feat(agent): expand LLM provider and wizard support

app/agent/llm/provider.py

@@ -117,6 +117,58 @@ class LLMProviderManager(metaclass=Singleton):
         "gpt-5.4-mini",
         "gpt-5.5",
     }
+    _MODELS_DEV_DYNAMIC_SKIP_IDS = {
+        "aihubmix",
+        "amazon-bedrock",
+        "azure",
+        "azure-cognitive-services",
+        "cloudflare-ai-gateway",
+        "cohere",
+        "gitlab",
+        "google-vertex",
+        "google-vertex-anthropic",
+        "kiro",
+        "sap-ai-core",
+        "v0",
+        "vercel",
+    }
+    _MODELS_DEV_DYNAMIC_PROVIDER_OVERRIDES = {
+        "bailing": {
+            "runtime": "openai_compatible",
+            "default_base_url": "https://api.tbox.cn/api/llm/v1",
+            "description": "Bailing OpenAI-compatible 端点。",
+        },
+        "cerebras": {
+            "runtime": "openai_compatible",
+            "default_base_url": "https://api.cerebras.ai/v1",
+            "description": "Cerebras 官方兼容端点。",
+        },
+        "deepinfra": {
+            "runtime": "openai_compatible",
+            "default_base_url": "https://api.deepinfra.com/v1/openai",
+            "description": "DeepInfra 官方兼容端点。",
+        },
+        "mistral": {
+            "runtime": "openai_compatible",
+            "default_base_url": "https://api.mistral.ai/v1",
+            "description": "Mistral 官方兼容端点。",
+        },
+        "perplexity": {
+            "runtime": "openai_compatible",
+            "default_base_url": "https://api.perplexity.ai/v1",
+            "description": "Perplexity 官方兼容端点。",
+        },
+        "togetherai": {
+            "runtime": "openai_compatible",
+            "default_base_url": "https://api.together.xyz/v1",
+            "description": "Together AI 官方兼容端点。",
+        },
+        "venice": {
+            "runtime": "openai_compatible",
+            "default_base_url": "https://api.venice.ai/api/v1",
+            "description": "Venice AI 官方兼容端点。",
+        },
+    }
 
     def __init__(self):
         self._lock = threading.RLock()
@@ -130,7 +182,7 @@ class LLMProviderManager(metaclass=Singleton):
         )
 
     @staticmethod
-    def _provider_specs() -> tuple[ProviderSpec, ...]:
+    def _builtin_provider_specs() -> tuple[ProviderSpec, ...]:
         """
         返回受支持的 provider 定义。
 
@@ -708,43 +760,253 @@ class LLMProviderManager(metaclass=Singleton):
         )
         return tuple(providers)
 
+    def _cached_models_dev_payload(self) -> dict[str, Any]:
+        if isinstance(self._models_dev_data, dict):
+            return self._models_dev_data
+
+        try:
+            if not self._models_dev_cache_path.exists():
+                return {}
+            payload = json.loads(self._models_dev_cache_path.read_text(encoding="utf-8"))
+        except Exception as err:
+            logger.warning(f"读取 models.dev provider 缓存失败: {err}")
+            return {}
+
+        if not isinstance(payload, dict):
+            return {}
+
+        self._models_dev_data = payload
+        return payload
+
+    @staticmethod
+    def _models_dev_env_names(payload: dict[str, Any]) -> tuple[str, ...]:
+        raw_env_names = payload.get("env")
+        if not isinstance(raw_env_names, list):
+            return ()
+        env_names = []
+        for item in raw_env_names:
+            value = str(item or "").strip()
+            if value:
+                env_names.append(value)
+        return tuple(env_names)
+
+    @classmethod
+    def _models_dev_reserved_provider_ids(
+            cls, specs: tuple[ProviderSpec, ...]
+    ) -> set[str]:
+        reserved_ids: set[str] = set()
+        for spec in specs:
+            if spec.models_dev_provider_id:
+                reserved_ids.add(spec.models_dev_provider_id)
+            for preset in spec.base_url_presets:
+                if preset.models_dev_provider_id:
+                    reserved_ids.add(preset.models_dev_provider_id)
+        return reserved_ids
+
+    @staticmethod
+    def _dynamic_api_key_label(env_names: tuple[str, ...]) -> str:
+        first_env = env_names[0].upper() if env_names else ""
+        if "TOKEN" in first_env and "KEY" not in first_env:
+            return "API Token"
+        return "API Key"
+
+    @classmethod
+    def _normalize_models_dev_base_url(
+            cls, runtime: str, base_url: Optional[str]
+    ) -> Optional[str]:
+        normalized = cls._sanitize_base_url(base_url)
+        if not normalized:
+            return None
+
+        suffixes = {
+            "openai_compatible": (
+                "/chat/completions",
+                "/completions",
+                "/responses",
+                "/embeddings",
+                "/audio/speech",
+                "/audio/transcriptions",
+            ),
+            "anthropic_compatible": (
+                "/messages",
+            ),
+        }
+
+        for suffix in suffixes.get(runtime, ()): 
+            if normalized.endswith(suffix):
+                normalized = normalized[: -len(suffix)]
+                break
+        return cls._sanitize_base_url(normalized)
+
+    @classmethod
+    def _models_dev_dynamic_provider_spec(
+            cls,
+            provider_id: str,
+            payload: dict[str, Any],
+            sort_order: int,
+    ) -> ProviderSpec | None:
+        normalized_id = str(provider_id or "").strip().lower()
+        if not normalized_id or normalized_id in cls._MODELS_DEV_DYNAMIC_SKIP_IDS:
+            return None
+
+        override = cls._MODELS_DEV_DYNAMIC_PROVIDER_OVERRIDES.get(normalized_id, {})
+        npm_package = str(payload.get("npm") or "").strip()
+        runtime = override.get("runtime")
+        if not runtime:
+            if npm_package == "@ai-sdk/openai-compatible":
+                runtime = "openai_compatible"
+            elif npm_package == "@ai-sdk/anthropic":
+                runtime = "anthropic_compatible"
+            else:
+                return None
+
+        model_list_strategy = override.get("model_list_strategy")
+        if not model_list_strategy:
+            model_list_strategy = (
+                "anthropic_compatible"
+                if runtime == "anthropic_compatible"
+                else "models_dev_only"
+            )
+
+        default_base_url = cls._normalize_models_dev_base_url(
+            runtime=runtime,
+            base_url=override.get("default_base_url") or payload.get("api"),
+        )
+        requires_base_url = not bool(default_base_url)
+        env_names = cls._models_dev_env_names(payload)
+        api_key_label = override.get("api_key_label") or cls._dynamic_api_key_label(
+            env_names
+        )
+        name = str(payload.get("name") or override.get("name") or normalized_id).strip()
+        description = override.get("description")
+        if not description:
+            transport_name = "Anthropic-compatible" if runtime == "anthropic_compatible" else "OpenAI-compatible"
+            description = f"{name} {transport_name} 端点（来自 models.dev 目录）。"
+
+        api_key_hint = override.get("api_key_hint")
+        if not api_key_hint:
+            api_key_hint = f"填写 {name} {api_key_label}。"
+            if requires_base_url:
+                api_key_hint = f"填写 {name} {api_key_label}，并手动填写 Base URL。"
+
+        return ProviderSpec(
+            id=normalized_id,
+            name=name,
+            runtime=runtime,
+            models_dev_provider_id=normalized_id,
+            default_base_url=default_base_url,
+            base_url_editable=True,
+            requires_base_url=requires_base_url,
+            api_key_label=api_key_label,
+            api_key_hint=api_key_hint,
+            model_list_strategy=model_list_strategy,
+            description=description,
+            sort_order=sort_order,
+        )
+
+    def _dynamic_provider_specs(
+            self, builtin_specs: tuple[ProviderSpec, ...]
+    ) -> tuple[ProviderSpec, ...]:
+        payload = self._cached_models_dev_payload()
+        if not payload:
+            return ()
+
+        explicit_ids = {spec.id for spec in builtin_specs}
+        reserved_ids = self._models_dev_reserved_provider_ids(builtin_specs)
+        candidates: list[tuple[str, str, dict[str, Any]]] = []
+
+        for provider_id, provider_payload in payload.items():
+            normalized_id = str(provider_id or "").strip().lower()
+            if not normalized_id or not isinstance(provider_payload, dict):
+                continue
+            if normalized_id in explicit_ids or normalized_id in reserved_ids:
+                continue
+
+            spec = self._models_dev_dynamic_provider_spec(
+                provider_id=normalized_id,
+                payload=provider_payload,
+                sort_order=0,
+            )
+            if not spec:
+                continue
+            candidates.append((spec.name.lower(), normalized_id, provider_payload))
+
+        dynamic_specs = []
+        for sort_order, (_, provider_id, provider_payload) in enumerate(
+                sorted(candidates),
+                start=700,
+        ):
+            spec = self._models_dev_dynamic_provider_spec(
+                provider_id=provider_id,
+                payload=provider_payload,
+                sort_order=sort_order,
+            )
+            if not spec:
+                continue
+            dynamic_specs.append(spec)
+        return tuple(dynamic_specs)
+
+    def _provider_specs(self) -> tuple[ProviderSpec, ...]:
+        builtin_specs = self._builtin_provider_specs()
+        return builtin_specs + self._dynamic_provider_specs(builtin_specs)
+
+    async def _get_provider_async(
+            self, provider_id: str, force_refresh: bool = False
+    ) -> ProviderSpec:
+        try:
+            return self.get_provider(provider_id)
+        except LLMProviderError:
+            await self.get_models_dev_data(force_refresh=force_refresh)
+            return self.get_provider(provider_id)
+
+    def _serialize_provider(self, spec: ProviderSpec) -> dict[str, Any]:
+        return {
+            "id": spec.id,
+            "name": spec.name,
+            "runtime": spec.runtime,
+            "default_base_url": self._default_base_url_for_provider(spec) or "",
+            "base_url_presets": [
+                {
+                    "label": preset.label,
+                    "value": self._sanitize_base_url(preset.value) or "",
+                }
+                for preset in spec.base_url_presets
+            ],
+            "base_url_editable": spec.base_url_editable,
+            "requires_base_url": spec.requires_base_url,
+            "supports_api_key": spec.supports_api_key,
+            "api_key_label": spec.api_key_label,
+            "api_key_hint": spec.api_key_hint,
+            "supports_model_refresh": spec.supports_model_refresh,
+            "oauth_methods": [
+                {
+                    "id": method.id,
+                    "type": method.type,
+                    "label": method.label,
+                    "description": method.description,
+                }
+                for method in spec.oauth_methods
+            ],
+            "description": spec.description,
+            "auth_status": self.get_auth_status(spec.id),
+        }
+
+    async def list_providers_async(
+            self, force_refresh: bool = False
+    ) -> list[dict[str, Any]]:
+        """返回前端可渲染的 provider 目录，并优先补齐 models.dev 动态平台。"""
+        try:
+            await self.get_models_dev_data(force_refresh=force_refresh)
+        except Exception as err:
+            logger.debug(f"加载 models.dev provider 目录失败，回退内置列表: {err}")
+        return self.list_providers()
+
     def list_providers(self) -> list[dict[str, Any]]:
         """返回前端可渲染的 provider 目录。"""
-        providers = []
-        for spec in sorted(self._provider_specs(), key=lambda item: item.sort_order):
-            providers.append(
-                {
-                    "id": spec.id,
-                    "name": spec.name,
-                    "runtime": spec.runtime,
-                    "default_base_url": self._default_base_url_for_provider(spec) or "",
-                    "base_url_presets": [
-                        {
-                            "label": preset.label,
-                            "value": self._sanitize_base_url(preset.value) or "",
-                        }
-                        for preset in spec.base_url_presets
-                    ],
-                    "base_url_editable": spec.base_url_editable,
-                    "requires_base_url": spec.requires_base_url,
-                    "supports_api_key": spec.supports_api_key,
-                    "api_key_label": spec.api_key_label,
-                    "api_key_hint": spec.api_key_hint,
-                    "supports_model_refresh": spec.supports_model_refresh,
-                    "oauth_methods": [
-                        {
-                            "id": method.id,
-                            "type": method.type,
-                            "label": method.label,
-                            "description": method.description,
-                        }
-                        for method in spec.oauth_methods
-                    ],
-                    "description": spec.description,
-                    "auth_status": self.get_auth_status(spec.id),
-                }
-            )
-        return providers
+        return [
+            self._serialize_provider(spec)
+            for spec in sorted(self._provider_specs(), key=lambda item: item.sort_order)
+        ]
 
     def get_provider(self, provider_id: str) -> ProviderSpec:
         """按 provider id 获取定义。"""
@@ -973,7 +1235,7 @@ class LLMProviderManager(metaclass=Singleton):
     async def _models_dev_provider_payload(
             self, provider_id: str, base_url: Optional[str] = None
     ) -> dict[str, Any]:
-        spec = self.get_provider(provider_id)
+        spec = await self._get_provider_async(provider_id)
         models_dev_provider_id = self._resolve_provider_models_dev_provider_id(
             spec,
             base_url,
@@ -1313,7 +1575,7 @@ class LLMProviderManager(metaclass=Singleton):
             force_refresh: bool = False,
     ) -> list[dict[str, Any]]:
         """返回标准化后的模型目录。"""
-        spec = self.get_provider(provider_id)
+        spec = await self._get_provider_async(provider_id, force_refresh=force_refresh)
         if self._resolve_provider_models_dev_provider_id(spec, base_url):
             # 对依赖 models.dev 的 provider 主动刷新一次缓存，保证“刷新模型列表”
             # 在使用目录型 provider 时也能拿到最新参数。
@@ -1449,7 +1711,7 @@ class LLMProviderManager(metaclass=Singleton):
 
         API Key 方式已经由普通设置表单覆盖，这里只处理需要交互式授权的 provider。
         """
-        provider = self.get_provider(provider_id)
+        provider = await self._get_provider_async(provider_id)
         method = next(
             (item for item in provider.oauth_methods if item.id == method_id),
             None,
@@ -1844,7 +2106,7 @@ class LLMProviderManager(metaclass=Singleton):
 
         返回统一结构，供 `LLMHelper` 创建具体 LangChain 模型实例时使用。
         """
-        spec = self.get_provider(provider_id)
+        spec = await self._get_provider_async(provider_id)
         normalized_api_key = str(api_key or "").strip() or None
         normalized_base_url = self._sanitize_base_url(base_url)
         model_record = None

app/api/endpoints/llm.py

@@ -98,7 +98,7 @@ async def get_llm_providers(
     返回前端可直接渲染的 provider 目录。
     """
     try:
-        providers = LLMProviderManager().list_providers()
+        providers = await LLMProviderManager().list_providers_async()
         return schemas.Response(success=True, data=providers)
     except Exception as err:
         return schemas.Response(success=False, message=str(err))

app/core/config.py

@@ -501,7 +501,7 @@ class ConfigModel(BaseModel):
     AI_AGENT_ENABLE: bool = False
     # 合局AI智能体
     AI_AGENT_GLOBAL: bool = False
-    # LLM提供商 (openai/google/deepseek)
+    # LLM提供商（支持内置 provider，以及从 models.dev 动态补充的平台）
     LLM_PROVIDER: str = "deepseek"
     # LLM模型名称
     LLM_MODEL: str = "deepseek-chat"

scripts/local_setup.py

@@ -5,6 +5,7 @@ from __future__ import annotations
 import argparse
 import asyncio
 import getpass
+import importlib.util
 import json
 import os
 import platform
@@ -66,6 +67,26 @@ LLM_PROVIDER_DEFAULTS = {
         "model": "gemini-2.5-flash",
         "base_url": "",
     },
+    "anthropic": {
+        "model": "claude-sonnet-4-0",
+        "base_url": "https://api.anthropic.com/v1",
+    },
+    "openrouter": {
+        "model": "openai/gpt-4.1-mini",
+        "base_url": "https://openrouter.ai/api/v1",
+    },
+    "groq": {
+        "model": "llama-3.3-70b-versatile",
+        "base_url": "https://api.groq.com/openai/v1",
+    },
+}
+LLM_PROVIDER_FALLBACK_CHOICES = {
+    "deepseek": "DeepSeek",
+    "openai": "OpenAI Compatible",
+    "google": "Google",
+    "anthropic": "Anthropic",
+    "openrouter": "OpenRouter",
+    "groq": "Groq",
 }
 RUNTIME_PACKAGE = {
     "name": "moviepilot-frontend-runtime",
@@ -1063,6 +1084,273 @@ def _prompt_choice(label: str, choices: dict[str, str], default: str) -> str:
         print("请输入列表中的可选值。")
 
 
+def _prompt_provider_choice(label: str, choices: dict[str, str], default: str) -> str:
+    labels = []
+    normalized_map: dict[str, str] = {}
+    for key, desc in choices.items():
+        labels.append(f"{key}({desc})")
+        normalized_map[_normalize_choice(key)] = key
+
+    preview_limit = 12
+    print("可用 LLM 提供商：")
+    for item in labels[:preview_limit]:
+        print(f"  {item}")
+    if len(labels) > preview_limit:
+        print(f"  ... 另有 {len(labels) - preview_limit} 个，可直接输入 provider id")
+
+    while True:
+        raw = input(f"{label} (默认 {default}，可直接输入 provider id): ").strip()
+        if not raw:
+            return default
+        normalized = _normalize_choice(raw)
+        if normalized in normalized_map:
+            return normalized_map[normalized]
+
+        provider_id = raw.strip().lower()
+        if re.fullmatch(r"[a-z0-9][a-z0-9._-]*", provider_id):
+            return provider_id
+        print("请输入列表中的可选值，或合法的 provider id（小写字母/数字/.-_）。")
+
+
+def _load_llm_provider_module():
+    provider_path = ROOT / "app" / "agent" / "llm" / "provider.py"
+    module_name = f"moviepilot_local_llm_provider_{uuid.uuid4().hex}"
+    spec = importlib.util.spec_from_file_location(module_name, provider_path)
+    if not spec or not spec.loader:
+        raise RuntimeError("无法加载 LLM provider 模块")
+    module = importlib.util.module_from_spec(spec)
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)
+    return module
+
+
+def _load_llm_provider_definitions_inner() -> list[dict[str, Any]]:
+    provider_module = _load_llm_provider_module()
+    providers = asyncio.run(provider_module.LLMProviderManager().list_providers_async())
+    return providers if isinstance(providers, list) else []
+
+
+def _load_llm_provider_definitions(
+    runtime_python: Optional[Path] = None,
+) -> list[dict[str, Any]]:
+    try:
+        return _load_llm_provider_definitions_inner()
+    except Exception as exc:
+        if runtime_python and not _current_python_matches(runtime_python):
+            try:
+                with TemporaryDirectory() as temp_dir:
+                    output_path = Path(temp_dir) / "llm-providers.json"
+                    subprocess.run(
+                        [
+                            str(runtime_python),
+                            str(Path(__file__).resolve()),
+                            "query-llm-providers",
+                            "--output-json-file",
+                            str(output_path),
+                        ],
+                        cwd=str(ROOT),
+                        check=True,
+                        stdout=subprocess.DEVNULL,
+                        stderr=subprocess.DEVNULL,
+                    )
+                    data = json.loads(output_path.read_text(encoding="utf-8"))
+                    if isinstance(data, list):
+                        return data
+            except Exception as runtime_exc:
+                print_step(
+                    f"当前环境暂时无法读取 LLM 提供商目录，已回退到常用平台列表：{runtime_exc}"
+                )
+                return []
+
+        print_step(f"当前环境暂时无法读取 LLM 提供商目录，已回退到常用平台列表：{exc}")
+        return []
+
+
+def _llm_provider_choice_map(
+    provider_definitions: list[dict[str, Any]],
+) -> dict[str, str]:
+    choices: dict[str, str] = {}
+    for item in provider_definitions:
+        if not isinstance(item, dict):
+            continue
+        if item.get("supports_api_key") is False:
+            continue
+        provider_id = str(item.get("id") or "").strip().lower()
+        name = str(item.get("name") or provider_id).strip()
+        if not provider_id or not name:
+            continue
+        choices[provider_id] = name
+    if choices:
+        return choices
+    return dict(LLM_PROVIDER_FALLBACK_CHOICES)
+
+
+def _llm_provider_defaults(
+    provider: str,
+    provider_definitions: list[dict[str, Any]],
+) -> dict[str, str]:
+    normalized_provider = str(provider or "").strip().lower()
+    defaults = dict(LLM_PROVIDER_DEFAULTS.get(normalized_provider) or {})
+    provider_meta = next(
+        (
+            item
+            for item in provider_definitions
+            if isinstance(item, dict)
+            and str(item.get("id") or "").strip().lower() == normalized_provider
+        ),
+        None,
+    )
+    if isinstance(provider_meta, dict):
+        default_base_url = str(provider_meta.get("default_base_url") or "").strip()
+        if default_base_url:
+            defaults["base_url"] = default_base_url
+
+    defaults.setdefault("model", _env_default("LLM_MODEL", ""))
+    defaults.setdefault("base_url", _env_default("LLM_BASE_URL", ""))
+    return defaults
+
+
+def _llm_provider_meta(
+    provider: str,
+    provider_definitions: list[dict[str, Any]],
+) -> dict[str, Any]:
+    normalized_provider = str(provider or "").strip().lower()
+    provider_meta = next(
+        (
+            item
+            for item in provider_definitions
+            if isinstance(item, dict)
+            and str(item.get("id") or "").strip().lower() == normalized_provider
+        ),
+        None,
+    )
+    return dict(provider_meta) if isinstance(provider_meta, dict) else {}
+
+
+def _load_llm_models_inner(payload: dict[str, Any]) -> list[dict[str, Any]]:
+    provider = str(payload.get("provider") or "").strip().lower()
+    if not provider:
+        raise RuntimeError("缺少 LLM provider")
+
+    provider_module = _load_llm_provider_module()
+    api_key = str(payload.get("api_key") or "").strip() or None
+    base_url = str(payload.get("base_url") or "").strip() or None
+    models = asyncio.run(
+        provider_module.LLMProviderManager().list_models(
+            provider_id=provider,
+            api_key=api_key,
+            base_url=base_url,
+            force_refresh=False,
+        )
+    )
+    return models if isinstance(models, list) else []
+
+
+def _load_llm_models(
+    *,
+    provider: str,
+    api_key: Optional[str],
+    base_url: Optional[str],
+    runtime_python: Optional[Path] = None,
+) -> list[dict[str, Any]]:
+    payload = {
+        "provider": str(provider or "").strip().lower(),
+        "api_key": str(api_key or "").strip(),
+        "base_url": str(base_url or "").strip(),
+    }
+    try:
+        return _load_llm_models_inner(payload)
+    except Exception as exc:
+        if runtime_python and not _current_python_matches(runtime_python):
+            try:
+                with TemporaryDirectory() as temp_dir:
+                    request_path = Path(temp_dir) / "llm-models-request.json"
+                    output_path = Path(temp_dir) / "llm-models.json"
+                    request_path.write_text(
+                        json.dumps(payload, ensure_ascii=False), encoding="utf-8"
+                    )
+                    subprocess.run(
+                        [
+                            str(runtime_python),
+                            str(Path(__file__).resolve()),
+                            "query-llm-models",
+                            "--request-json-file",
+                            str(request_path),
+                            "--output-json-file",
+                            str(output_path),
+                        ],
+                        cwd=str(ROOT),
+                        check=True,
+                        stdout=subprocess.DEVNULL,
+                        stderr=subprocess.DEVNULL,
+                    )
+                    data = json.loads(output_path.read_text(encoding="utf-8"))
+                    if isinstance(data, list):
+                        return data
+            except Exception as runtime_exc:
+                print_step(
+                    f"当前环境暂时无法获取 {payload['provider']} 模型目录，已回退为手动输入模型名称：{runtime_exc}"
+                )
+                return []
+
+        print_step(
+            f"当前环境暂时无法获取 {payload['provider']} 模型目录，已回退为手动输入模型名称：{exc}"
+        )
+        return []
+
+
+def _print_llm_models(models: list[dict[str, Any]], limit: int = 20) -> None:
+    print("可用模型：")
+    for index, item in enumerate(models[:limit], start=1):
+        if not isinstance(item, dict):
+            continue
+        model_id = str(item.get("id") or "").strip()
+        if not model_id:
+            continue
+        model_name = str(item.get("name") or model_id).strip()
+        extras: list[str] = []
+        if item.get("context_tokens_k"):
+            extras.append(f"{item['context_tokens_k']}K")
+        if item.get("supports_reasoning"):
+            extras.append("reasoning")
+        if item.get("supports_tools"):
+            extras.append("tools")
+        if item.get("supports_image_input"):
+            extras.append("vision")
+        extra_text = f" [{' / '.join(extras)}]" if extras else ""
+        if model_name != model_id:
+            print(f"  {index}. {model_id} ({model_name}){extra_text}")
+        else:
+            print(f"  {index}. {model_id}{extra_text}")
+    if len(models) > limit:
+        print(f"  ... 共 {len(models)} 个模型，可输入编号或直接输入模型名称")
+
+
+def _prompt_model_choice(models: list[dict[str, Any]], default: Optional[str] = None) -> str:
+    valid_models = [item for item in models if isinstance(item, dict) and item.get("id")]
+    if not valid_models:
+        return _prompt_text("LLM 模型名称", default=default)
+
+    indexed_models = {
+        str(index): str(item.get("id")).strip()
+        for index, item in enumerate(valid_models, start=1)
+    }
+    default_model = str(default or indexed_models.get("1") or "").strip()
+    _print_llm_models(valid_models)
+
+    while True:
+        raw = input(
+            f"LLM 模型名称/编号{' [' + default_model + ']' if default_model else ''}: "
+        ).strip()
+        if not raw and default_model:
+            return default_model
+        if raw in indexed_models:
+            return indexed_models[raw]
+        if raw:
+            return raw
+        print("请输入有效模型编号或模型名称。")
+
+
 def _env_llm_thinking_level_default() -> str:
     value = _normalize_choice(_env_default("LLM_THINKING_LEVEL", ""))
     alias_map = {
@@ -1459,7 +1747,9 @@ def _collect_notification_config() -> Optional[dict[str, Any]]:
     }
 
 
-def _collect_agent_config() -> dict[str, Any]:
+def _collect_agent_config(
+    runtime_python: Optional[Path] = None,
+) -> dict[str, Any]:
     print_step("AI Agent 配置")
     enabled = _prompt_yes_no(
         "是否启用 AI 智能体",
@@ -1471,22 +1761,42 @@ def _collect_agent_config() -> dict[str, Any]:
             "AI_AGENT_GLOBAL": False,
         }
 
+    provider_definitions = _load_llm_provider_definitions(runtime_python=runtime_python)
+    provider_choices = _llm_provider_choice_map(provider_definitions)
     current_provider = _env_default("LLM_PROVIDER", "deepseek").lower()
-    if current_provider not in LLM_PROVIDER_DEFAULTS:
+    if current_provider not in provider_choices:
         current_provider = "deepseek"
 
-    provider = _prompt_choice(
-        "选择 LLM 提供商",
-        {
-            "deepseek": "DeepSeek",
-            "openai": "OpenAI",
-            "google": "Google",
-        },
-        default=current_provider,
-    )
-    defaults = LLM_PROVIDER_DEFAULTS[provider]
+    while True:
+        provider = _prompt_provider_choice(
+            "选择 LLM 提供商",
+            provider_choices,
+            default=current_provider,
+        )
+        provider_meta = _llm_provider_meta(provider, provider_definitions)
+        if provider_meta.get("supports_api_key") is False:
+            print_step(
+                f"{provider_meta.get('name') or provider} 当前仅支持交互式授权，安装向导暂不支持，请改选可填写 API Key 的 provider。"
+            )
+            current_provider = "deepseek"
+            continue
+        break
+
+    defaults = _llm_provider_defaults(provider, provider_definitions)
     current_model = _env_default("LLM_MODEL", defaults["model"])
     current_base_url = _env_default("LLM_BASE_URL", defaults["base_url"])
+    api_key_label = str(provider_meta.get("api_key_label") or "API Key").strip() or "API Key"
+    api_key_hint = str(provider_meta.get("api_key_hint") or "").strip()
+    requires_base_url = bool(provider_meta.get("requires_base_url"))
+    base_url_label = (
+        "自定义 Google API Base URL（可选）"
+        if provider == "google"
+        else "LLM Base URL（必填）"
+        if requires_base_url
+        else "LLM Base URL"
+    )
+    if api_key_hint:
+        print_step(api_key_hint)
 
     config: dict[str, Any] = {
         "AI_AGENT_ENABLE": True,
@@ -1495,12 +1805,8 @@ def _collect_agent_config() -> dict[str, Any]:
             default=_env_bool("AI_AGENT_GLOBAL", False),
         ),
         "LLM_PROVIDER": provider,
-        "LLM_MODEL": _prompt_text(
-            "LLM 模型名称",
-            default=current_model,
-        ),
         "LLM_API_KEY": _prompt_secret_text(
-            "LLM API Key",
+            f"LLM {api_key_label}",
             current_value=read_env_value("LLM_API_KEY"),
             required=True,
         ),
@@ -1524,18 +1830,18 @@ def _collect_agent_config() -> dict[str, Any]:
         ),
     }
 
-    if provider == "google":
-        config["LLM_BASE_URL"] = _prompt_text(
-            "自定义 Google API Base URL（可选）",
-            default=current_base_url,
-            allow_empty=True,
-        )
-    else:
-        config["LLM_BASE_URL"] = _prompt_text(
-            "LLM Base URL",
-            default=current_base_url,
-            allow_empty=True,
-        )
+    config["LLM_BASE_URL"] = _prompt_text(
+        base_url_label,
+        default=current_base_url,
+        allow_empty=not requires_base_url,
+    )
+    models = _load_llm_models(
+        provider=provider,
+        api_key=config["LLM_API_KEY"],
+        base_url=config["LLM_BASE_URL"],
+        runtime_python=runtime_python,
+    )
+    config["LLM_MODEL"] = _prompt_model_choice(models, default=current_model)
 
     return config
 
@@ -1744,7 +2050,7 @@ def run_setup_wizard(
                 preset_password=preset_superuser_password,
             ),
             **_collect_database_config(),
-            **_collect_agent_config(),
+            **_collect_agent_config(runtime_python=runtime_python),
         },
         "directories": [_collect_directory_config()],
         "downloader": _collect_downloader_config(),
@@ -3276,6 +3582,23 @@ def build_parser() -> argparse.ArgumentParser:
         "--output-json-file", required=True, help=argparse.SUPPRESS
     )
 
+    query_llm_providers_parser = subparsers.add_parser(
+        "query-llm-providers", help=argparse.SUPPRESS
+    )
+    query_llm_providers_parser.add_argument(
+        "--output-json-file", required=True, help=argparse.SUPPRESS
+    )
+
+    query_llm_models_parser = subparsers.add_parser(
+        "query-llm-models", help=argparse.SUPPRESS
+    )
+    query_llm_models_parser.add_argument(
+        "--request-json-file", required=True, help=argparse.SUPPRESS
+    )
+    query_llm_models_parser.add_argument(
+        "--output-json-file", required=True, help=argparse.SUPPRESS
+    )
+
     return parser
 
 
@@ -3461,6 +3784,25 @@ def main() -> int:
                 encoding="utf-8",
             )
             return 0
+
+        if args.command == "query-llm-providers":
+            payload = _load_llm_provider_definitions_inner()
+            Path(args.output_json_file).write_text(
+                json.dumps(payload, ensure_ascii=False, indent=2),
+                encoding="utf-8",
+            )
+            return 0
+
+        if args.command == "query-llm-models":
+            payload = json.loads(Path(args.request_json_file).read_text(encoding="utf-8"))
+            if not isinstance(payload, dict):
+                raise RuntimeError("模型查询负载格式错误")
+            models = _load_llm_models_inner(payload)
+            Path(args.output_json_file).write_text(
+                json.dumps(models, ensure_ascii=False, indent=2),
+                encoding="utf-8",
+            )
+            return 0
     except subprocess.CalledProcessError as exc:
         print(f"命令执行失败，退出码：{exc.returncode}", file=sys.stderr)
         return exc.returncode

tests/test_llm_provider_registry.py

@@ -0,0 +1,215 @@
+import asyncio
+import importlib.util
+import sys
+import unittest
+from pathlib import Path
+from types import ModuleType, SimpleNamespace
+from unittest.mock import AsyncMock, patch
+
+
+def _stub_module(name: str, **attrs):
+    module = sys.modules.get(name)
+    if module is None:
+        module = ModuleType(name)
+        sys.modules[name] = module
+    for key, value in attrs.items():
+        setattr(module, key, value)
+    return module
+
+
+class _DummyLogger:
+    def __getattr__(self, _name):
+        return lambda *args, **kwargs: None
+
+
+class _DummySystemConfigOper:
+    def get(self, _key):
+        return {}
+
+    async def async_set(self, _key, _value):
+        return None
+
+
+for _module_name in ("aiofiles", "jwt"):
+    _stub_module(_module_name)
+
+_stub_module(
+    "app.core.config",
+    settings=SimpleNamespace(
+        TEMP_PATH="/tmp",
+        PROXY_HOST=None,
+        LLM_MAX_CONTEXT_TOKENS=64,
+    ),
+)
+_stub_module("app.db.systemconfig_oper", SystemConfigOper=_DummySystemConfigOper)
+_stub_module("app.log", logger=_DummyLogger())
+_stub_module("app.schemas.types", SystemConfigKey=SimpleNamespace(AIAgentConfig="agent"))
+
+provider_path = Path(__file__).resolve().parents[1] / "app" / "agent" / "llm" / "provider.py"
+spec = importlib.util.spec_from_file_location("test_llm_provider_module", provider_path)
+provider_module = importlib.util.module_from_spec(spec)
+assert spec and spec.loader
+sys.modules[spec.name] = provider_module
+spec.loader.exec_module(provider_module)
+
+LLMProviderError = provider_module.LLMProviderError
+LLMProviderManager = provider_module.LLMProviderManager
+
+
+class LlmProviderRegistryTest(unittest.TestCase):
+    def setUp(self):
+        LLMProviderManager._instances.clear()
+
+    def tearDown(self):
+        LLMProviderManager._instances.clear()
+
+    def test_dynamic_provider_is_exposed_from_models_dev_cache(self):
+        manager = LLMProviderManager()
+        manager._models_dev_data = {
+            "frogbot": {
+                "id": "frogbot",
+                "name": "FrogBot",
+                "npm": "@ai-sdk/openai-compatible",
+                "env": ["FROGBOT_API_KEY"],
+                "api": "https://app.frogbot.ai/api/v1",
+                "models": {},
+            }
+        }
+
+        provider = manager.get_provider("frogbot")
+
+        self.assertEqual(provider.id, "frogbot")
+        self.assertEqual(provider.runtime, "openai_compatible")
+        self.assertEqual(provider.default_base_url, "https://app.frogbot.ai/api/v1")
+        self.assertFalse(provider.requires_base_url)
+        self.assertTrue(provider.base_url_editable)
+        self.assertEqual(provider.model_list_strategy, "models_dev_only")
+
+    def test_dynamic_provider_override_normalizes_chat_endpoint_base_url(self):
+        manager = LLMProviderManager()
+        manager._models_dev_data = {
+            "bailing": {
+                "id": "bailing",
+                "name": "Bailing",
+                "npm": "@ai-sdk/openai-compatible",
+                "env": ["BAILING_API_TOKEN"],
+                "api": "https://api.tbox.cn/api/llm/v1/chat/completions",
+                "models": {},
+            }
+        }
+
+        provider = manager.get_provider("bailing")
+
+        self.assertEqual(provider.default_base_url, "https://api.tbox.cn/api/llm/v1")
+        self.assertEqual(provider.api_key_label, "API Token")
+
+    def test_dynamic_provider_skips_alias_only_models_dev_ids(self):
+        manager = LLMProviderManager()
+        manager._models_dev_data = {
+            "moonshotai": {
+                "id": "moonshotai",
+                "name": "Moonshot AI Intl",
+                "npm": "@ai-sdk/openai-compatible",
+                "env": ["MOONSHOT_API_KEY"],
+                "api": "https://api.moonshot.ai/v1",
+                "models": {},
+            }
+        }
+
+        with self.assertRaises(LLMProviderError):
+            manager.get_provider("moonshotai")
+
+    def test_dynamic_provider_skips_incompatible_models_dev_provider(self):
+        manager = LLMProviderManager()
+        manager._models_dev_data = {
+            "azure": {
+                "id": "azure",
+                "name": "Azure",
+                "npm": "@ai-sdk/azure",
+                "env": ["AZURE_API_KEY"],
+                "models": {},
+            }
+        }
+
+        with self.assertRaises(LLMProviderError):
+            manager.get_provider("azure")
+
+    def test_dynamic_provider_without_known_base_url_requires_manual_input(self):
+        manager = LLMProviderManager()
+        manager._models_dev_data = {
+            "custom-anthropic": {
+                "id": "custom-anthropic",
+                "name": "Custom Anthropic",
+                "npm": "@ai-sdk/anthropic",
+                "env": ["CUSTOM_ANTHROPIC_KEY"],
+                "models": {},
+            }
+        }
+
+        provider = manager.get_provider("custom-anthropic")
+
+        self.assertEqual(provider.runtime, "anthropic_compatible")
+        self.assertTrue(provider.requires_base_url)
+        self.assertTrue(provider.base_url_editable)
+        self.assertEqual(provider.model_list_strategy, "anthropic_compatible")
+
+    def test_list_providers_async_loads_models_dev_before_serializing(self):
+        manager = LLMProviderManager()
+        payload = {
+            "frogbot": {
+                "id": "frogbot",
+                "name": "FrogBot",
+                "npm": "@ai-sdk/openai-compatible",
+                "env": ["FROGBOT_API_KEY"],
+                "api": "https://app.frogbot.ai/api/v1",
+                "models": {},
+            }
+        }
+
+        with patch.object(
+            manager,
+            "get_models_dev_data",
+            AsyncMock(side_effect=lambda force_refresh=False: manager.__dict__.update({"_models_dev_data": payload}) or payload),
+        ) as fetch_mock:
+            providers = asyncio.run(manager.list_providers_async())
+
+        fetch_mock.assert_awaited_once_with(force_refresh=False)
+        self.assertIn("frogbot", {item["id"] for item in providers})
+
+    def test_list_models_uses_dynamic_provider_after_refresh(self):
+        manager = LLMProviderManager()
+        payload = {
+            "frogbot": {
+                "id": "frogbot",
+                "name": "FrogBot",
+                "npm": "@ai-sdk/openai-compatible",
+                "env": ["FROGBOT_API_KEY"],
+                "api": "https://app.frogbot.ai/api/v1",
+                "models": {
+                    "frog-1": {
+                        "name": "Frog 1",
+                        "limit": {"context": 131072},
+                    }
+                },
+            }
+        }
+
+        async def _load_models_dev(force_refresh: bool = False):
+            manager._models_dev_data = payload
+            return payload
+
+        with patch.object(manager, "get_models_dev_data", AsyncMock(side_effect=_load_models_dev)):
+            models = asyncio.run(
+                manager.list_models(
+                    provider_id="frogbot",
+                    api_key="sk-test",
+                    force_refresh=True,
+                )
+            )
+
+        self.assertEqual([item["id"] for item in models], ["frog-1"])
+        self.assertEqual(models[0]["source"], "models.dev")
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_local_setup_llm_provider_prompt.py

@@ -0,0 +1,189 @@
+from __future__ import annotations
+
+import importlib.util
+import unittest
+import uuid
+from pathlib import Path
+from unittest.mock import patch
+
+
+MODULE_PATH = Path(__file__).resolve().parents[1] / "scripts" / "local_setup.py"
+
+
+def load_local_setup_module():
+    module_name = f"moviepilot_local_setup_llm_{uuid.uuid4().hex}"
+    spec = importlib.util.spec_from_file_location(module_name, MODULE_PATH)
+    module = importlib.util.module_from_spec(spec)
+    assert spec and spec.loader
+    spec.loader.exec_module(module)
+    return module
+
+
+class LocalSetupLlmProviderPromptTests(unittest.TestCase):
+    def test_collect_agent_config_prefers_loaded_provider_directory(self):
+        module = load_local_setup_module()
+
+        provider_definitions = [
+            {
+                "id": "frogbot",
+                "name": "FrogBot",
+                "default_base_url": "https://app.frogbot.ai/api/v1",
+                "api_key_label": "API Key",
+            }
+        ]
+        models = [
+            {"id": "frog-1", "name": "Frog 1", "context_tokens_k": 128},
+            {"id": "frog-2", "name": "Frog 2"},
+        ]
+
+        with patch.object(module, "print_step"), patch.object(
+            module, "_prompt_yes_no", side_effect=[True, False, True]
+        ), patch.object(
+            module, "_load_llm_provider_definitions", return_value=provider_definitions
+        ), patch.object(
+            module, "_prompt_provider_choice", return_value="frogbot"
+        ) as provider_prompt, patch.object(
+            module, "_prompt_text", side_effect=["https://override.example.com/v1"]
+        ), patch.object(
+            module, "_prompt_secret_text", return_value="sk-frog"
+        ), patch.object(
+            module, "_load_llm_models", return_value=models
+        ) as load_models, patch.object(
+            module, "_prompt_model_choice", return_value="frog-2"
+        ) as model_prompt, patch.object(
+            module, "read_env_value", return_value=None
+        ), patch.object(
+            module, "_env_default", side_effect=lambda key, default="": default
+        ), patch.object(
+            module, "_env_bool", side_effect=lambda key, default: default
+        ), patch.object(
+            module, "_env_llm_thinking_level_default", return_value="auto"
+        ), patch.object(
+            module, "_prompt_choice", return_value="auto"
+        ):
+            config = module._collect_agent_config(runtime_python=Path("/tmp/runtime-python"))
+
+        provider_prompt.assert_called_once()
+        load_models.assert_called_once_with(
+            provider="frogbot",
+            api_key="sk-frog",
+            base_url="https://override.example.com/v1",
+            runtime_python=Path("/tmp/runtime-python"),
+        )
+        model_prompt.assert_called_once_with(models, default="")
+        self.assertEqual(config["LLM_PROVIDER"], "frogbot")
+        self.assertEqual(config["LLM_MODEL"], "frog-2")
+        self.assertEqual(config["LLM_API_KEY"], "sk-frog")
+        self.assertEqual(config["LLM_BASE_URL"], "https://override.example.com/v1")
+
+    def test_collect_agent_config_falls_back_to_common_provider_choices(self):
+        module = load_local_setup_module()
+
+        with patch.object(module, "print_step"), patch.object(
+            module, "_prompt_yes_no", side_effect=[True, False, True]
+        ), patch.object(
+            module, "_load_llm_provider_definitions", return_value=[]
+        ), patch.object(
+            module, "_prompt_provider_choice", return_value="anthropic"
+        ), patch.object(
+            module, "_prompt_text", side_effect=["https://api.anthropic.com/v1"]
+        ), patch.object(
+            module, "_prompt_secret_text", return_value="sk-anthropic"
+        ), patch.object(
+            module, "_load_llm_models", return_value=[]
+        ), patch.object(
+            module, "_prompt_model_choice", return_value="claude-sonnet-4-0"
+        ), patch.object(
+            module, "read_env_value", return_value=None
+        ), patch.object(
+            module, "_env_default", side_effect=lambda key, default="": default
+        ), patch.object(
+            module, "_env_bool", side_effect=lambda key, default: default
+        ), patch.object(
+            module, "_env_llm_thinking_level_default", return_value="off"
+        ), patch.object(
+            module, "_prompt_choice", return_value="off"
+        ):
+            config = module._collect_agent_config()
+
+        self.assertEqual(config["LLM_PROVIDER"], "anthropic")
+        self.assertEqual(config["LLM_MODEL"], "claude-sonnet-4-0")
+        self.assertEqual(config["LLM_BASE_URL"], "https://api.anthropic.com/v1")
+
+    def test_prompt_model_choice_accepts_index_selection(self):
+        module = load_local_setup_module()
+
+        with patch.object(module, "_print_llm_models") as print_models, patch(
+            "builtins.input", return_value="2"
+        ):
+            model = module._prompt_model_choice(
+                [
+                    {"id": "model-a", "name": "Model A"},
+                    {"id": "model-b", "name": "Model B"},
+                ],
+                default="model-a",
+            )
+
+        print_models.assert_called_once()
+        self.assertEqual(model, "model-b")
+
+    def test_prompt_model_choice_falls_back_to_text_input_when_empty(self):
+        module = load_local_setup_module()
+
+        with patch.object(module, "_prompt_text", return_value="custom-model") as prompt_text:
+            model = module._prompt_model_choice([], default="")
+
+        prompt_text.assert_called_once_with("LLM 模型名称", default="")
+        self.assertEqual(model, "custom-model")
+
+    def test_load_llm_provider_definitions_inner_uses_direct_provider_module_loader(self):
+        module = load_local_setup_module()
+
+        class _FakeManager:
+            async def list_providers_async(self, force_refresh: bool = False):
+                return [{"id": "frogbot", "name": "FrogBot"}]
+
+        class _FakeProviderModule:
+            @staticmethod
+            def LLMProviderManager():
+                return _FakeManager()
+
+        fake_provider_module = _FakeProviderModule()
+
+        with patch.object(
+            module,
+            "_load_llm_provider_module",
+            return_value=fake_provider_module,
+        ) as loader:
+            providers = module._load_llm_provider_definitions_inner()
+
+        loader.assert_called_once_with()
+        self.assertEqual(providers, [{"id": "frogbot", "name": "FrogBot"}])
+
+    def test_llm_provider_choice_map_skips_oauth_only_provider(self):
+        module = load_local_setup_module()
+
+        choices = module._llm_provider_choice_map(
+            [
+                {"id": "chatgpt", "name": "ChatGPT", "supports_api_key": True},
+                {"id": "github-copilot", "name": "GitHub Copilot", "supports_api_key": False},
+            ]
+        )
+
+        self.assertEqual(choices, {"chatgpt": "ChatGPT"})
+
+    def test_prompt_provider_choice_accepts_custom_provider_id(self):
+        module = load_local_setup_module()
+
+        with patch("builtins.input", return_value="my-provider_01"), patch("builtins.print"):
+            provider = module._prompt_provider_choice(
+                "选择 LLM 提供商",
+                {"deepseek": "DeepSeek", "google": "Google"},
+                default="deepseek",
+            )
+
+        self.assertEqual(provider, "my-provider_01")
+
+
+if __name__ == "__main__":
+    unittest.main()