03-17 14:30:05" in content
+    assert "**Модель**: large-v3" in content
+    assert "**Язык**: ru (detected)" in content
+    assert "**Длительность**: 02:00" in content
+    assert "**Устройство**: CUDA (NVIDIA GeForce RTX 3060)" in content
+    assert "---" in content
+    assert "[00:00.00 - 00:04.82] Добрый день, коллеги." in content
+    assert "[00:04.82 - 00:09.15] Первый вопрос." in content
+
+
+def test_format_transcript_empty():
+    result = TranscribeResult(
+        segments=[],
+        language="ru",
+        language_probability=0.5,
+        duration=30.0,
+        device_used="cpu",
+    )
+    content = format_transcript(
+        result,
+        source_filename="silence.wav",
+        model_name="tiny",
+        device_info="CPU",
+        language_mode="detected",
+        transcription_date=datetime(2026, 1, 1, 0, 0, 0),
+    )
+
+    assert "# Транскрипт: silence.wav" in content
+    assert "*Речь не обнаружена.*" in content
+    assert "**Модель**: tiny" in content
+
+
+def test_format_transcript_long():
+    result = TranscribeResult(
+        segments=[
+            Segment(start=0.0, end=10.5, text=" Начало."),
+            Segment(start=3700.0, end=3710.25, text=" Конец."),
+        ],
+        language="en",
+        language_probability=0.99,
+        duration=3800.0,
+        device_used="cuda",
+    )
+    content = format_transcript(
+        result,
+        source_filename="long.mp4",
+        model_name="large-v3",
+        device_info="CUDA",
+        language_mode="forced",
+        transcription_date=datetime(2026, 3, 17, 10, 0, 0),
+    )
+
+    assert "**Длительность**: 01:03:20" in content
+    assert "**Язык**: en (forced)" in content
+    # Timestamps should use hours format
+    assert "[00:00:00.00 - 00:00:10.50] Начало." in content
+    assert "[01:01:40.00 - 01:01:50.25] Конец." in content
+
+
+def test_write_transcript(tmp_path):
+    out = tmp_path / "output.md"
+    write_transcript("# Test content\n", out)
+    assert out.read_text(encoding="utf-8") == "# Test content\n"