✨ Feat: 增强 LLM、渲染与广播功能并优化性能 (#2071)

* ⚡️ perf(image_utils): 优化图片哈希获取避免阻塞异步

* ✨ feat(llm): 增强 LLM 管理功能，支持纯文本列表输出，优化模型能力识别并新增提供商

- 【LLM 管理器】为 `llm list` 命令添加 `--text` 选项，支持以纯文本格式输出模型列表。
- 【LLM 配置】新增 `OpenRouter` LLM 提供商的默认配置。
- 【模型能力】增强 `get_model_capabilities` 函数的查找逻辑，支持模型名称分段匹配和更灵活的通配符匹配。
- 【模型能力】为 `Gemini` 模型能力注册表使用更通用的通配符模式。
- 【模型能力】新增 `GPT` 系列模型的详细能力定义，包括多模态输入输出和工具调用支持。

* ✨ feat(renderer): 添加 Jinja2 `inline_asset` 全局函数

- 新增 `RendererService._inline_asset_global` 方法，并注册为 Jinja2 全局函数 `inline_asset`。
- 允许模板通过 `{{ inline_asset('@namespace/path/to/asset.svg') }}` 直接内联已注册命名空间下的资源文件内容。
- 主要用于解决内联 SVG 时可能遇到的跨域安全问题。
- 【重构】优化 `ResourceResolver.resolve_asset_uri` 中对命名空间资源 (以 `@` 开头) 的解析逻辑，确保能够正确获取文件绝对路径并返回 URI。
- 改进 `RenderableComponent.get_extra_css`，使其在组件定义 `component_css` 时自动返回该 CSS 内容。
- 清理 `Renderable` 协议和 `RenderableComponent` 基类中已存在方法的 `[新增]` 标记。

* ✨ feat(tag): 添加标签克隆功能

- 新增 `tag clone <源标签名> <新标签名>` 命令，用于复制现有标签。
- 【优化】在 `tag create`, `tag edit --add`, `tag edit --set` 命令中，自动去重传入的群组ID，避免重复关联。

* ✨ feat(broadcast): 实现标签定向广播、强制发送及并发控制

- 【新功能】
  - 新增标签定向广播功能，支持通过 `-t <标签名>` 或 `广播到 <标签名>` 命令向指定标签的群组发送消息
  - 引入广播强制发送模式，允许绕过群组的任务阻断设置
  - 实现广播并发控制，通过配置限制同时发送任务数量，避免API速率限制
  - 优化视频消息处理，支持从URL下载视频内容并作为原始数据发送，提高跨平台兼容性
- 【配置】
  - 添加 `DEFAULT_BROADCAST` 配置项，用于设置群组进群时广播功能的默认开关状态
  - 添加 `BROADCAST_CONCURRENCY_LIMIT` 配置项，用于控制广播时的最大并发任务数

* ✨ feat(renderer): 支持组件变体样式收集

* ✨ feat(tag): 实现群组标签自动清理及手动清理功能

* 🐛 fix(gemini): 增加响应验证以处理内容过滤（promptFeedback）

* 🐛 fix(codeql): 移除对 JavaScript 和 TypeScript 的分析支持

* 🚨 auto fix by pre-commit hooks

---------

Co-authored-by: webjoin111 <455457521@qq.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

zhenxun/services/llm/adapters/gemini.py

@@ -354,6 +354,24 @@ class GeminiAdapter(BaseAdapter):
 
         return safety_settings if safety_settings else None
 
+    def validate_response(self, response_json: dict[str, Any]) -> None:
+        """验证 Gemini API 响应，增加对 promptFeedback 的检查"""
+        super().validate_response(response_json)
+
+        if prompt_feedback := response_json.get("promptFeedback"):
+            if block_reason := prompt_feedback.get("blockReason"):
+                logger.warning(
+                    f"Gemini 内容因 promptFeedback 被安全过滤: {block_reason}"
+                )
+                raise LLMException(
+                    f"内容被安全过滤: {block_reason}",
+                    code=LLMErrorCode.CONTENT_FILTERED,
+                    details={
+                        "block_reason": block_reason,
+                        "safety_ratings": prompt_feedback.get("safetyRatings"),
+                    },
+                )
+
     def parse_response(
         self,
         model: "LLMModel",

zhenxun/services/llm/config/providers.py

@@ -192,10 +192,20 @@ def get_default_providers() -> list[dict[str, Any]]:
             "api_base": "https://generativelanguage.googleapis.com",
             "api_type": "gemini",
             "models": [
-                {"model_name": "gemini-2.0-flash"},
                 {"model_name": "gemini-2.5-flash"},
                 {"model_name": "gemini-2.5-pro"},
-                {"model_name": "gemini-2.5-flash-lite-preview-06-17"},
+                {"model_name": "gemini-2.5-flash-lite"},
+            ],
+        },
+        {
+            "name": "OpenRouter",
+            "api_key": "YOUR_OPENROUTER_API_KEY",
+            "api_base": "https://openrouter.ai/api",
+            "api_type": "openrouter",
+            "models": [
+                {"model_name": "google/gemini-2.5-pro"},
+                {"model_name": "google/gemini-2.5-flash"},
+                {"model_name": "x-ai/grok-4"},
             ],
         },
     ]

zhenxun/services/llm/types/capabilities.py

@@ -9,6 +9,8 @@ import fnmatch
 
 from pydantic import BaseModel, Field
 
+from zhenxun.services.log import logger
+
 
 class ModelModality(str, Enum):
     TEXT = "text"
@@ -50,6 +52,46 @@ GEMINI_IMAGE_GEN_CAPABILITIES = ModelCapabilities(
     supports_tool_calling=True,
 )
 
+GPT_ADVANCED_TEXT_IMAGE_CAPABILITIES = ModelCapabilities(
+    input_modalities={ModelModality.TEXT, ModelModality.IMAGE},
+    output_modalities={ModelModality.TEXT},
+    supports_tool_calling=True,
+)
+
+GPT_MULTIMODAL_IO_CAPABILITIES = ModelCapabilities(
+    input_modalities={ModelModality.TEXT, ModelModality.AUDIO, ModelModality.IMAGE},
+    output_modalities={ModelModality.TEXT, ModelModality.AUDIO},
+    supports_tool_calling=True,
+)
+
+GPT_IMAGE_GENERATION_CAPABILITIES = ModelCapabilities(
+    input_modalities={ModelModality.TEXT, ModelModality.IMAGE},
+    output_modalities={ModelModality.IMAGE},
+    supports_tool_calling=True,
+)
+
+GPT_VIDEO_GENERATION_CAPABILITIES = ModelCapabilities(
+    input_modalities={ModelModality.TEXT, ModelModality.IMAGE, ModelModality.VIDEO},
+    output_modalities={ModelModality.VIDEO},
+    supports_tool_calling=True,
+)
+
+DEFAULT_PERMISSIVE_CAPABILITIES = ModelCapabilities(
+    input_modalities={
+        ModelModality.TEXT,
+        ModelModality.IMAGE,
+        ModelModality.AUDIO,
+        ModelModality.VIDEO,
+    },
+    output_modalities={
+        ModelModality.TEXT,
+        ModelModality.IMAGE,
+        ModelModality.AUDIO,
+        ModelModality.VIDEO,
+    },
+    supports_tool_calling=True,
+)
+
 
 DOUBAO_ADVANCED_MULTIMODAL_CAPABILITIES = ModelCapabilities(
     input_modalities={ModelModality.TEXT, ModelModality.IMAGE, ModelModality.VIDEO},
@@ -91,11 +133,8 @@ MODEL_CAPABILITIES_REGISTRY: dict[str, ModelCapabilities] = {
         is_embedding_model=True,
     ),
     "*gemini-*-image-preview*": GEMINI_IMAGE_GEN_CAPABILITIES,
-    "gemini-2.5-pro*": GEMINI_CAPABILITIES,
-    "gemini-1.5-pro*": GEMINI_CAPABILITIES,
-    "gemini-2.5-flash*": GEMINI_CAPABILITIES,
-    "gemini-2.0-flash*": GEMINI_CAPABILITIES,
-    "gemini-1.5-flash*": GEMINI_CAPABILITIES,
+    "gemini-*-pro*": GEMINI_CAPABILITIES,
+    "gemini-*-flash*": GEMINI_CAPABILITIES,
     "GLM-4V-Flash": ModelCapabilities(
         input_modalities={ModelModality.TEXT, ModelModality.IMAGE},
         output_modalities={ModelModality.TEXT},
@@ -112,6 +151,13 @@ MODEL_CAPABILITIES_REGISTRY: dict[str, ModelCapabilities] = {
     "doubao-1-5-thinking-vision-pro": DOUBAO_ADVANCED_MULTIMODAL_CAPABILITIES,
     "deepseek-chat": STANDARD_TEXT_TOOL_CAPABILITIES,
     "deepseek-reasoner": STANDARD_TEXT_TOOL_CAPABILITIES,
+    "gpt-5*": GPT_ADVANCED_TEXT_IMAGE_CAPABILITIES,
+    "gpt-4.1*": GPT_ADVANCED_TEXT_IMAGE_CAPABILITIES,
+    "gpt-4o*": GPT_MULTIMODAL_IO_CAPABILITIES,
+    "o3*": GPT_ADVANCED_TEXT_IMAGE_CAPABILITIES,
+    "o4-mini*": GPT_ADVANCED_TEXT_IMAGE_CAPABILITIES,
+    "gpt image*": GPT_IMAGE_GENERATION_CAPABILITIES,
+    "sora*": GPT_VIDEO_GENERATION_CAPABILITIES,
 }
 
 
@@ -126,11 +172,25 @@ def get_model_capabilities(model_name: str) -> ModelCapabilities:
             canonical_name = c_name
             break
 
-    if canonical_name in MODEL_CAPABILITIES_REGISTRY:
-        return MODEL_CAPABILITIES_REGISTRY[canonical_name]
+    parts = canonical_name.split("/")
+    names_to_check = ["/".join(parts[i:]) for i in range(len(parts))]
 
-    for pattern, capabilities in MODEL_CAPABILITIES_REGISTRY.items():
-        if "*" in pattern and fnmatch.fnmatch(model_name, pattern):
-            return capabilities
+    logger.trace(f"为 '{model_name}' 生成的检查列表: {names_to_check}")
 
-    return ModelCapabilities()
+    for name in names_to_check:
+        if name in MODEL_CAPABILITIES_REGISTRY:
+            logger.debug(f"模型 '{model_name}' 通过精确匹配 '{name}' 找到能力定义。")
+            return MODEL_CAPABILITIES_REGISTRY[name]
+
+        for pattern, capabilities in MODEL_CAPABILITIES_REGISTRY.items():
+            if "*" in pattern and fnmatch.fnmatch(name, pattern):
+                logger.debug(
+                    f"模型 '{model_name}' 通过通配符匹配 '{name}'(pattern: '{pattern}')"
+                    f"找到能力定义。"
+                )
+                return capabilities
+
+    logger.warning(
+        f"模型 '{model_name}' 的能力定义未在注册表中找到，将使用默认的'全功能'回退配置"
+    )
+    return DEFAULT_PERMISSIVE_CAPABILITIES

zhenxun/services/renderer/protocols.py

@@ -40,7 +40,7 @@ class Renderable(ABC):
     @abstractmethod
     def get_children(self) -> Iterable["Renderable"]:
         """
-        [新增] 返回一个包含所有直接子组件的可迭代对象。
+        返回一个包含所有直接子组件的可迭代对象。
 
         这使得渲染服务能够递归地遍历整个组件树，以执行依赖收集（CSS、JS）等任务。
         非容器组件应返回一个空列表。

zhenxun/services/renderer/service.py

@@ -75,6 +75,7 @@ class RendererService:
         self._custom_globals: dict[str, Callable] = {}
 
         self.filter("dump_json")(self._pydantic_tojson_filter)
+        self.global_function("inline_asset")(self._inline_asset_global)
 
     def _create_jinja_env(self) -> Environment:
         """
@@ -176,9 +177,24 @@ class RendererService:
 
         return decorator
 
+    async def _inline_asset_global(self, namespaced_path: str) -> str:
+        """
+        一个Jinja2全局函数，用于读取并内联一个已注册命名空间下的资源文件内容。
+        主要用于内联SVG，以解决浏览器的跨域安全问题。
+        """
+        if not self._jinja_env or not self._jinja_env.loader:
+            return f"<!-- Error: Jinja env not ready for {namespaced_path} -->"
+        try:
+            source, _, _ = self._jinja_env.loader.get_source(
+                self._jinja_env, namespaced_path
+            )
+            return source
+        except TemplateNotFound:
+            return f"<!-- Asset not found: {namespaced_path} -->"
+
     async def initialize(self):
         """
-        [新增] 延迟初始化方法，在 on_startup 钩子中调用。
+        延迟初始化方法，在 on_startup 钩子中调用。
 
         负责初始化截图引擎和主题管理器，确保在首次渲染前所有依赖都已准备就绪。
         使用锁来防止并发初始化。
@@ -223,27 +239,36 @@ class RendererService:
         )
 
         style_paths_to_load = []
-        if manifest and "styles" in manifest:
-            styles = (
-                [manifest["styles"]]
-                if isinstance(manifest["styles"], str)
-                else manifest["styles"]
-            )
-            for style_path in styles:
-                full_style_path = str(Path(component_path_base) / style_path).replace(
-                    "\\", "/"
+        if manifest and manifest.get("styles"):
+            styles = manifest["styles"]
+            styles = [styles] if isinstance(styles, str) else styles
+
+            resolution_base_path = Path(component_path_base)
+            if variant:
+                skin_manifest_path = str(Path(component_path_base) / "skins" / variant)
+                skin_manifest = await context.theme_manager._load_single_manifest(
+                    skin_manifest_path
                 )
-                style_paths_to_load.append(full_style_path)
+                if skin_manifest and "styles" in skin_manifest:
+                    resolution_base_path = Path(skin_manifest_path)
+
+            style_paths_to_load.extend(
+                str(resolution_base_path / style).replace("\\", "/") for style in styles
+            )
         else:
-            resolved_template_name = (
+            base_template_path = (
                 await context.theme_manager._resolve_component_template(
                     component, context
                 )
             )
-            conventional_style_path = str(
-                Path(resolved_template_name).with_name("style.css")
+            base_style_path = str(
+                Path(base_template_path).with_name("style.css")
             ).replace("\\", "/")
-            style_paths_to_load.append(conventional_style_path)
+            style_paths_to_load.append(base_style_path)
+
+            if variant:
+                skin_style_path = f"{component_path_base}/skins/{variant}/style.css"
+                style_paths_to_load.append(skin_style_path)
 
         for css_template_path in style_paths_to_load:
             try:

zhenxun/services/renderer/theme.py

@@ -172,24 +172,45 @@ class ResourceResolver:
 
         if asset_path.startswith("@"):
             try:
-                full_asset_path = self.theme_manager.jinja_env.join_path(
-                    asset_path, current_template_name
-                )
-                _source, file_abs_path, _uptodate = (
-                    self.theme_manager.jinja_env.loader.get_source(
-                        self.theme_manager.jinja_env, full_asset_path
+                if "/" not in asset_path:
+                    raise TemplateNotFound(f"无效的命名空间路径: {asset_path}")
+
+                namespace, rel_path = asset_path.split("/", 1)
+
+                loader = self.theme_manager.jinja_env.loader
+                if (
+                    isinstance(loader, ChoiceLoader)
+                    and loader.loaders
+                    and isinstance(loader.loaders[0], PrefixLoader)
+                ):
+                    prefix_loader = loader.loaders[0]
+                    if namespace in prefix_loader.mapping:
+                        loader_for_namespace = prefix_loader.mapping[namespace]
+                        if isinstance(loader_for_namespace, FileSystemLoader):
+                            base_path = Path(loader_for_namespace.searchpath[0])
+                            file_abs_path = (base_path / rel_path).resolve()
+
+                            if file_abs_path.is_file():
+                                logger.debug(
+                                    f"Resolved namespaced asset"
+                                    f" '{asset_path}' -> '{file_abs_path}'"
+                                )
+                                return file_abs_path.as_uri()
+                            else:
+                                raise TemplateNotFound(asset_path)
+                        else:
+                            raise TemplateNotFound(
+                                f"Unsupported loader type for namespace '{namespace}'."
+                            )
+                    else:
+                        raise TemplateNotFound(f"Namespace '{namespace}' not found.")
+                else:
+                    raise TemplateNotFound(
+                        f"无法解析命名空间资源 '{asset_path}'，加载器结构不符合预期。"
                     )
-                )
-                if file_abs_path:
-                    logger.debug(
-                        f"Jinja Loader resolved asset '{asset_path}'->'{file_abs_path}'"
-                    )
-                    return Path(file_abs_path).absolute().as_uri()
+
             except TemplateNotFound:
-                logger.warning(
-                    f"资源文件在命名空间中未找到: '{asset_path}'"
-                    f"(在模板 '{current_template_name}' 中引用)"
-                )
+                logger.warning(f"资源文件在命名空间中未找到: '{asset_path}'")
                 return ""
 
         search_paths: list[tuple[str, Path]] = []

zhenxun/services/tags/manager.py

@@ -9,6 +9,7 @@ from typing import Any, ClassVar
 
 from aiocache import Cache, cached
 from arclet.alconna import Alconna, Args
+import nonebot
 from nonebot.adapters import Bot
 from tortoise.exceptions import IntegrityError
 from tortoise.expressions import Q
@@ -156,8 +157,9 @@ class TagManager:
                 dynamic_rule=dynamic_rule,
             )
             if group_ids:
+                unique_group_ids = list(dict.fromkeys(group_ids))
                 await GroupTagLink.bulk_create(
-                    [GroupTagLink(tag=tag, group_id=gid) for gid in group_ids]
+                    [GroupTagLink(tag=tag, group_id=gid) for gid in unique_group_ids]
                 )
             return tag
 
@@ -175,6 +177,49 @@ class TagManager:
         deleted_count = await GroupTag.filter(name=name).delete()
         return deleted_count > 0
 
+    @invalidate_on_change
+    async def remove_group_from_all_tags(self, group_id: str) -> int:
+        """
+        从所有静态标签中移除一个指定的群组ID。
+        主要用于机器人退群时的实时清理。
+
+        参数：
+            group_id: 要移除的群组ID。
+
+        返回：
+            被删除的关联数量。
+        """
+        deleted_count = await GroupTagLink.filter(group_id=group_id).delete()
+        if deleted_count > 0:
+            logger.info(f"已从 {deleted_count} 个标签中移除群组 {group_id} 的关联。")
+        return deleted_count
+
+    @invalidate_on_change
+    async def prune_stale_group_links(self) -> int:
+        """
+        清理所有静态标签中无效的群组关联。
+        无效指的是机器人已不再任何一个已连接的Bot的群组列表中。
+
+        返回：
+            被清理的无效关联的总数。
+        """
+        all_bot_group_ids = set()
+        for bot in nonebot.get_bots().values():
+            groups, _ = await PlatformUtils.get_group_list(bot)
+            all_bot_group_ids.update(g.group_id for g in groups if g.group_id)
+
+        all_static_links = await GroupTagLink.filter(tag__tag_type="STATIC").all()
+
+        stale_link_ids = [
+            link.id
+            for link in all_static_links
+            if link.group_id not in all_bot_group_ids
+        ]
+
+        if stale_link_ids:
+            return await GroupTagLink.filter(id__in=stale_link_ids).delete()
+        return 0
+
     @invalidate_on_change
     async def add_groups_to_tag(self, name: str, group_ids: list[str]) -> int:  # type: ignore
         """
@@ -186,11 +231,12 @@ class TagManager:
         if tag.tag_type == "DYNAMIC":
             raise ValueError("不能向动态标签手动添加群组。")
 
+        unique_group_ids = list(dict.fromkeys(group_ids))
         await GroupTagLink.bulk_create(
-            [GroupTagLink(tag=tag, group_id=gid) for gid in group_ids],
+            [GroupTagLink(tag=tag, group_id=gid) for gid in unique_group_ids],
             ignore_conflicts=True,
         )
-        return len(group_ids)
+        return len(unique_group_ids)
 
     @invalidate_on_change
     async def remove_groups_from_tag(self, name: str, group_ids: list[str]) -> int:
@@ -205,6 +251,72 @@ class TagManager:
         ).delete()
         return deleted_count
 
+    @invalidate_on_change
+    async def clone_tag(
+        self,
+        source_name: str,
+        new_name: str,
+        bot: Bot,
+        add_groups: list[str] | None = None,
+        remove_groups: list[str] | None = None,
+        as_dynamic: bool = False,
+        description: str | None = None,
+        mode: str | None = None,
+    ) -> GroupTag:
+        """
+        克隆一个标签，支持动态转静态、修改群组等。
+        """
+        source_tag = await GroupTag.get_or_none(name=source_name)
+        if not source_tag:
+            raise ValueError(f"源标签 '{source_name}' 不存在。")
+
+        if await GroupTag.exists(name=new_name):
+            raise IntegrityError(f"目标标签 '{new_name}' 已存在。")
+
+        tag_type = "STATIC"
+        group_ids_to_set: list[str] | None = None
+        dynamic_rule: str | dict | None = None
+
+        if source_tag.tag_type == "STATIC":
+            if as_dynamic:
+                raise ValueError("不能将静态标签克隆为动态标签。")
+            group_ids_to_set = await GroupTagLink.filter(tag=source_tag).values_list(  # type: ignore
+                "group_id", flat=True
+            )
+        else:
+            if as_dynamic:
+                tag_type = "DYNAMIC"
+                dynamic_rule = source_tag.dynamic_rule
+                if add_groups or remove_groups:
+                    raise ValueError(
+                        "克隆为动态标签时，不支持 --add 或 --remove 操作。"
+                    )
+            else:
+                group_ids_to_set = await self.resolve_tag_to_group_ids(
+                    source_name, bot=bot
+                )
+
+        if group_ids_to_set is not None:
+            final_group_set = set(group_ids_to_set)
+            if add_groups:
+                final_group_set.update(add_groups)
+            if remove_groups:
+                final_group_set.difference_update(remove_groups)
+            group_ids_to_set = list(final_group_set)
+
+        is_blacklist = (
+            (mode == "black") if mode is not None else source_tag.is_blacklist
+        )
+
+        return await self.create_tag(
+            name=new_name,
+            is_blacklist=is_blacklist,
+            description=description,
+            group_ids=group_ids_to_set,
+            tag_type=tag_type,
+            dynamic_rule=dynamic_rule,
+        )
+
     async def list_tags_with_counts(self) -> list[dict]:
         """列出所有标签及其关联的群组数量。"""
         tags = await GroupTag.all().prefetch_related("groups")
@@ -514,11 +626,13 @@ class TagManager:
             raise ValueError("不能为动态标签设置静态群组列表。")
         async with in_transaction():
             await GroupTagLink.filter(tag=tag).delete()
-            await GroupTagLink.bulk_create(
-                [GroupTagLink(tag=tag, group_id=gid) for gid in group_ids],
-                ignore_conflicts=True,
-            )
-        return len(group_ids)
+            unique_group_ids = list(dict.fromkeys(group_ids))
+            if unique_group_ids:
+                await GroupTagLink.bulk_create(
+                    [GroupTagLink(tag=tag, group_id=gid) for gid in unique_group_ids],
+                    ignore_conflicts=True,
+                )
+        return len(unique_group_ids)
 
     @invalidate_on_change
     async def clear_all_tags(self) -> int: