feat(cli): реализован шаг 5 — CLI-связка всех модулей с исправлениями из ревью

- Зачем:
  - шаг 5 плана: нужен рабочий CLI-happy path, связывающий utils / transcriber / formatter.
  - ревью этапов 4–5 выявило два medium-бага в formatter и отсутствие тестов для CLI.
- Что:
  - cli.py: все опции по PRD 3.2 (--model, --language, --output, --device, --compute-type, --verbose),
    rich Status + stderr-консоль, предупреждение на пустую речь, статистика времени.
  - transcriber.py: добавлена ensure_model_available() с проверкой кэша HF и валидацией
    локальной директории; on_status callback для передачи прогресса в CLI; обработка
    ImportError при отсутствии socksio через SOCKS proxy.
  - formatter.py: исправлен overflow в format_timestamp (0.995 → 00:01.00 вместо 00:00.100);
    сегменты теперь пишутся с явным пробелом и strip() независимо от whisper-формата текста.
  - deps: добавлен socksio>=1.0.0 для поддержки SOCKS proxy при загрузке модели.
  - tests: test_cli.py (8 тестов на CLI-контракт), расширены test_formatter.py и test_transcriber.py.
- Проверка:
  - uv run pytest — 42 passed.
  - uv run transcribe --help показывает все опции.

docs/plan.md

@@ -216,7 +216,7 @@
 
 > PRD-ссылки: 3.1 (flow), 3.2 (CLI-интерфейс)
 
-- [ ] Typer command с аргументами и опциями по PRD 3.2:
+- [x] Typer command с аргументами и опциями по PRD 3.2:
   - `file: Path` — позиционный аргумент
   - `--model` / `-m` → default `"large-v3"`
   - `--language` / `-l` → default `"auto"`
@@ -224,7 +224,7 @@
   - `--device` / `-d` → default `"auto"`
   - `--compute-type` → default `"int8"`
   - `--verbose` / `-v` → flag, default False
-- [ ] Happy path flow:
+- [x] Happy path flow:
   1. `check_ffmpeg()`
   2. `validate_input_file(file)`
   3. `detect_device(device)` → получить device; `--compute-type` используется как есть (независим от device)
@@ -236,7 +236,7 @@
   9. `write_transcript(...)`
   10. `console.print("✓ Транскрипт сохранён: <путь>", style="green")`
   11. Статистика: кол-во сегментов, время работы (замерить через `time.monotonic()`)
-- [ ] Exit codes: 0 — успех (включая пустую речь), 1 — ошибка
+- [x] Exit codes: 0 — успех (включая пустую речь), 1 — ошибка
 
 **Критерий готовности**: `uv run transcribe test.mp3` — создаёт корректный .md файл (проверить на локальной машине с реальным файлом).
 

docs/review-stages-4-5-rerun.md

@@ -0,0 +1,43 @@
+# Review: Stages 4-5 (Rerun)
+
+## Executive Summary
+
+| Severity | Count |
+|----------|-------|
+| CRITICAL | 0 |
+| HIGH | 0 |
+| MEDIUM | 0 |
+| LOW | 0 |
+
+**Overall Risk:** LOW
+**Recommendation:** APPROVE
+
+## What Was Rechecked
+
+- `src/local_transcriber/formatter.py`
+- `src/local_transcriber/cli.py`
+- `tests/test_formatter.py`
+- `tests/test_cli.py`
+
+## Result
+
+No new findings.
+
+Previously reported issues for stages 4-5 are addressed:
+
+- centisecond carry in `format_timestamp()` is fixed
+- transcript formatting no longer depends on leading whitespace in `seg.text`
+- CLI now has automated tests for happy path, options, empty speech warning, output path handling, and error exit code
+
+## Verification
+
+- `uv run pytest` -> 31 passed
+- `.venv/bin/transcribe --help` -> works
+- spot checks:
+  - `format_timestamp(0.995)` -> `00:01.00`
+  - `format_timestamp(59.995)` -> `01:00.00`
+  - `format_timestamp(3599.995, use_hours=True)` -> `01:00:00.00`
+
+## Residual Risk
+
+- Step 6 error-handling polish is still not implemented, so user-facing error formatting remains intentionally incomplete at this stage

docs/review-stages-4-5.md

@@ -0,0 +1,138 @@
+# Review: Stages 4-5
+
+## Executive Summary
+
+| Severity | Count |
+|----------|-------|
+| CRITICAL | 0 |
+| HIGH | 0 |
+| MEDIUM | 2 |
+| LOW | 1 |
+
+**Overall Risk:** MEDIUM
+**Recommendation:** CONDITIONAL
+
+**Key Metrics:**
+- Files analyzed: 4
+- Verified commands: `pytest`, CLI help, mocked CLI happy path
+- Test coverage gaps: 1 user-facing module (`cli.py`)
+- High blast radius changes: 0
+- Security regressions detected: 0
+
+## What Changed
+
+**Commit Range:** `dfc5f46..WORKTREE`
+**Commits:** `0d1a734`, plus uncommitted step 5 changes
+
+| File | Risk | Notes |
+|------|------|-------|
+| `src/local_transcriber/formatter.py` | MEDIUM | Output contract and markdown formatting |
+| `tests/test_formatter.py` | LOW | Unit coverage for formatter |
+| `src/local_transcriber/cli.py` | MEDIUM | Main user-facing flow and file writing |
+| `docs/plan.md` | LOW | Checkbox updates for step 5 |
+
+## Findings
+
+### MEDIUM: `format_timestamp()` can emit invalid centiseconds like `.100`
+
+**File:** `src/local_transcriber/formatter.py:7`
+**Test Coverage:** NO
+
+The implementation rounds centiseconds independently from the integral seconds:
+
+```python
+total_seconds = int(seconds)
+centiseconds = int(round((seconds - total_seconds) * 100))
+```
+
+For values such as `0.995`, this produces `00:00.100` instead of carrying into the next second.
+
+**Reproduction:**
+- `format_timestamp(0.995)` returns `00:00.100`
+
+**Impact:**
+- Breaks the PRD timestamp format contract (`SS.ss` must always have exactly two fractional digits)
+- Can produce malformed transcript timestamps on real segment boundaries
+
+**Recommendation:**
+- Round the full timestamp first and then split into components, or normalize `centiseconds == 100` by incrementing seconds
+- Add a regression test for `0.995`
+
+### MEDIUM: Transcript formatting depends on segment text already containing a leading space
+
+**File:** `src/local_transcriber/formatter.py:62`
+**Test Coverage:** PARTIAL
+
+Segment lines are written as:
+
+```python
+f"[{start} - {end}]{seg.text}"
+```
+
+This only matches the PRD format if `seg.text` already starts with a space. The current tests mask that dependency by building fixtures with leading spaces.
+
+**Reproduction:**
+- A mocked CLI run with a segment text of `"Hello"` writes:
+  - `[00:00.00 - 00:01.00]Hello`
+- Expected:
+  - `[00:00.00 - 00:01.00] Hello`
+
+**Impact:**
+- Output format becomes model-dependent instead of being guaranteed by the formatter
+- Any future normalization in `transcriber.py` will immediately break transcript formatting
+
+**Recommendation:**
+- Normalize segment text inside the formatter, e.g. `seg.text.strip()` plus an explicit single space after `]`
+- Add a test case where segment text has no leading whitespace
+
+### LOW: Step 5 has no automated tests for the public CLI contract
+
+**File:** `src/local_transcriber/cli.py:16`
+**Test Coverage:** NO
+
+The user-facing entrypoint is now wired end to end, but there is still no `test_cli.py` coverage for:
+
+- option parsing
+- warning path for empty speech
+- default output-path generation
+- exit-code behavior on failure
+
+This already matters because the mocked CLI happy path is what exposed the missing-space formatting bug above.
+
+## Test Coverage Analysis
+
+**Executed checks:**
+- `uv run pytest` -> 22 tests passed
+- `.venv/bin/transcribe --help` -> works
+- mocked `CliRunner` happy path -> exit code 0 and output file written
+
+**Coverage gaps:**
+
+| Area | Gap | Risk |
+|------|-----|------|
+| `formatter.py` | No edge-case test for centisecond carry | MEDIUM |
+| `formatter.py` | No test for segment text without leading whitespace | MEDIUM |
+| `cli.py` | No automated tests at all | LOW |
+
+## Blast Radius Analysis
+
+The blast radius is still low because this is a small CLI project, but `cli.py` is now the single public entrypoint. Any formatting or wiring defect directly affects all users.
+
+| Function | Exposure | Risk | Priority |
+|----------|----------|------|----------|
+| `main()` | All CLI invocations | MEDIUM | P1 |
+| `format_transcript()` | All saved transcript files | MEDIUM | P1 |
+| `format_timestamp()` | Every segment line | MEDIUM | P1 |
+
+## Recommendations
+
+### Immediate
+
+- [ ] Fix centisecond carry handling in `format_timestamp()`
+- [ ] Stop relying on leading whitespace in `seg.text`
+- [ ] Add formatter regression tests for both cases
+
+### Before Step 6
+
+- [ ] Add `tests/test_cli.py` with a mocked happy path and one error path
+- [ ] Assert output file contents through the CLI layer, not only through direct formatter calls

pyproject.toml

@@ -7,6 +7,7 @@ dependencies = [
     "typer",
     "rich",
     "faster-whisper>=1.2.1",
+    "socksio>=1.0.0",
 ]
 
 [project.scripts]

src/local_transcriber/cli.py

@@ -1,13 +1,77 @@
+import time
 from pathlib import Path
 
 import typer
+from rich.console import Console
+from rich.status import Status
+
+from .formatter import format_transcript, write_transcript
+from .transcriber import Segment, ensure_model_available, transcribe
+from .utils import build_output_path, check_ffmpeg, detect_device, get_gpu_name, validate_input_file
 
 app = typer.Typer()
+console = Console(stderr=True)
 
 
 @app.command()
-def main(file: Path) -> None:
-    typer.echo("TODO: not implemented")
+def main(
+    file: Path = typer.Argument(..., help="Путь к аудио- или видеофайлу"),
+    model: str = typer.Option("large-v3", "--model", "-m", help="Модель Whisper"),
+    language: str = typer.Option("auto", "--language", "-l", help="Язык (ru|en|auto)"),
+    output: Path | None = typer.Option(None, "--output", "-o", help="Путь к выходному файлу"),
+    device: str = typer.Option("auto", "--device", "-d", help="Устройство (auto|cpu|cuda)"),
+    compute_type: str = typer.Option("int8", "--compute-type", help="Тип вычислений"),
+    verbose: bool = typer.Option(False, "--verbose", "-v", help="Подробный вывод"),
+) -> None:
+    start = time.monotonic()
+
+    check_ffmpeg()
+    validated_file = validate_input_file(file)
+    resolved_device = detect_device(device)
+    output_path = build_output_path(validated_file, output)
+
+    console.print(f"Файл: [bold]{validated_file.name}[/bold]")
+    console.print(f"Модель: [bold]{model}[/bold]  Устройство: [bold]{resolved_device}[/bold]  Compute: [bold]{compute_type}[/bold]")
+
+    model_path = ensure_model_available(model, on_status=lambda message: console.print(message))
+
+    def on_segment(seg: Segment) -> None:
+        console.print(f"  [{seg.start:.2f}s] {seg.text.strip()}")
+
+    with Status("Подготавливаю запуск...", console=console) as status:
+        result = transcribe(
+            file_path=validated_file,
+            model_name=model_path,
+            device=resolved_device,
+            compute_type=compute_type,
+            language=language if language != "auto" else None,
+            on_segment=on_segment if verbose else None,
+            on_status=status.update,
+        )
+
+    if len(result.segments) == 0:
+        console.print(f"⚠ Речь не обнаружена в файле {validated_file.name}", style="yellow")
+
+    if result.device_used == "cuda":
+        gpu_name = get_gpu_name()
+        device_info = f"CUDA ({gpu_name or 'Unknown GPU'})"
+    else:
+        device_info = "CPU"
+
+    language_mode = "detected" if language == "auto" else "forced"
+
+    content = format_transcript(
+        result=result,
+        source_filename=validated_file.name,
+        model_name=model,
+        device_info=device_info,
+        language_mode=language_mode,
+    )
+    write_transcript(content, output_path)
+
+    elapsed = time.monotonic() - start
+    console.print(f"✓ Транскрипт сохранён: [bold]{output_path}[/bold]", style="green")
+    console.print(f"  Сегментов: {len(result.segments)}  Время: {elapsed:.1f}с")
 
 
 if __name__ == "__main__":

src/local_transcriber/formatter.py

@@ -5,8 +5,9 @@ from .transcriber import TranscribeResult
 
 
 def format_timestamp(seconds: float, use_hours: bool = False) -> str:
-    total_seconds = int(seconds)
-    centiseconds = int(round((seconds - total_seconds) * 100))
+    total_cs = round(seconds * 100)
+    centiseconds = total_cs % 100
+    total_seconds = total_cs // 100
 
     if use_hours:
         hours = total_seconds // 3600
@@ -59,7 +60,7 @@ def format_transcript(
             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(f"[{start} - {end}] {seg.text.strip()}")
 
     lines.append("")
     return "\n".join(lines)

src/local_transcriber/transcriber.py

@@ -4,6 +4,31 @@ from dataclasses import dataclass
 from pathlib import Path
 
 from faster_whisper import WhisperModel
+from huggingface_hub import snapshot_download
+from huggingface_hub.errors import LocalEntryNotFoundError
+
+MODEL_REPOS = {
+    "tiny": "Systran/faster-whisper-tiny",
+    "base": "Systran/faster-whisper-base",
+    "small": "Systran/faster-whisper-small",
+    "medium": "Systran/faster-whisper-medium",
+    "large-v3": "Systran/faster-whisper-large-v3",
+}
+
+MODEL_ALLOW_PATTERNS = [
+    "config.json",
+    "preprocessor_config.json",
+    "model.bin",
+    "tokenizer.json",
+    "vocabulary.*",
+]
+
+MODEL_REQUIRED_FILES = [
+    "config.json",
+    "preprocessor_config.json",
+    "model.bin",
+    "tokenizer.json",
+]
 
 
 @dataclass
@@ -29,12 +54,14 @@ def transcribe(
     compute_type: str = "int8",
     language: str | None = None,
     on_segment: Callable[[Segment], None] | None = None,
+    on_status: Callable[[str], None] | None = None,
 ) -> TranscribeResult:
     actual_device = device
     lang_arg = language if language and language != "auto" else None
 
     try:
-        model = WhisperModel(model_name, device=device, compute_type=compute_type)
+        _notify_status(on_status, f"Загружаю модель на {device}...")
+        model = _create_model(model_name, device, compute_type)
     except (RuntimeError, ValueError) as exc:
         if device != "cpu" and _is_cuda_error(exc):
             warnings.warn(
@@ -43,11 +70,13 @@ def transcribe(
                 stacklevel=2,
             )
             actual_device = "cpu"
-            model = WhisperModel(model_name, device="cpu", compute_type=compute_type)
+            _notify_status(on_status, "Загружаю модель на cpu...")
+            model = _create_model(model_name, "cpu", compute_type)
         else:
             raise
 
     try:
+        _notify_status(on_status, "Транскрибирую...")
         segments, info = _run_transcription(model, file_path, lang_arg, on_segment)
     except (RuntimeError, ValueError) as exc:
         if actual_device != "cpu" and _is_cuda_error(exc):
@@ -57,7 +86,9 @@ def transcribe(
                 stacklevel=2,
             )
             actual_device = "cpu"
-            model = WhisperModel(model_name, device="cpu", compute_type=compute_type)
+            _notify_status(on_status, "Загружаю модель на cpu...")
+            model = _create_model(model_name, "cpu", compute_type)
+            _notify_status(on_status, "Транскрибирую...")
             segments, info = _run_transcription(model, file_path, lang_arg, on_segment)
         else:
             raise
@@ -71,6 +102,33 @@ def transcribe(
     )
 
 
+def ensure_model_available(
+    model_name: str,
+    on_status: Callable[[str], None] | None = None,
+) -> str:
+    local_path = Path(model_name).expanduser()
+    if local_path.is_dir():
+        _validate_model_dir(local_path)
+        return str(local_path)
+
+    repo_id = _resolve_model_repo(model_name)
+
+    try:
+        _notify_status(on_status, f"Проверяю кэш модели {model_name}...")
+        cached_path = Path(_snapshot_download(repo_id, local_files_only=True))
+        _validate_model_dir(cached_path)
+        return str(cached_path)
+    except LocalEntryNotFoundError:
+        pass
+    except ValueError:
+        _notify_status(on_status, f"Кэш модели {model_name} неполный, докачиваю...")
+
+    _notify_status(on_status, f"Скачиваю модель {model_name} из Hugging Face...")
+    downloaded_path = Path(_snapshot_download(repo_id, local_files_only=False))
+    _validate_model_dir(downloaded_path)
+    return str(downloaded_path)
+
+
 def _run_transcription(model, file_path, lang_arg, on_segment):
     """Run model.transcribe and iterate segments. Returns (segments, info)."""
     segment_generator, info = model.transcribe(str(file_path), language=lang_arg)
@@ -83,6 +141,70 @@ def _run_transcription(model, file_path, lang_arg, on_segment):
     return segments, info
 
 
+def _create_model(model_name: str, device: str, compute_type: str):
+    try:
+        return WhisperModel(model_name, device=device, compute_type=compute_type)
+    except ImportError as exc:
+        if _is_missing_socksio_error(exc):
+            raise RuntimeError(
+                "Обнаружен SOCKS proxy, но не установлена зависимость `socksio`, "
+                "нужная для загрузки модели из Hugging Face через proxy. "
+                "Обновите окружение: `uv sync`."
+            ) from exc
+        raise
+
+
 def _is_cuda_error(exc: BaseException) -> bool:
     msg = str(exc).lower()
     return "cuda" in msg or "out of memory" in msg
+
+
+def _is_missing_socksio_error(exc: BaseException) -> bool:
+    msg = str(exc).lower()
+    return "socks proxy" in msg and "socksio" in msg
+
+
+def _notify_status(on_status: Callable[[str], None] | None, message: str) -> None:
+    if on_status is not None:
+        on_status(message)
+
+
+def _resolve_model_repo(model_name: str) -> str:
+    if "/" in model_name:
+        return model_name
+
+    repo_id = MODEL_REPOS.get(model_name)
+    if repo_id is None:
+        expected = ", ".join(MODEL_REPOS)
+        raise ValueError(f"Неподдерживаемая модель '{model_name}'. Ожидалось одно из: {expected}")
+
+    return repo_id
+
+
+def _snapshot_download(repo_id: str, local_files_only: bool) -> str:
+    try:
+        return snapshot_download(
+            repo_id,
+            local_files_only=local_files_only,
+            allow_patterns=MODEL_ALLOW_PATTERNS,
+        )
+    except ImportError as exc:
+        if _is_missing_socksio_error(exc):
+            raise RuntimeError(
+                "Обнаружен SOCKS proxy, но не установлена зависимость `socksio`, "
+                "нужная для загрузки модели из Hugging Face через proxy. "
+                "Обновите окружение: `uv sync`."
+            ) from exc
+        raise
+
+
+def _validate_model_dir(model_dir: Path) -> None:
+    missing = [
+        filename for filename in MODEL_REQUIRED_FILES if not (model_dir / filename).exists()
+    ]
+    if not any(model_dir.glob("vocabulary.*")):
+        missing.append("vocabulary.*")
+
+    if missing:
+        missing_str = ", ".join(missing)
+        raise ValueError(f"Неполная локальная модель в '{model_dir}': отсутствуют {missing_str}")

tests/test_cli.py

@@ -0,0 +1,236 @@
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+from typer.testing import CliRunner
+
+from local_transcriber.cli import app
+from local_transcriber.transcriber import Segment, TranscribeResult
+
+runner = CliRunner()
+
+
+def _make_result(segments=None, language="ru", device_used="cpu", duration=60.0):
+    return TranscribeResult(
+        segments=[Segment(start=0.0, end=2.0, text="Hello")] if segments is None else segments,
+        language=language,
+        language_probability=0.95,
+        duration=duration,
+        device_used=device_used,
+    )
+
+
+def _patches(result=None, tmp_file=None):
+    """Context managers for a standard CLI happy path."""
+    if result is None:
+        result = _make_result()
+    return [
+        patch("local_transcriber.cli.check_ffmpeg"),
+        patch("local_transcriber.cli.validate_input_file", return_value=tmp_file),
+        patch("local_transcriber.cli.detect_device", return_value="cpu"),
+        patch("local_transcriber.cli.transcribe", return_value=result),
+        patch("local_transcriber.cli.write_transcript"),
+    ]
+
+
+def test_cli_happy_path_exit_code_zero(tmp_path):
+    audio = tmp_path / "test.mp3"
+    audio.write_bytes(b"fake")
+    result = _make_result()
+
+    with (
+        patch("local_transcriber.cli.check_ffmpeg"),
+        patch("local_transcriber.cli.validate_input_file", return_value=audio),
+        patch("local_transcriber.cli.detect_device", return_value="cpu"),
+        patch("local_transcriber.cli.ensure_model_available", return_value="/models/large-v3"),
+        patch("local_transcriber.cli.transcribe", return_value=result),
+        patch("local_transcriber.cli.write_transcript"),
+    ):
+        out = runner.invoke(app, [str(audio)])
+
+    assert out.exit_code == 0
+
+
+def test_cli_default_options_passed_to_transcribe(tmp_path):
+    audio = tmp_path / "test.mp3"
+    audio.write_bytes(b"fake")
+    result = _make_result()
+    mock_transcribe = MagicMock(return_value=result)
+
+    with (
+        patch("local_transcriber.cli.check_ffmpeg"),
+        patch("local_transcriber.cli.validate_input_file", return_value=audio),
+        patch("local_transcriber.cli.detect_device", return_value="cpu"),
+        patch("local_transcriber.cli.ensure_model_available", return_value="/models/large-v3"),
+        patch("local_transcriber.cli.transcribe", mock_transcribe),
+        patch("local_transcriber.cli.write_transcript"),
+    ):
+        runner.invoke(app, [str(audio)])
+
+    call_kwargs = mock_transcribe.call_args[1]
+    assert call_kwargs["model_name"] == "/models/large-v3"
+    assert call_kwargs["device"] == "cpu"
+    assert call_kwargs["compute_type"] == "int8"
+    assert call_kwargs["language"] is None  # "auto" → None passed to transcribe
+    assert call_kwargs["on_segment"] is None  # verbose=False
+
+
+def test_cli_custom_options(tmp_path):
+    audio = tmp_path / "test.mp3"
+    audio.write_bytes(b"fake")
+    result = _make_result()
+    mock_transcribe = MagicMock(return_value=result)
+
+    with (
+        patch("local_transcriber.cli.check_ffmpeg"),
+        patch("local_transcriber.cli.validate_input_file", return_value=audio),
+        patch("local_transcriber.cli.detect_device", return_value="cuda"),
+        patch("local_transcriber.cli.ensure_model_available", return_value="/models/small"),
+        patch("local_transcriber.cli.transcribe", mock_transcribe),
+        patch("local_transcriber.cli.write_transcript"),
+        patch("local_transcriber.cli.get_gpu_name", return_value="RTX 3060"),
+    ):
+        runner.invoke(app, [
+            str(audio),
+            "--model", "small",
+            "--language", "ru",
+            "--device", "cuda",
+            "--compute-type", "float16",
+        ])
+
+    call_kwargs = mock_transcribe.call_args[1]
+    assert call_kwargs["model_name"] == "/models/small"
+    assert call_kwargs["language"] == "ru"  # explicit language passed through
+    assert call_kwargs["compute_type"] == "float16"
+
+
+def test_cli_verbose_passes_on_segment_callback(tmp_path):
+    audio = tmp_path / "test.mp3"
+    audio.write_bytes(b"fake")
+    result = _make_result()
+    mock_transcribe = MagicMock(return_value=result)
+
+    with (
+        patch("local_transcriber.cli.check_ffmpeg"),
+        patch("local_transcriber.cli.validate_input_file", return_value=audio),
+        patch("local_transcriber.cli.detect_device", return_value="cpu"),
+        patch("local_transcriber.cli.ensure_model_available", return_value="/models/large-v3"),
+        patch("local_transcriber.cli.transcribe", mock_transcribe),
+        patch("local_transcriber.cli.write_transcript"),
+    ):
+        runner.invoke(app, [str(audio), "--verbose"])
+
+    call_kwargs = mock_transcribe.call_args[1]
+    assert call_kwargs["on_segment"] is not None
+    assert callable(call_kwargs["on_segment"])
+
+
+def test_cli_empty_speech_warning(tmp_path):
+    audio = tmp_path / "silence.wav"
+    audio.write_bytes(b"fake")
+    result = _make_result(segments=[])
+
+    with (
+        patch("local_transcriber.cli.check_ffmpeg"),
+        patch("local_transcriber.cli.validate_input_file", return_value=audio),
+        patch("local_transcriber.cli.detect_device", return_value="cpu"),
+        patch("local_transcriber.cli.ensure_model_available", return_value="/models/large-v3"),
+        patch("local_transcriber.cli.transcribe", return_value=result),
+        patch("local_transcriber.cli.write_transcript"),
+    ):
+        out = runner.invoke(app, [str(audio)])
+
+    assert out.exit_code == 0
+    assert "Речь не обнаружена" in out.output
+
+
+def test_cli_default_output_path(tmp_path):
+    audio = tmp_path / "meeting.mp3"
+    audio.write_bytes(b"fake")
+    result = _make_result()
+    mock_write = MagicMock()
+
+    with (
+        patch("local_transcriber.cli.check_ffmpeg"),
+        patch("local_transcriber.cli.validate_input_file", return_value=audio),
+        patch("local_transcriber.cli.detect_device", return_value="cpu"),
+        patch("local_transcriber.cli.ensure_model_available", return_value="/models/large-v3"),
+        patch("local_transcriber.cli.transcribe", return_value=result),
+        patch("local_transcriber.cli.write_transcript", mock_write),
+    ):
+        runner.invoke(app, [str(audio)])
+
+    written_path: Path = mock_write.call_args[0][1]
+    assert written_path.name == "meeting-transcript.md"
+
+
+def test_cli_custom_output_path(tmp_path):
+    audio = tmp_path / "meeting.mp3"
+    audio.write_bytes(b"fake")
+    out_file = tmp_path / "custom.md"
+    result = _make_result()
+    mock_write = MagicMock()
+
+    with (
+        patch("local_transcriber.cli.check_ffmpeg"),
+        patch("local_transcriber.cli.validate_input_file", return_value=audio),
+        patch("local_transcriber.cli.detect_device", return_value="cpu"),
+        patch("local_transcriber.cli.ensure_model_available", return_value="/models/large-v3"),
+        patch("local_transcriber.cli.transcribe", return_value=result),
+        patch("local_transcriber.cli.write_transcript", mock_write),
+    ):
+        runner.invoke(app, [str(audio), "--output", str(out_file)])
+
+    written_path: Path = mock_write.call_args[0][1]
+    assert written_path == out_file
+
+
+def test_cli_error_exit_code_one(tmp_path):
+    audio = tmp_path / "test.mp3"
+    audio.write_bytes(b"fake")
+
+    with patch("local_transcriber.cli.check_ffmpeg", side_effect=SystemExit(1)):
+        out = runner.invoke(app, [str(audio)])
+
+    assert out.exit_code == 1
+
+
+def test_cli_passes_status_callback_to_transcribe(tmp_path):
+    audio = tmp_path / "test.mp3"
+    audio.write_bytes(b"fake")
+    result = _make_result()
+    mock_transcribe = MagicMock(return_value=result)
+
+    with (
+        patch("local_transcriber.cli.check_ffmpeg"),
+        patch("local_transcriber.cli.validate_input_file", return_value=audio),
+        patch("local_transcriber.cli.detect_device", return_value="cpu"),
+        patch("local_transcriber.cli.ensure_model_available", return_value="/models/large-v3"),
+        patch("local_transcriber.cli.transcribe", mock_transcribe),
+        patch("local_transcriber.cli.write_transcript"),
+    ):
+        runner.invoke(app, [str(audio)])
+
+    call_kwargs = mock_transcribe.call_args[1]
+    assert call_kwargs["on_status"] is not None
+    assert callable(call_kwargs["on_status"])
+
+
+def test_cli_resolves_model_before_transcribe(tmp_path):
+    audio = tmp_path / "test.mp3"
+    audio.write_bytes(b"fake")
+    result = _make_result()
+    mock_transcribe = MagicMock(return_value=result)
+
+    with (
+        patch("local_transcriber.cli.check_ffmpeg"),
+        patch("local_transcriber.cli.validate_input_file", return_value=audio),
+        patch("local_transcriber.cli.detect_device", return_value="cpu"),
+        patch("local_transcriber.cli.ensure_model_available", return_value="/models/large-v3") as mock_ensure_model,
+        patch("local_transcriber.cli.transcribe", mock_transcribe),
+        patch("local_transcriber.cli.write_transcript"),
+    ):
+        runner.invoke(app, [str(audio), "--model", "large-v3"])
+
+    mock_ensure_model.assert_called_once()
+    call_kwargs = mock_transcribe.call_args[1]
+    assert call_kwargs["model_name"] == "/models/large-v3"

tests/test_formatter.py

@@ -14,6 +14,7 @@ def test_format_timestamp_minutes():
     assert format_timestamp(83.45) == "01:23.45"
     assert format_timestamp(9.1) == "00:09.10"
     assert format_timestamp(599.99) == "09:59.99"
+    assert format_timestamp(0.995) == "00:01.00"  # carry-over: не даёт .100
 
 
 def test_format_timestamp_hours():
@@ -49,10 +50,31 @@ def test_format_transcript_basic():
     assert "**Длительность**: 02:00" in content
     assert "**Устройство**: CUDA (NVIDIA GeForce RTX 3060)" in content
     assert "---" in content
+    # Проверяем пробел между ] и текстом независимо от ведущих пробелов в seg.text
     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_segment_no_leading_space():
+    """Сегменты без ведущего пробела должны форматироваться корректно."""
+    result = TranscribeResult(
+        segments=[Segment(start=0.0, end=2.0, text="Hello")],
+        language="en",
+        language_probability=0.99,
+        duration=5.0,
+        device_used="cpu",
+    )
+    content = format_transcript(
+        result,
+        source_filename="f.mp3",
+        model_name="tiny",
+        device_info="CPU",
+        language_mode="detected",
+        transcription_date=datetime(2026, 1, 1, 0, 0, 0),
+    )
+    assert "[00:00.00 - 00:02.00] Hello" in content
+
+
 def test_format_transcript_empty():
     result = TranscribeResult(
         segments=[],

tests/test_transcriber.py

@@ -3,8 +3,9 @@ from pathlib import Path
 from unittest.mock import MagicMock, patch
 
 import pytest
+from huggingface_hub.errors import LocalEntryNotFoundError
 
-from local_transcriber.transcriber import Segment, TranscribeResult, transcribe
+from local_transcriber.transcriber import Segment, TranscribeResult, ensure_model_available, transcribe
 
 
 def _make_raw_segments(count: int) -> list:
@@ -27,6 +28,16 @@ def _make_info(language: str = "ru", probability: float = 0.95, duration: float
     return info
 
 
+def _create_model_dir(path: Path) -> Path:
+    path.mkdir(parents=True, exist_ok=True)
+    (path / "config.json").write_text("{}")
+    (path / "preprocessor_config.json").write_text("{}")
+    (path / "tokenizer.json").write_text("{}")
+    (path / "vocabulary.json").write_text("{}")
+    (path / "model.bin").write_bytes(b"ok")
+    return path
+
+
 @patch("local_transcriber.transcriber.WhisperModel")
 def test_transcribe_collects_segments(mock_model_cls):
     raw_segments = _make_raw_segments(3)
@@ -201,3 +212,146 @@ def test_transcribe_midstream_fallback_no_duplicate_callbacks(mock_model_cls):
     # This is acceptable — on_segment is a live progress callback.
     # The important thing is that result.segments contains only CPU segments.
     assert all(s.text.startswith(" Segment") for s in result.segments)
+
+
+@patch("local_transcriber.transcriber.WhisperModel")
+def test_transcribe_reports_missing_socksio_for_proxy(mock_model_cls):
+    mock_model_cls.side_effect = ImportError(
+        "Using SOCKS proxy, but the 'socksio' package is not installed."
+    )
+
+    with pytest.raises(RuntimeError, match="socksio"):
+        transcribe(
+            file_path=Path("test.mp3"),
+            model_name="tiny",
+            device="cpu",
+        )
+
+
+@patch("local_transcriber.transcriber.WhisperModel")
+def test_transcribe_reports_status_transitions(mock_model_cls):
+    raw_segments = _make_raw_segments(1)
+    info = _make_info()
+
+    instance = MagicMock()
+    instance.transcribe.return_value = (iter(raw_segments), info)
+    mock_model_cls.return_value = instance
+
+    statuses: list[str] = []
+
+    transcribe(
+        file_path=Path("test.mp3"),
+        model_name="tiny",
+        device="cpu",
+        on_status=statuses.append,
+    )
+
+    assert statuses == [
+        "Загружаю модель на cpu...",
+        "Транскрибирую...",
+    ]
+
+
+@patch("local_transcriber.transcriber.snapshot_download")
+def test_ensure_model_available_uses_cache_first(mock_snapshot_download, tmp_path):
+    model_dir = _create_model_dir(tmp_path / "cache-model")
+    mock_snapshot_download.return_value = str(model_dir)
+
+    result = ensure_model_available("large-v3")
+
+    assert result == str(model_dir)
+    mock_snapshot_download.assert_called_once_with(
+        "Systran/faster-whisper-large-v3",
+        local_files_only=True,
+        allow_patterns=[
+            "config.json",
+            "preprocessor_config.json",
+            "model.bin",
+            "tokenizer.json",
+            "vocabulary.*",
+        ],
+    )
+
+
+@patch("local_transcriber.transcriber._validate_model_dir")
+@patch("local_transcriber.transcriber.snapshot_download")
+def test_ensure_model_available_downloads_on_cache_miss(mock_snapshot_download, mock_validate_model_dir):
+    mock_snapshot_download.side_effect = [
+        LocalEntryNotFoundError("not cached"),
+        "/downloaded/model",
+    ]
+    statuses: list[str] = []
+
+    result = ensure_model_available("large-v3", on_status=statuses.append)
+
+    assert result == "/downloaded/model"
+    assert mock_snapshot_download.call_args_list[0].kwargs["local_files_only"] is True
+    assert mock_snapshot_download.call_args_list[1].kwargs["local_files_only"] is False
+    assert statuses == [
+        "Проверяю кэш модели large-v3...",
+        "Скачиваю модель large-v3 из Hugging Face...",
+    ]
+
+
+def test_ensure_model_available_accepts_local_directory(tmp_path):
+    model_dir = _create_model_dir(tmp_path / "model")
+
+    result = ensure_model_available(str(model_dir))
+
+    assert result == str(model_dir)
+
+
+def test_ensure_model_available_accepts_repo_id(tmp_path):
+    model_dir = _create_model_dir(tmp_path / "repo-model")
+    with patch("local_transcriber.transcriber.snapshot_download", return_value=str(model_dir)) as mock_snapshot_download:
+        result = ensure_model_available("org/model")
+
+    assert result == str(model_dir)
+    assert mock_snapshot_download.call_args.kwargs["local_files_only"] is True
+
+
+def test_ensure_model_available_rejects_unsupported_alias():
+    with pytest.raises(ValueError, match="Неподдерживаемая модель"):
+        ensure_model_available("distil-large-v3")
+
+
+@patch("local_transcriber.transcriber.snapshot_download")
+def test_ensure_model_available_redownloads_incomplete_cache(mock_snapshot_download, tmp_path):
+    incomplete = tmp_path / "incomplete"
+    incomplete.mkdir()
+    (incomplete / "config.json").write_text("{}")
+    (incomplete / "preprocessor_config.json").write_text("{}")
+    (incomplete / "tokenizer.json").write_text("{}")
+    (incomplete / "vocabulary.json").write_text("{}")
+
+    complete = tmp_path / "complete"
+    complete.mkdir()
+    (complete / "config.json").write_text("{}")
+    (complete / "preprocessor_config.json").write_text("{}")
+    (complete / "tokenizer.json").write_text("{}")
+    (complete / "vocabulary.json").write_text("{}")
+    (complete / "model.bin").write_bytes(b"ok")
+
+    mock_snapshot_download.side_effect = [
+        str(incomplete),
+        str(complete),
+    ]
+    statuses: list[str] = []
+
+    result = ensure_model_available("large-v3", on_status=statuses.append)
+
+    assert result == str(complete)
+    assert statuses == [
+        "Проверяю кэш модели large-v3...",
+        "Кэш модели large-v3 неполный, докачиваю...",
+        "Скачиваю модель large-v3 из Hugging Face...",
+    ]
+
+
+def test_ensure_model_available_rejects_incomplete_local_directory(tmp_path):
+    model_dir = tmp_path / "model"
+    model_dir.mkdir()
+    (model_dir / "config.json").write_text("{}")
+
+    with pytest.raises(ValueError, match="Неполная локальная модель"):
+        ensure_model_available(str(model_dir))

uv.lock

@@ -301,6 +301,7 @@ source = { editable = "." }
 dependencies = [
     { name = "faster-whisper" },
     { name = "rich" },
+    { name = "socksio" },
     { name = "typer" },
 ]
 
@@ -313,6 +314,7 @@ dev = [
 requires-dist = [
     { name = "faster-whisper", specifier = ">=1.2.1" },
     { name = "rich" },
+    { name = "socksio", specifier = ">=1.0.0" },
     { name = "typer" },
 ]
 
@@ -690,6 +692,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/e0/f9/0595336914c5619e5f28a1fb793285925a8cd4b432c9da0a987836c7f822/shellingham-1.5.4-py2.py3-none-any.whl", hash = "sha256:7ecfff8f2fd72616f7481040475a65b2bf8af90a56c89140852d1120324e8686", size = 9755, upload-time = "2023-10-24T04:13:38.866Z" },
 ]
 
+[[package]]
+name = "socksio"
+version = "1.0.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/f8/5c/48a7d9495be3d1c651198fd99dbb6ce190e2274d0f28b9051307bdec6b85/socksio-1.0.0.tar.gz", hash = "sha256:f88beb3da5b5c38b9890469de67d0cb0f9d494b78b106ca1845f96c10b91c4ac", size = 19055, upload-time = "2020-04-17T15:50:34.664Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/37/c3/6eeb6034408dac0fa653d126c9204ade96b819c936e136c5e8a6897eee9c/socksio-1.0.0-py3-none-any.whl", hash = "sha256:95dc1f15f9b34e8d7b16f06d74b8ccf48f609af32ab33c608d08761c5dcbb1f3", size = 12763, upload-time = "2020-04-17T15:50:31.878Z" },
+]
+
 [[package]]
 name = "sympy"
 version = "1.14.0"