Refactor CLI and LLM runtime handling for improved user experience

- Removed deprecated console usage in favor of structured logging functions for error handling and user prompts.
- Enhanced CLI prompts for LLM runtime and preset selection, utilizing menu helpers for better user interaction.
- Updated GPU pool scanning output with improved formatting and error indication for clarity.
- Refactored setup wizard to streamline LLM runtime and preset configuration, ensuring a more intuitive setup process.
- Improved documentation and user feedback in CLI outputs to enhance overall usability.

src/gpu_rent/cli.py

@@ -10,7 +10,6 @@ from datetime import datetime, timezone
 from typing import Optional
 
 import typer
-from rich.console import Console
 from rich.table import Table
 
 from gpu_rent import __version__
@@ -23,22 +22,15 @@ from gpu_rent.ssh_ops import interactive_ssh, run_ssh
 from gpu_rent.state import load_state, preempt_window_end
 from gpu_rent.provision import ensure_swarmui_running, seed_civitai, seed_extensions
 from gpu_rent.sync_files import pull_tree, push_tree
+from gpu_rent.term import console, err, log, ok, warn
 from gpu_rent.tunnel import run_tunnel
 
-if sys.platform == "win32":
-    for _stream in (sys.stdout, sys.stderr):
-        try:
-            _stream.reconfigure(encoding="utf-8", errors="replace")
-        except (AttributeError, OSError):
-            pass
-
 app = typer.Typer(
     no_args_is_help=True,
     pretty_exceptions_enable=False,
     add_completion=False,
     help="Прерываемый GPU Selectel + SwarmUI на localhost:17801. Сначала: gpu-rent setup / doctor. Ключи: docs/setup.md",
 )
-console = Console(highlight=False, legacy_windows=False)
 
 _DEBUG = False
 
@@ -54,7 +46,7 @@ def _root(
 def _die(exc: BaseException) -> None:
     if _DEBUG:
         traceback.print_exc()
-    console.print(f"[red]{exc}[/red]")
+    err(str(exc))
     raise typer.Exit(1)
 
 
@@ -368,7 +360,7 @@ def setup(
             install_watchdog=watchdog if watchdog is not None else (False if yes else None),
             confirm=None if yes and watchdog is None else confirm,
             ask=None if yes and llm is not None else ask,
-            log=lambda m: console.print(m),
+            log=log,
         )
     except Exception as exc:
         _die(exc)
@@ -412,7 +404,6 @@ def up(
         from dataclasses import replace
 
         from gpu_rent.llm_runtime import (
-            PRESET_HELP,
             append_vars_llm_runtime,
             decide_runtime,
             ensure_ollama_manifest_from_example,
@@ -438,11 +429,24 @@ def up(
 
         asked_model_preset = False
         if not yes and runtime == "none" and not llm and not ollama and not llamacpp:
-            choice = typer.prompt(
+            from gpu_rent.llm_runtime import (
-                "Поднять LLM рядом со SwarmUI? [none/ollama/llamacpp]",
+                llamacpp_preset_menu,
-                default="none",
+                llm_runtime_menu,
+                ollama_preset_menu,
             )
+            from gpu_rent.prompts import prompt_menu
+
+            def _ask(msg: str, default: str = "") -> str:
+                return typer.prompt(msg, default=default)
+
             try:
+                choice = prompt_menu(
+                    "LLM рядом со SwarmUI",
+                    llm_runtime_menu(),
+                    default="none",
+                    ask=_ask,
+                    show=log,
+                )
                 runtime = decide_runtime(
                     flag=choice, ollama_flag=False, llamacpp_flag=False, from_config="none"
                 )
@@ -452,63 +456,82 @@ def up(
                 append_vars_llm_runtime(vars_path(), runtime)
             if runtime == "ollama":
                 ensure_ollama_manifest_from_example()
-                console.print(PRESET_HELP)
+                try:
-                preset = typer.prompt("Ollama preset", default="recommended")
+                    preset = prompt_menu(
-                if preset.strip().lower() not in {"keep", "example"}:
+                        "Ollama preset",
-                    write_ollama_models_preset(
+                        ollama_preset_menu(include_keep=False),
-                        cfg.ollama_models_manifest, preset.strip().lower()
+                        default="recommended",
+                        ask=_ask,
+                        show=log,
                     )
+                except ValueError as exc:
+                    raise GpuRentError(str(exc)) from exc
+                if preset not in {"keep", "example"}:
+                    write_ollama_models_preset(cfg.ollama_models_manifest, preset)
                 asked_model_preset = True
             elif runtime == "llamacpp":
                 from gpu_rent.llm_runtime import (
-                    LLAMACPP_PRESET_HELP,
                     ensure_llamacpp_manifest_from_example,
                     write_llamacpp_models_preset,
                 )
 
                 ensure_llamacpp_manifest_from_example()
-                console.print(LLAMACPP_PRESET_HELP)
+                try:
-                preset = typer.prompt(
+                    preset = prompt_menu(
-                    "llama.cpp GGUF preset [recommended/light/stock/empty]",
+                        "llama.cpp GGUF",
-                    default="recommended",
+                        llamacpp_preset_menu(include_keep=False),
-                )
+                        default="recommended",
-                if preset.strip().lower() not in {"keep", "example"}:
+                        ask=_ask,
-                    write_llamacpp_models_preset(
+                        show=log,
-                        cfg.llamacpp_models_manifest, preset.strip().lower()
                     )
+                except ValueError as exc:
+                    raise GpuRentError(str(exc)) from exc
+                if preset not in {"keep", "example"}:
+                    write_llamacpp_models_preset(cfg.llamacpp_models_manifest, preset)
                 asked_model_preset = True
 
-        # Runtime уже в vars (напр. llamacpp) — всё равно спросить модель, default=keep.
+        # Runtime уже в vars — спросить пресет, default=keep.
-        if not yes and not asked_model_preset and runtime == "llamacpp":
+        if not yes and not asked_model_preset and runtime in {"llamacpp", "ollama"}:
             from gpu_rent.llm_runtime import (
-                LLAMACPP_PRESET_HELP,
                 ensure_llamacpp_manifest_from_example,
+                llamacpp_preset_menu,
+                ollama_preset_menu,
                 write_llamacpp_models_preset,
             )
+            from gpu_rent.prompts import prompt_menu
 
-            ensure_llamacpp_manifest_from_example()
+            def _ask2(msg: str, default: str = "") -> str:
-            console.print(LLAMACPP_PRESET_HELP)
+                return typer.prompt(msg, default=default)
-            preset = typer.prompt(
+
-                "llama.cpp GGUF preset [recommended/light/stock/empty/keep]",
+            try:
-                default="keep",
+                if runtime == "llamacpp":
-            )
+                    ensure_llamacpp_manifest_from_example()
-            key = preset.strip().lower()
+                    key = prompt_menu(
-            if key not in {"keep", "example", ""}:
+                        "llama.cpp GGUF",
-                write_llamacpp_models_preset(cfg.llamacpp_models_manifest, key)
+                        llamacpp_preset_menu(include_keep=True),
-        elif not yes and not asked_model_preset and runtime == "ollama":
+                        default="keep",
-            ensure_ollama_manifest_from_example()
+                        ask=_ask2,
-            console.print(PRESET_HELP)
+                        show=log,
-            preset = typer.prompt(
+                    )
-                "Ollama preset [recommended/light/stock/alt/empty/keep]",
+                    if key not in {"keep", "example", ""}:
-                default="keep",
+                        write_llamacpp_models_preset(cfg.llamacpp_models_manifest, key)
-            )
+                else:
-            key = preset.strip().lower()
+                    ensure_ollama_manifest_from_example()
-            if key not in {"keep", "example", ""}:
+                    key = prompt_menu(
-                write_ollama_models_preset(cfg.ollama_models_manifest, key)
+                        "Ollama preset",
+                        ollama_preset_menu(include_keep=True),
+                        default="keep",
+                        ask=_ask2,
+                        show=log,
+                    )
+                    if key not in {"keep", "example", ""}:
+                        write_ollama_models_preset(cfg.ollama_models_manifest, key)
+            except ValueError as exc:
+                raise GpuRentError(str(exc)) from exc
 
         cfg = replace(cfg, llm_runtime=runtime)
         if runtime != "none":
-            console.print(f"LLM runtime: {runtime}")
+            ok(f"LLM runtime: {runtime}")
 
         def confirm(msg: str) -> bool:
             return typer.confirm(msg)
@@ -525,7 +548,7 @@ def up(
             update=False if no_update else None,
             confirm=confirm,
             ask=None if yes else ask,
-            log=lambda m: console.print(m),
+            log=log,
         )
         if no_tunnel:
             from gpu_rent.access_card import print_access_card
@@ -550,7 +573,7 @@ def up(
             cfg,
             state.floating_ip,
             open_browser=open_browser,
-            log=lambda m: console.print(m),
+            log=log,
         )
     except GpuRentError as exc:
         _die(exc)
@@ -563,7 +586,7 @@ def stop(
     """Удалить compute и FIP, диски оставить."""
     try:
         cfg = load_config(require_auth=True)
-        cmd_stop(cfg, no_pull=no_pull, log=lambda m: console.print(m))
+        cmd_stop(cfg, no_pull=no_pull, log=log)
     except GpuRentError as exc:
         _die(exc)
 
@@ -579,7 +602,7 @@ def destroy(
         raise typer.Exit(1)
     try:
         cfg = load_config(require_auth=True)
-        cmd_stop(cfg, destroy_disks=True, no_pull=no_pull, log=lambda m: console.print(m))
+        cmd_stop(cfg, destroy_disks=True, no_pull=no_pull, log=log)
     except GpuRentError as exc:
         _die(exc)
 
@@ -636,7 +659,7 @@ def tunnel(
             cfg,
             state.floating_ip,
             open_browser=open_browser,
-            log=lambda m: console.print(m),
+            log=log,
         )
     except GpuRentError as exc:
         _die(exc)
@@ -654,9 +677,9 @@ def hold(
 
         cfg, host = _live()
         if clear:
-            clear_hold(cfg, host, log=lambda m: console.print(m))
+            clear_hold(cfg, host, log=log)
             return
-        set_hold(cfg, host, minutes=minutes, until=until, log=lambda m: console.print(m))
+        set_hold(cfg, host, minutes=minutes, until=until, log=log)
     except GpuRentError as exc:
         _die(exc)
 
@@ -666,7 +689,7 @@ def seed_models() -> None:
     """Докачать Civitai-манифест на живой диск."""
     try:
         cfg, host = _live()
-        seed_civitai(cfg, host, log=lambda m: console.print(m))
+        seed_civitai(cfg, host, log=log)
     except GpuRentError as exc:
         _die(exc)
 
@@ -677,9 +700,6 @@ def push_all() -> None:
     try:
         cfg, host = _live()
 
-        def log(msg: str) -> None:
-            console.print(msg)
-
         push_tree(cfg, host, cfg.local_models_dir, "/mnt/swarm_data/Models", log, models=True)
         push_tree(cfg, host, cfg.local_wildcards_dir, "/mnt/swarm_data/Data/Wildcards", log, models=False)
         push_tree(cfg, host, cfg.local_workflows_dir, "/mnt/swarm_data/CustomWorkflows", log, models=False)
@@ -697,7 +717,7 @@ def push_models() -> None:
             host,
             cfg.local_models_dir,
             "/mnt/swarm_data/Models",
-            lambda m: console.print(m),
+            log,
             models=True,
         )
     except GpuRentError as exc:
@@ -714,7 +734,7 @@ def pull_output_cmd() -> None:
             host,
             "/mnt/swarm_data/Output",
             cfg.local_output_dir,
-            lambda m: console.print(m),
+            log,
         )
     except GpuRentError as exc:
         _die(exc)
@@ -726,9 +746,6 @@ def seed_extensions_cmd() -> None:
     try:
         cfg, host = _live()
 
-        def log(msg: str) -> None:
-            console.print(msg)
-
         seed_extensions(cfg, host, log)
         ensure_swarmui_running(cfg, host, log, restart=True)
     except GpuRentError as exc:
@@ -768,9 +785,9 @@ def capture_root(
             host,
             dry_run=dry_run,
             kind_filter=kind,
-            log=lambda m: console.print(m),
+            log=log,
         )
-        print_report(report, lambda m: console.print(m), dry_run=dry_run)
+        print_report(report, log, dry_run=dry_run)
     except GpuRentError as exc:
         _die(exc)
 
@@ -796,11 +813,11 @@ def capture_models_cmd(
             None,
             dry_run=dry_run,
             kind_filter=kind,
-            log=lambda m: console.print(m),
+            log=log,
         )
         print_report(
             report,
-            lambda m: console.print(m),
+            log,
             dry_run=dry_run,
             show_models=True,
             show_extensions=False,
@@ -823,11 +840,11 @@ def capture_extensions_cmd(
             host,
             None,
             dry_run=dry_run,
-            log=lambda m: console.print(m),
+            log=log,
         )
         print_report(
             report,
-            lambda m: console.print(m),
+            log,
             dry_run=dry_run,
             show_models=False,
             show_extensions=True,
@@ -854,9 +871,9 @@ def capture_all_cmd(
             host,
             dry_run=dry_run,
             kind_filter=kind,
-            log=lambda m: console.print(m),
+            log=log,
         )
-        print_report(report, lambda m: console.print(m), dry_run=dry_run)
+        print_report(report, log, dry_run=dry_run)
     except GpuRentError as exc:
         _die(exc)
 
@@ -868,7 +885,7 @@ def resize_data(gb: int = typer.Option(..., "--gb", help="Новый разме
         from gpu_rent.resize import resize_data_volume
 
         cfg = load_config(require_auth=True)
-        resize_data_volume(cfg, gb, log=lambda m: console.print(m))
+        resize_data_volume(cfg, gb, log=log)
     except GpuRentError as exc:
         _die(exc)
 
@@ -905,9 +922,9 @@ def watchdog_install(
     try:
         from gpu_rent.local_watchdog import install_watchdog
 
-        install_watchdog(interval_minutes=interval, log=lambda m: console.print(m))
+        install_watchdog(interval_minutes=interval, log=log)
     except Exception as exc:
-        console.print(f"[red]install fail:[/red] {exc}")
+        err(f"install fail: {exc}")
         raise typer.Exit(1) from exc
 
 
@@ -917,9 +934,9 @@ def watchdog_uninstall() -> None:
     try:
         from gpu_rent.local_watchdog import uninstall_watchdog
 
-        uninstall_watchdog(log=lambda m: console.print(m))
+        uninstall_watchdog(log=log)
     except Exception as exc:
-        console.print(f"[red]uninstall fail:[/red] {exc}")
+        err(f"uninstall fail: {exc}")
         raise typer.Exit(1) from exc
 
 
@@ -946,7 +963,7 @@ def watchdog_tick(
     try:
         from gpu_rent.local_watchdog import run_tick
 
-        decision = run_tick(dry_run=dry_run, log=lambda m: console.print(m))
+        decision = run_tick(dry_run=dry_run, log=log)
         if decision.kind == "stop" and not dry_run:
             raise typer.Exit(0)
     except GpuRentError as exc:

src/gpu_rent/llm_runtime.py

@@ -26,12 +26,18 @@ OLLAMA_PRESETS: dict[str, list[str]] = {
     "empty": [],
 }
 
-PRESET_HELP = (
+OLLAMA_PRESET_LABELS: dict[str, str] = {
-    "recommended — Qwen2.5 7B abliterate (RU/EN, мало отказов, ~5GB)\n"
+    "recommended": "Qwen2.5 7B abliterate (RU/EN, мало отказов, ~5GB)",
-    "light — qwen2.5:3b (быстрее, слабее)\n"
+    "light": "qwen2.5:3b (быстрее, слабее)",
-    "stock — официальный qwen2.5:7b (больше цензуры)\n"
+    "stock": "официальный qwen2.5:7b (больше цензуры)",
-    "alt — другой abliterate-пак 7B\n"
+    "alt": "другой abliterate-пак 7B",
-    "empty — только runtime, без pull"
+    "empty": "только runtime, без pull",
+    "keep": "не менять ollama-models.yaml",
+}
+
+# Deprecated text blob — prefer menu helpers below.
+PRESET_HELP = "\n".join(
+    f"{k} — {v}" for k, v in OLLAMA_PRESET_LABELS.items() if k != "keep"
 )
 
 LLAMACPP_PRESETS: dict[str, list[str]] = {
@@ -47,13 +53,48 @@ LLAMACPP_PRESETS: dict[str, list[str]] = {
     "empty": [],
 }
 
-LLAMACPP_PRESET_HELP = (
+LLAMACPP_PRESET_LABELS: dict[str, str] = {
-    "recommended — Qwen2.5 7B abliterate GGUF Q4_K_M (~4.7GB, мало отказов)\n"
+    "recommended": "Qwen2.5 7B abliterate GGUF Q4_K_M (~4.7GB, мало отказов)",
-    "light — Qwen2.5 3B Instruct Q4_K_M (~2GB)\n"
+    "light": "Qwen2.5 3B Instruct Q4_K_M (~2GB)",
-    "stock — официальный Qwen2.5 7B Instruct Q4_K_M\n"
+    "stock": "официальный Qwen2.5 7B Instruct Q4_K_M",
-    "empty — только llama-server, GGUF положи вручную / правь llamacpp-models.yaml"
+    "empty": "только llama-server, GGUF вручную",
+    "keep": "не менять llamacpp-models.yaml",
+}
+
+LLAMACPP_PRESET_HELP = "\n".join(
+    f"{k} — {v}" for k, v in LLAMACPP_PRESET_LABELS.items() if k != "keep"
 )
 
+LLM_RUNTIME_LABELS: dict[str, str] = {
+    "none": "только SwarmUI",
+    "ollama": "Ollama (+ pull моделей)",
+    "llamacpp": "llama.cpp server (+ GGUF)",
+}
+
+
+def llm_runtime_menu() -> list:
+    from gpu_rent.prompts import MenuItem
+
+    return [MenuItem(k, f"{k} — {LLM_RUNTIME_LABELS[k]}") for k in ("none", "ollama", "llamacpp")]
+
+
+def ollama_preset_menu(*, include_keep: bool = False) -> list:
+    from gpu_rent.prompts import MenuItem
+
+    keys = list(OLLAMA_PRESETS.keys())
+    if include_keep:
+        keys.append("keep")
+    return [MenuItem(k, OLLAMA_PRESET_LABELS.get(k, k)) for k in keys]
+
+
+def llamacpp_preset_menu(*, include_keep: bool = False) -> list:
+    from gpu_rent.prompts import MenuItem
+
+    keys = list(LLAMACPP_PRESETS.keys())
+    if include_keep:
+        keys.append("keep")
+    return [MenuItem(k, LLAMACPP_PRESET_LABELS.get(k, k)) for k in keys]
+
 
 @dataclass(frozen=True)
 class OllamaModelEntry:

src/gpu_rent/pools.py

@@ -186,31 +186,35 @@ def scan_pools(cfg: Config, pools: tuple[str, ...] | None = None) -> list[PoolGp
 
 def format_pool_scan(offers: list[PoolGpuOffer], preference: tuple[str, ...]) -> list[str]:
     lines = [
-        "скан пулов (мультизональные + кандидаты) × характеристики GPU:",
+        "[bold cyan]скан пулов[/bold cyan] (мультизональные + кандидаты) × характеристики GPU:",
-        "  (серые кнопки панели ≠ Nova; available_count бывает только у части flavors)",
+        "  [dim](серые кнопки панели ≠ Nova; available_count бывает только у части flavors)[/dim]",
     ]
     for offer in offers:
         tag = "multizone" if offer.multizone else "pool"
-        prefix = f"  {offer.region} ({tag}):"
+        prefix = f"  [bold]{offer.region}[/bold] ({tag}):"
         if offer.error:
-            lines.append(f"{prefix} ОШИБКА — {offer.error}")
+            lines.append(f"{prefix} [red]ОШИБКА[/red] — {offer.error}")
             continue
         types_s = ", ".join(offer.gpu_types) if offer.gpu_types else "—"
         lines.append(f"{prefix} типы: {types_s}")
         if offer.gpu_labels_found:
             best = offer.matched[0] if offer.matched else None
-            best_s = f" → лучший {best.name} ({best.label})" if best else ""
+            best_s = (
+                f" → [green]лучший {best.name} ({best.label})[/green]" if best else ""
+            )
             lines.append(
                 f"           preference: {', '.join(offer.gpu_labels_found)}{best_s}"
             )
         else:
-            lines.append(f"           preference: нет совпадений с {','.join(preference)}")
+            lines.append(
+                f"           [yellow]preference: нет совпадений с {','.join(preference)}[/yellow]"
+            )
     for offer in offers:
         if offer.matched and not offer.error:
             lines.append(
-                f"рекомендация: OS_REGION_NAME={offer.region}  "
+                f"[green]рекомендация:[/green] OS_REGION_NAME={offer.region}  "
                 f"GPU_RENT_AZ={offer.region}a  "
-                f"(сегмент a/b/c — в панели; диски и VM в одном AZ)"
+                f"[dim](сегмент a/b/c — в панели; диски и VM в одном AZ)[/dim]"
             )
             break
     return lines

src/gpu_rent/prompts.py

@@ -0,0 +1,72 @@
+"""Numbered interactive menus (Enter = default index)."""
+
+from __future__ import annotations
+
+from collections.abc import Callable, Sequence
+from dataclasses import dataclass
+
+from rich.markup import escape
+
+Ask = Callable[[str, str], str]
+Show = Callable[[str], None]
+
+
+@dataclass(frozen=True)
+class MenuItem:
+    key: str
+    label: str
+
+
+def resolve_menu_choice(
+    raw: str,
+    items: Sequence[MenuItem],
+    *,
+    default_key: str,
+) -> str:
+    """Accept 1-based index or exact key (case-insensitive). Empty → default."""
+    text = (raw or "").strip().lower()
+    if not text:
+        return default_key
+    if text.isdigit():
+        idx = int(text)
+        if 1 <= idx <= len(items):
+            return items[idx - 1].key
+        raise ValueError(f"выбор: жду 1…{len(items)}, получили {idx}")
+    for item in items:
+        if item.key.lower() == text:
+            return item.key
+    keys = ", ".join(f"{i}:{it.key}" for i, it in enumerate(items, 1))
+    raise ValueError(f"выбор: жду номер или ключ ({keys}), получили {raw!r}")
+
+
+def prompt_menu(
+    title: str,
+    items: Sequence[MenuItem],
+    *,
+    default: str,
+    ask: Ask,
+    show: Show | None = None,
+) -> str:
+    """Print numbered list, ask for index (default = current recommendation)."""
+    if not items:
+        raise ValueError(f"{title}: пустое меню")
+    default_idx = 1
+    for i, item in enumerate(items, 1):
+        if item.key == default:
+            default_idx = i
+            break
+    safe_title = escape(title)
+    lines = [f"[bold cyan]{safe_title}[/bold cyan]"]
+    for i, item in enumerate(items, 1):
+        label = escape(item.label)
+        if i == default_idx:
+            lines.append(f"  [cyan]{i}.[/cyan] {label} [green]←[/green]")
+        else:
+            lines.append(f"  [dim]{i}.[/dim] {label}")
+    emit = show
+    if emit is None:
+        from gpu_rent.term import log as emit  # type: ignore[assignment]
+    for line in lines:
+        emit(line)
+    raw = ask(f"{title} [1-{len(items)}]", str(default_idx))
+    return resolve_menu_choice(raw, items, default_key=items[default_idx - 1].key)

src/gpu_rent/setup_wizard.py

@@ -7,12 +7,13 @@ from collections.abc import Callable
 from pathlib import Path
 
 from gpu_rent.llm_runtime import (
-    LLAMACPP_PRESET_HELP,
-    PRESET_HELP,
     append_vars_llm_runtime,
     ensure_llamacpp_manifest_from_example,
     ensure_ollama_manifest_from_example,
+    llamacpp_preset_menu,
+    llm_runtime_menu,
     normalize_runtime,
+    ollama_preset_menu,
     write_llamacpp_models_preset,
     write_ollama_models_preset,
 )
@@ -28,6 +29,7 @@ from gpu_rent.paths import (
     vars_example_path,
     vars_path,
 )
+from gpu_rent.prompts import prompt_menu
 
 Log = Callable[[str], None]
 
@@ -74,9 +76,12 @@ def run_setup(
     runtime = llm
     if runtime is None:
         if ask:
-            runtime = ask(
+            runtime = prompt_menu(
-                "LLM runtime [none/ollama/llamacpp]",
+                "LLM runtime",
-                "none",
+                llm_runtime_menu(),
+                default="none",
+                ask=ask,
+                show=log,
             )
         else:
             runtime = "none"
@@ -87,8 +92,13 @@ def run_setup(
     if runtime == "ollama":
         preset = ollama_preset
         if preset is None and ask:
-            log(PRESET_HELP)
+            preset = prompt_menu(
-            preset = ask("Ollama preset [recommended/light/stock/alt/empty]", "recommended")
+                "Ollama preset",
+                ollama_preset_menu(include_keep=False),
+                default="recommended",
+                ask=ask,
+                show=log,
+            )
         if preset is None:
             preset = "recommended"
         if preset.strip().lower() in {"keep", "example", ""}:
@@ -98,12 +108,14 @@ def run_setup(
             write_ollama_models_preset(ollama_models_manifest_path(), preset)
             log(f"ollama-models.yaml пресет={preset}")
     elif runtime == "llamacpp":
-        preset = ollama_preset  # reuse --ollama-preset flag as generic LLM preset in setup
+        preset = ollama_preset
         if preset is None and ask:
-            log(LLAMACPP_PRESET_HELP)
+            preset = prompt_menu(
-            preset = ask(
+                "llama.cpp GGUF",
-                "llama.cpp GGUF preset [recommended/light/stock/empty]",
+                llamacpp_preset_menu(include_keep=False),
-                "recommended",
+                default="recommended",
+                ask=ask,
+                show=log,
             )
         if preset is None:
             preset = "recommended"

src/gpu_rent/term.py

@@ -0,0 +1,105 @@
+"""Colored terminal output (Rich). Plain log lines get severity colors."""
+
+from __future__ import annotations
+
+import re
+import sys
+
+from rich.console import Console
+from rich.markup import escape
+
+if sys.platform == "win32":
+    for _stream in (sys.stdout, sys.stderr):
+        try:
+            _stream.reconfigure(encoding="utf-8", errors="replace")
+        except (AttributeError, OSError):
+            pass
+
+console = Console(highlight=False, legacy_windows=False)
+
+# Our own markup — leave as-is.
+_MARKUP = re.compile(
+    r"\[(/?)(bold|dim|red|green|yellow|cyan|magenta|blue|white|"
+    r"bold [a-z]+|dim [a-z]+)\]"
+)
+
+_ERR = re.compile(
+    r"(?i)\b("
+    r"fail|failed|error|exception|traceback|отмен|ошиб|не\s+удал|"
+    r"блокирующ|refuse|denied|abort"
+    r")\b"
+)
+_WARN = re.compile(
+    r"(?i)("
+    r"^!\s|"
+    r"\bwarn|\bwarning|предупрежд|внимание|"
+    r"тарифицир|idle-killer|preemptible:|"
+    r"₽:|диск data|серые кнопки|не\s+гасит|"
+    r"сейчас OS_REGION|лучше закрывается"
+    r")"
+)
+_OK = re.compile(
+    r"(?i)("
+    r"^ok\b|готово|создал|создан|doctor ok|SSH ок|"
+    r"туннель 127|← выберем|рекомендация:|"
+    r"выбрано:|armed|bootstrap уже|слушает 127|"
+    r"pull Output:|push .+: \d|wrote "
+    r")"
+)
+
+
+def has_markup(text: str) -> bool:
+    return bool(_MARKUP.search(text))
+
+
+def paint(msg: str) -> str:
+    """Wrap a plain log line in Rich markup by severity heuristics."""
+    if msg == "" or msg.isspace():
+        return msg
+    if has_markup(msg):
+        return msg
+
+    stripped = msg.lstrip()
+    body = escape(msg)
+
+    if stripped.startswith("!"):
+        return f"[yellow]{body}[/yellow]"
+    if _ERR.search(stripped):
+        return f"[red]{body}[/red]"
+    if _WARN.search(stripped):
+        return f"[yellow]{body}[/yellow]"
+    if _OK.search(stripped):
+        return f"[green]{body}[/green]"
+
+    # Numbered menus / lists
+    if re.match(r"^\s*\d+\.\s", msg):
+        return f"[white]{body}[/white]"
+    if stripped.endswith(":") and len(stripped) < 80 and not stripped.startswith("http"):
+        return f"[bold cyan]{body}[/bold cyan]"
+
+    return body
+
+
+def log(msg: str = "") -> None:
+    """CLI log callback — colored print."""
+    console.print(paint(msg))
+
+
+def ok(msg: str) -> None:
+    console.print(f"[green]{escape(msg)}[/green]")
+
+
+def warn(msg: str) -> None:
+    console.print(f"[yellow]{escape(msg)}[/yellow]")
+
+
+def err(msg: str) -> None:
+    console.print(f"[red]{escape(msg)}[/red]")
+
+
+def info(msg: str) -> None:
+    console.print(f"[cyan]{escape(msg)}[/cyan]")
+
+
+def dim(msg: str) -> None:
+    console.print(f"[dim]{escape(msg)}[/dim]")

src/gpu_rent/ux.py

@@ -6,6 +6,8 @@ from collections.abc import Callable
 from dataclasses import dataclass
 from typing import Any
 
+from rich.markup import escape
+
 from gpu_rent.config import Config
 from gpu_rent.inventory import FlavorInfo, looks_like_gpu, rank_flavors, resolve_flavor
 
@@ -21,14 +23,19 @@ def list_ranked_flavors(flavors: list[Any], cfg: Config) -> list[FlavorInfo]:
 
 def format_flavor_lines(ranked: list[FlavorInfo], picked: FlavorInfo | None = None) -> list[str]:
     if not ranked:
-        return ["flavors: пусто (проверь регион / FLAVOR_PREFERENCE)"]
+        return ["[yellow]flavors: пусто (проверь регион / FLAVOR_PREFERENCE)[/yellow]"]
-    lines = ["flavors (по FLAVOR_PREFERENCE):"]
+    lines = ["[bold cyan]flavors (по FLAVOR_PREFERENCE):[/bold cyan]"]
     for i, info in enumerate(ranked, 1):
-        mark = " ← выберем" if picked and info.id == picked.id else ""
+        mark = " [green]← выберем[/green]" if picked and info.id == picked.id else ""
         ram = f"{info.ram_mb // 1024}GB" if info.ram_mb else "?"
         vcpu = str(info.vcpus) if info.vcpus is not None else "?"
-        label = info.label or "—"
+        label = escape(info.label or "—")
-        lines.append(f"  {i}. [{label}] {info.name}  vCPU={vcpu} RAM={ram}  id={info.id}{mark}")
+        name = escape(info.name)
+        fid = escape(info.id)
+        lines.append(
+            f"  [cyan]{i}.[/cyan] [bold]{label}[/bold] {name}  "
+            f"vCPU={vcpu} RAM={ram}  id={fid}{mark}"
+        )
     return lines
 
 
@@ -114,21 +121,18 @@ def prompt_server_plan(
     if data_gb < 20:
         raise ValueError("Data disk GB: минимум 20")
 
-    spot_default = "Y" if spot else "n"
+    from gpu_rent.prompts import MenuItem, prompt_menu
-    raw_spot = (
+
-        ask(
+    spot_key = prompt_menu(
-            "Preemptible GPU (дешевле, могут усыпить ~24ч)? [Y/n]",
+        "Тариф GPU",
-            spot_default,
+        [
-        )
+            MenuItem("yes", "preemptible (дешевле, могут усыпить ~24ч)"),
-        .strip()
+            MenuItem("no", "обычный on-demand"),
-        .lower()
+        ],
+        default="yes" if spot else "no",
+        ask=ask,
     )
-    if raw_spot in {"", "y", "yes", "1", "true", "on"}:
+    use_spot = spot_key == "yes"
-        use_spot = True
-    elif raw_spot in {"n", "no", "0", "false", "off"}:
-        use_spot = False
-    else:
-        raise ValueError(f"preemptible: жду Y/n, получили {raw_spot!r}")
 
     changed = (
         chosen.id != picked.id

tests/test_prompts.py

@@ -0,0 +1,32 @@
+from gpu_rent.prompts import MenuItem, prompt_menu, resolve_menu_choice
+
+
+def test_resolve_by_number():
+    items = [MenuItem("a", "A"), MenuItem("b", "B"), MenuItem("keep", "K")]
+    assert resolve_menu_choice("2", items, default_key="keep") == "b"
+    assert resolve_menu_choice("", items, default_key="keep") == "keep"
+    assert resolve_menu_choice("keep", items, default_key="a") == "keep"
+
+
+def test_prompt_menu_prints_and_picks():
+    items = [
+        MenuItem("recommended", "rec"),
+        MenuItem("light", "lit"),
+        MenuItem("keep", "keep yaml"),
+    ]
+    shown: list[str] = []
+
+    def ask(msg: str, default: str = "") -> str:
+        assert default == "3"  # keep is default
+        return "1"
+
+    key = prompt_menu(
+        "GGUF",
+        items,
+        default="keep",
+        ask=ask,
+        show=shown.append,
+    )
+    assert key == "recommended"
+    assert any("1. rec" in line for line in shown)
+    assert any("←" in line for line in shown)

tests/test_server_plan.py

@@ -31,7 +31,7 @@ def test_prompt_server_plan_with_ranked(monkeypatch, tmp_path):
         "gpu_rent.ux.list_ranked_flavors",
         lambda flavors, cfg: [f1, f2],
     )
-    answers = iter(["2", "200", "n"])
+    answers = iter(["2", "200", "2"])
 
     def ask(msg: str, default: str = "") -> str:
         return next(answers)