""" schemas.py - 数据模型和 API Schema 定义 职责: - 定义所有数据结构的类型(TypedDict) - 定义 API 请求/响应模型(Pydantic) - 提供类型检查支持 """ import copy from typing import TypedDict, List, Optional, Literal, Dict, Any, Union try: # Python 3.11+ from typing import NotRequired except ImportError: # Python 3.10(最低支持版本,见 start.py MIN_PYTHON) from typing_extensions import NotRequired from pydantic import BaseModel, ConfigDict # ================= 动作类型 ================= ActionType = Literal[ "FILL_INPUT", "SELECT_MODEL", "CLICK", "COORD_CLICK", "COORD_SCROLL", "STREAM_WAIT", "STREAM_OUTPUT", "KEY_PRESS", "WAIT", "JS_EXEC", "READONLY_HINT", "PAGE_FETCH", "SET", "CAPTURE", "IF", "SWITCH", "GROUP", "GUARD", "TRY", "LABEL", ] # ================= 选择器字段名称 ================= REQUIRED_SELECTOR_KEYS = [ "input_box", "send_btn", "result_container", ] OPTIONAL_SELECTOR_KEYS = [ "new_chat_btn", "retry_send_btn", "stop_btn", "message_wrapper", "generating_indicator", "upload_btn", "file_input", "drop_zone", "composer_root", ] ALL_SELECTOR_KEYS = REQUIRED_SELECTOR_KEYS + OPTIONAL_SELECTOR_KEYS # ================= 工作流步骤 ================= class WorkflowStep(TypedDict): """工作流单步定义""" action: ActionType target: str optional: bool value: Optional[Any] execution: NotRequired[Dict[str, Any]] selector: NotRequired[str] label: NotRequired[str] flow_version: NotRequired[int] retry_safe: NotRequired[bool] # ================= 元素定义 ================= class SelectorDefinition(TypedDict): """选择器定义 - 用于 AI 分析页面时的查找目标""" key: str description: str enabled: bool required: bool class GlobalConfig(TypedDict, total=False): """全局配置 - 存储在 sites.json 的 _global 节点""" selector_definitions: List[SelectorDefinition] # ================= 默认元素定义 ================= DEFAULT_SELECTOR_DEFINITIONS: List[SelectorDefinition] = [ { "key": "input_box", "description": "用户输入文本的输入框(textarea 或 contenteditable 元素)", "enabled": True, "required": True }, { "key": "send_btn", "description": "发送消息的按钮(通常是 type=submit 或带有发送图标的按钮)", "enabled": True, "required": True }, { "key": "result_container", "description": "AI 回复内容的容器(仅包含 AI 的输出文本,不含用户消息)", "enabled": True, "required": True }, { "key": "new_chat_btn", "description": "新建对话的按钮(点击后开始新的对话)", "enabled": True, "required": False }, { "key": "retry_send_btn", "description": "重新运行/重试当前回复的按钮(通常是回复卡片顶部的圆形箭头)", "enabled": False, "required": False }, { "key": "stop_btn", "description": "停止/打断大模型输出的按钮(通常在生成过程中显示)", "enabled": False, "required": False }, { "key": "message_wrapper", "description": "消息完整容器(包裹单条消息的外层元素,用于多节点拼接)", "enabled": False, "required": False }, { "key": "generating_indicator", "description": "生成中指示器(如停止按钮、加载动画,用于检测是否还在输出)", "enabled": False, "required": False }, { "key": "upload_btn", "description": "打开文件选择器的上传按钮(点击后通常会弹出原生选文件)", "enabled": False, "required": False }, { "key": "file_input", "description": "原生文件输入框(input[type=file]),用于直接注入文件", "enabled": False, "required": False }, { "key": "composer_root", "description": "当前消息编辑区根节点,限定附件输入框和就绪观察范围", "enabled": False, "required": False }, { "key": "drop_zone", "description": "支持拖拽上传的区域(某些站点不支持粘贴但支持拖拽)", "enabled": False, "required": False } ] def get_default_selector_definitions() -> List[SelectorDefinition]: """获取默认的元素定义列表(深拷贝)""" import copy return copy.deepcopy(DEFAULT_SELECTOR_DEFINITIONS) # ================= 图片提取相关模型(Phase A 新增)================= class ImageData(BaseModel): """ 图片数据模型 kind 决定使用 url 还是 data_uri: - kind="url": 使用 url 字段 - kind="data_uri": 使用 data_uri 字段 """ kind: Literal["url", "data_uri"] url: Optional[str] = None data_uri: Optional[str] = None mime: Optional[str] = None byte_size: Optional[int] = None alt: Optional[str] = None width: Optional[int] = None height: Optional[int] = None index: int = 0 detected_at: Optional[str] = None # ISO 格式时间戳 source: Optional[Literal["currentSrc", "src", "blob", "data_uri", "relative"]] = None model_config = ConfigDict( json_schema_extra={ "example": { "kind": "url", "url": "https://example.com/image.png", "mime": "image/png", "width": 800, "height": 600, "index": 0, "source": "currentSrc" } } ) class FilePasteConfig(TypedDict, total=False): """ 文件粘贴配置 当文本长度超过阈值时,将文本写入临时 txt/pdf 文件并上传, 按 chunk 类型分块连续发送,或按 error 类型直接返回自定义错误信息。 上传文件后,自动在输入框中追加一句引导文本,确保能正常发送。 附件发送确认规则也优先挂在这里,避免和网络监听配置混在一起。 用于 sites.json 中的 file_paste 字段 """ attachments: Dict[str, Any] # 通用附件能力、限额、上传策略(独立于超长文本开关) enabled: bool # 是否启用文件粘贴模式,默认 False threshold: int # 字符数阈值,超过此值时使用文件粘贴,默认 50000 temp_file_type: Literal["txt", "pdf", "chunk", "error"] # 临时文件类型/超长处理策略,默认 txt hint_text: str # 上传后引导文本;error 类型时作为自定义错误信息 txt_hint_text: str # TXT 临时文件引导文本 pdf_hint_text: str # PDF 临时文件引导文本 error_hint_text: str # ERROR 类型错误信息 reacquire_input_after_upload: bool # 上传完成后是否重新定位输入框 post_upload_input_selector: str # 上传后专用输入框选择器 post_upload_settle: float # 上传完成后额外稳定等待 upload_signal_timeout: float # 上传信号确认超时 upload_signal_grace: float # 弱信号额外宽限 send_confirmation: Dict[str, Any] # 附件发送后的成功判定策略 attachment_monitor: Dict[str, Any] # 附件预览 / pending / busy 规则 state_probe: Dict[str, Any] # 自定义 JS 状态探针 def get_default_file_paste_config() -> 'FilePasteConfig': """获取默认的文件粘贴配置""" return { "enabled": False, "threshold": 50000, "temp_file_type": "txt", "hint_text": "完全专注于文件内容", "txt_hint_text": "完全专注于文件内容", "pdf_hint_text": "完全专注于文件内容", "error_hint_text": "输入文本长度超过限制,已中止发送", "reacquire_input_after_upload": False, "post_upload_input_selector": "", "post_upload_settle": 0.0, "upload_signal_timeout": 2.5, "upload_signal_grace": 3.0, "state_probe": { "enabled": False, "code": "", }, } class PromptPaddingConfig(TypedDict, total=False): """ 提示词开头注入配置 用于 sites.json 中的 prompt_padding 字段 """ enabled: bool marker_text: str segments_per_side: int random_insert_enabled: bool random_insert_chars: str def get_default_prompt_padding_config() -> 'PromptPaddingConfig': """获取默认的提示词开头注入配置""" return { "enabled": False, "marker_text": "测试号,无实际意义", "segments_per_side": 12, "random_insert_enabled": False, "random_insert_chars": "", } ModalityRunPolicy = Literal[ "disabled", "generic_only", "on_signal", "probe_if_trigger_found", "always_probe", ] class ModalityPolicyConfig(TypedDict, total=False): """单个模态的运行策略。""" enabled: bool run_policy: ModalityRunPolicy selector: str container_selector: Optional[str] quick_probe_timeout_seconds: float capture_timeout_seconds: float late_wait_timeout_seconds: float blind_wait_timeout_seconds: float class ExtractionModalitiesConfig(TypedDict, total=False): """多模态提取开关与运行策略。旧版 bool 配置仍兼容。""" image: Union[bool, ModalityPolicyConfig] audio: Union[bool, ModalityPolicyConfig] video: Union[bool, ModalityPolicyConfig] class AudioNetworkCaptureConfig(TypedDict, total=False): """页面内网络音频捕获配置""" enabled: bool timeout_seconds: float transport: Literal["page_websocket_probe"] url_patterns: List[str] extractor: Literal["voicegenie_ogg_pages", "voicegenie_binary_stream"] settle_seconds: float max_payload_bytes: int class AudioBrowserTtsFallbackConfig(TypedDict, total=False): """浏览器上下文内直接拉取 TTS 的兜底配置。""" enabled: bool provider: Literal["doubao_samantha"] speaker: str speech_rate: int pitch: int format: Literal["aac"] timeout_seconds: float pc_version: str aid: str real_aid: str language: str device_platform: str pkg_type: str region: str sys_region: str use_olympus_account: str samantha_web: str MODALITY_TYPES = ("image", "audio", "video") MODALITY_RUN_POLICY_VALUES = { "disabled", "generic_only", "on_signal", "probe_if_trigger_found", "always_probe", } def _coerce_bool(value: Any, default: bool = False) -> bool: if isinstance(value, bool): return value if value is None: return default if isinstance(value, str): lowered = value.strip().lower() if lowered in {"1", "true", "yes", "y", "on"}: return True if lowered in {"0", "false", "no", "n", "off"}: return False return default if isinstance(value, (int, float)): return value != 0 return bool(value) def get_default_modality_policy(media_type: str, enabled: bool = False) -> ModalityPolicyConfig: """获取单模态默认运行策略。""" media_type = str(media_type or "").strip().lower() policy: ModalityPolicyConfig = { "enabled": bool(enabled), "run_policy": "disabled" if not enabled else "on_signal", "quick_probe_timeout_seconds": 1.0, } if media_type == "audio": policy.update({ "run_policy": "probe_if_trigger_found" if enabled else "disabled", "capture_timeout_seconds": 12.0, }) elif media_type == "video": policy.update({ "run_policy": "on_signal" if enabled else "disabled", "late_wait_timeout_seconds": 90.0, }) elif media_type == "image": policy.update({ "run_policy": "on_signal" if enabled else "disabled", "late_wait_timeout_seconds": 45.0, "blind_wait_timeout_seconds": 1.0, }) return policy def normalize_modality_policy(media_type: str, raw_value: Any) -> ModalityPolicyConfig: """把旧 bool 或新对象配置规范化成策略对象。""" if isinstance(raw_value, dict): enabled = _coerce_bool(raw_value.get("enabled"), False) result = get_default_modality_policy(media_type, enabled=enabled) run_policy = str(raw_value.get("run_policy") or result.get("run_policy") or "").strip().lower() if run_policy not in MODALITY_RUN_POLICY_VALUES: run_policy = str(result.get("run_policy") or "disabled") if not enabled: run_policy = "disabled" result["run_policy"] = run_policy # type: ignore[typeddict-item] if "quick_probe_timeout_seconds" in raw_value: try: result["quick_probe_timeout_seconds"] = max(0.1, min(float(raw_value["quick_probe_timeout_seconds"]), 10.0)) except (TypeError, ValueError): pass if "capture_timeout_seconds" in raw_value: try: result["capture_timeout_seconds"] = max(0.2, min(float(raw_value["capture_timeout_seconds"]), 180.0)) except (TypeError, ValueError): pass if "late_wait_timeout_seconds" in raw_value: try: result["late_wait_timeout_seconds"] = max(0.2, min(float(raw_value["late_wait_timeout_seconds"]), 300.0)) except (TypeError, ValueError): pass if "blind_wait_timeout_seconds" in raw_value: try: result["blind_wait_timeout_seconds"] = max(0.0, min(float(raw_value["blind_wait_timeout_seconds"]), 300.0)) except (TypeError, ValueError): pass if "selector" in raw_value: selector = str(raw_value.get("selector") or "").strip() if selector: result["selector"] = selector if "container_selector" in raw_value: container_selector = str(raw_value.get("container_selector") or "").strip() result["container_selector"] = container_selector or None return result return get_default_modality_policy(media_type, enabled=_coerce_bool(raw_value, False)) def normalize_modalities_config(raw_modalities: Any) -> ExtractionModalitiesConfig: """规范化 image/audio/video 三个模态配置。""" raw = raw_modalities if isinstance(raw_modalities, dict) else {} return { media_type: normalize_modality_policy(media_type, raw.get(media_type, False)) for media_type in MODALITY_TYPES } def get_modality_policy(modalities: Any, media_type: str) -> ModalityPolicyConfig: """读取单模态策略,兼容旧 bool。""" raw = modalities.get(media_type) if isinstance(modalities, dict) else False return normalize_modality_policy(media_type, raw) def is_modality_enabled(modalities: Any, media_type: str) -> bool: """判断模态是否启用,避免 bool(dict) 误判。""" return bool(get_modality_policy(modalities, media_type).get("enabled", False)) def get_modality_run_policy(modalities: Any, media_type: str) -> str: """读取模态运行策略。""" return str(get_modality_policy(modalities, media_type).get("run_policy") or "disabled") def get_enabled_modalities(modalities: Any) -> List[str]: """返回已启用模态列表。""" return [ media_type for media_type in MODALITY_TYPES if is_modality_enabled(modalities, media_type) ] class ImageExtractionConfig(TypedDict, total=False): """ 多模态提取配置 用于 sites.json 中的 image_extraction 字段 """ enabled: bool # 是否启用多模态提取(兼容旧字段) modalities: ExtractionModalitiesConfig # 各模态开关 selector: str # 图片选择器,默认 "img" audio_selector: str # 音频选择器 video_selector: str # 视频选择器 container_selector: Optional[str] # 容器选择器,限定查找范围 final_target_strategy: Literal["container", "latest_reply", "latest_visual_reply"] # 最终提取时锁定整容器或当前回复节点 latest_visual_column: Literal["left", "right"] # latest_visual_reply 同一行内优先左栏或右栏 allow_container_fallback: bool # 当前回复内无媒体时是否回退到容器/整页 request_baseline_exclude_existing_nodes: bool # 发送前已存在的图片节点即使换源也视为输入/历史图片 force_postprocess: bool # 是否强制执行收尾多模态后处理(由配置显式声明) arena_image_generation: bool # 启用 Arena 生图的严格完成条件 direct_postprocess_modalities: List[Literal["image", "audio", "video"]] # 允许无额外信号直接执行 DOM 收尾提取的模态 debounce_seconds: float # 文本稳定后等待时间 wait_for_load: bool # 是否等待媒体加载完成 load_timeout_seconds: float # 等待加载的超时时间 download_blobs: bool # 是否下载 blob 转 data_uri max_size_mb: int # blob 最大允许大小(MB) src_allow_patterns: List[str] # 可选:按 src 正则白名单过滤 mode: Literal["all", "first", "last"] # 每种模态的提取模式 audio_capture_enabled: bool # 是否启用通用页面播放音频捕获回退 audio_capture_mute_playback: bool # 捕获时是否静音页面播放 audio_capture_preload_enabled: bool # 是否在页面加载前预注入捕获脚本 audio_capture_reload_before_workflow: bool # 捕获前是否刷新页面以接管早期音频图 audio_capture_preserve_graph: bool # 重置捕获时是否保留已接管的音频图 audio_capture_terminal_settle_seconds: float # 播放结束后额外等待一小段时间,避免截断尾音 audio_trigger_selector: str # 可选:朗读/播放按钮选择器 audio_trigger_labels: List[str] # 可选:朗读/播放按钮文本候选 audio_capture_max_wait_seconds: float # 最长等待音频播放完成 audio_capture_min_wait_seconds: float # 按文本估算等待时的最短窗口 audio_capture_hard_max_wait_seconds: float # 按文本估算等待时的硬上限 audio_capture_estimated_chars_per_second: float # 按文本估算朗读时长的字符/秒 audio_capture_wait_padding_seconds: float # 按文本估算等待时额外冗余 audio_network_capture: AudioNetworkCaptureConfig # 页面内网络音频捕获配置 audio_browser_tts_fallback: AudioBrowserTtsFallbackConfig # 浏览器页内直接拉 TTS 的兜底 audio_capture_poll_seconds: float # 捕获状态轮询间隔 audio_capture_silence_seconds: float # 检测到静默后结束捕获 audio_capture_activity_threshold: float # 音量活动检测阈值 audio_capture_activity_silence_seconds: float # 音量静默多久后结束捕获 # ================= 流式监控配置 ================= class SendConfirmationConfig(TypedDict, total=False): """Post-click send confirmation strategy.""" attachment_sensitivity: Literal["low", "medium", "high"] post_click_observe_window: float pre_retry_probe_window: float retry_observe_window: float attachment_observe_window: float max_retry_count: int retry_interval: float retry_cooldown_window: float retry_action: Literal["click_send_btn", "key_press"] retry_key_combo: str retry_on_unconfirmed_send: bool accept_attachment_change: bool accept_attachment_disappear: bool accept_probe_confirmation: bool retry_block_on_stop_button: bool retry_block_if_generating: bool trust_network_activity: bool trust_generating_indicator: bool trust_send_disabled_with_input_shrink: bool class AttachmentMonitorConfig(TypedDict, total=False): """Per-site attachment readiness heuristics.""" root_selectors: List[str] attachment_selectors: List[str] pending_selectors: List[str] status_selectors: List[str] content_exclusion_selectors: List[str] busy_text_markers: List[str] ignored_busy_text_markers: List[str] send_button_disabled_markers: List[str] require_attachment_present: bool require_upload_signal_before_ready: bool continue_once_on_unconfirmed_send: bool idle_timeout: float hard_max_wait: float class NetworkConfig(TypedDict, total=False): """Network stream capture configuration.""" listen_pattern: str stream_match_mode: Literal["keyword", "regex"] stream_match_pattern: str parser: str silence_threshold: float response_interval: float class StreamConfig(TypedDict, total=False): """流式监控配置(可选字段)""" mode: Literal["dom", "network"] request_transport: Dict[str, Any] hard_timeout: float network: NetworkConfig send_confirmation: SendConfirmationConfig attachment_monitor: AttachmentMonitorConfig # ================= 站点高级配置 ================= class SiteAdvancedConfig(TypedDict, total=False): """站点/预设高级功能配置。""" independent_cookies: bool independent_cookies_auto_takeover: bool input_box_stability_wait_enabled: bool input_box_stability_wait_after_new_chat_only: bool input_box_stability_wait_timeout: float url_transition_wait_on_new_chat: bool url_transition_wait_patterns: List[str] send_confirmation_check_enabled: bool send_confirmation_check_timeout: float skip_new_chat_on_retry: bool SITE_ADVANCED_FIELDS = { "independent_cookies", "independent_cookies_auto_takeover", } PRESET_ADVANCED_FIELDS = { "input_box_stability_wait_enabled", "input_box_stability_wait_after_new_chat_only", "input_box_stability_wait_timeout", "url_transition_wait_on_new_chat", "url_transition_wait_patterns", "send_confirmation_check_enabled", "send_confirmation_check_timeout", "skip_new_chat_on_retry", } ADVANCED_FIELDS = SITE_ADVANCED_FIELDS | PRESET_ADVANCED_FIELDS def get_default_site_advanced_config() -> 'SiteAdvancedConfig': """获取默认站点高级配置""" return { "independent_cookies": False, "independent_cookies_auto_takeover": False, "input_box_stability_wait_enabled": False, "input_box_stability_wait_after_new_chat_only": True, "input_box_stability_wait_timeout": 1.5, "url_transition_wait_on_new_chat": False, "url_transition_wait_patterns": [], "send_confirmation_check_enabled": False, "send_confirmation_check_timeout": 1.5, "skip_new_chat_on_retry": False, } # ================= 站点配置 ================= class SiteConfig(TypedDict, total=False): """站点配置结构""" advanced: SiteAdvancedConfig selectors: Dict[str, Optional[str]] workflow: List[WorkflowStep] stealth: bool stream_config: StreamConfig image_extraction: ImageExtractionConfig # 🆕 新增 file_paste: FilePasteConfig prompt_padding: PromptPaddingConfig extractor_id: str # 提取器 ID(已有) extractor_verified: bool # 提取器验证状态(已有) # ================= 选择器验证结果 ================= class SelectorValidationResult(TypedDict): """选择器验证结果""" key: str selector: Optional[str] valid: bool reason: Optional[str] repaired: Optional[str] # ================= AI 分析结果 ================= class AIAnalysisResult(TypedDict, total=False): """AI 分析返回的选择器结构""" input_box: Optional[str] send_btn: Optional[str] result_container: Optional[str] new_chat_btn: Optional[str] retry_send_btn: Optional[str] message_wrapper: Optional[str] generating_indicator: Optional[str] upload_btn: Optional[str] file_input: Optional[str] drop_zone: Optional[str] composer_root: Optional[str] # ================= 健康检查结果 ================= class HealthCheckResult(TypedDict): """健康检查结果""" status: Literal["healthy", "unhealthy"] connected: bool port: int tab_url: Optional[str] tab_title: Optional[str] error: Optional[str] # ================= 页面状态检查结果 ================= class PageStatusResult(TypedDict): """页面状态检查结果""" ready: bool reason: Optional[str] # ================= API 请求模型 ================= class ChatMessage(TypedDict, total=False): """聊天消息""" role: Literal["user", "assistant", "system"] content: str images: List[Dict] # 🆕 新增:图片列表(可选字段) class ChatCompletionRequest(TypedDict): """聊天补全请求""" model: str messages: List[ChatMessage] stream: Optional[bool] temperature: Optional[float] max_tokens: Optional[int] # ================= SSE 响应模型 ================= class DeltaContent(TypedDict, total=False): """流式响应的增量内容""" content: str images: List[Dict] # 🆕 新增:最后一个 chunk 携带 class StreamChoice(TypedDict): """流式响应的选项""" index: int delta: DeltaContent finish_reason: Optional[Literal["stop", None]] class StreamResponse(TypedDict): """流式响应结构""" id: str object: str created: int model: str choices: List[StreamChoice] # ================= 非流式响应模型 ================= class MessageContent(TypedDict, total=False): """消息内容""" role: Literal["assistant"] content: str images: List[Dict] # 🆕 新增 class NonStreamChoice(TypedDict): """非流式响应的选项""" index: int message: MessageContent finish_reason: Literal["stop"] class UsageInfo(TypedDict): """Token 使用信息(占位)""" prompt_tokens: int completion_tokens: int total_tokens: int class NonStreamResponse(TypedDict): """非流式响应结构""" id: str object: str created: int model: str choices: List[NonStreamChoice] usage: UsageInfo # ================= 错误响应模型 ================= class ErrorDetail(TypedDict): """错误详情""" message: str type: str code: str class ErrorResponse(TypedDict): """错误响应结构""" error: ErrorDetail # ================= 模型信息 ================= class ModelInfo(TypedDict): """模型信息""" id: str object: str created: int owned_by: str class ModelsResponse(TypedDict): """模型列表响应""" object: str data: List[ModelInfo] # ================= 提取器相关模型 ================= class ExtractorConfigDict(TypedDict, total=False): """提取器配置参数(内部类型定义)""" enable_latex: bool enable_shadow_dom: bool chunk_threshold: int class ExtractorDefinition(TypedDict): """提取器定义(extractors.json 中的结构)""" id: str name: str description: str class_: str # Python 类名(注意:JSON 中是 "class") module: str enabled: bool config: ExtractorConfigDict # ================= Pydantic API 模型(FastAPI 用)================= class ExtractorListResponse(BaseModel): """API 响应:提取器列表""" extractors: List[Dict[str, Any]] default: str model_config = ConfigDict( json_schema_extra={ "example": { "extractors": [ { "id": "deep_mode_v1", "name": "深度模式", "description": "JS 注入提取", "enabled": True } ], "default": "deep_mode_v1" } } ) class ExtractorTestRequest(BaseModel): """API 请求:测试提取器""" site_id: str extractor_id: str test_prompt: str = "Hello, test." model_config = ConfigDict( json_schema_extra={ "example": { "site_id": "chatgpt.com", "extractor_id": "deep_mode_v1", "test_prompt": "Write a short poem." } } ) class ExtractorVerifyRequest(BaseModel): """API 请求:验证提取结果""" extracted_text: str expected_text: str model_config = ConfigDict( json_schema_extra={ "example": { "extracted_text": "Roses are red...", "expected_text": "Roses are red..." } } ) class ExtractorVerifyResponse(BaseModel): """API 响应:验证结果""" similarity: float # 0.0 - 1.0 passed: bool # >= 0.95 视为通过 message: str model_config = ConfigDict( json_schema_extra={ "example": { "similarity": 0.973, "passed": True, "message": "验证通过" } } ) class ExtractorAssignRequest(BaseModel): """API 请求:为站点分配提取器""" extractor_id: str model_config = ConfigDict( json_schema_extra={ "example": { "extractor_id": "deep_mode_v1" } } ) # ================= 工具函数 ================= def validate_workflow_step(step: Dict[str, Any]) -> bool: """验证工作流步骤是否有效""" required_keys = {"action", "target", "optional"} return all(key in step for key in required_keys) def get_default_image_extraction_config() -> ImageExtractionConfig: """获取默认的多模态提取配置""" return { "enabled": False, "modalities": normalize_modalities_config({}), "selector": "img", "audio_selector": "audio, audio source", "video_selector": "video, video source", "container_selector": None, "force_postprocess": False, "debounce_seconds": 2.0, "wait_for_load": True, "load_timeout_seconds": 5.0, "download_blobs": True, "max_size_mb": 10, "canvas_export_mime": "image/jpeg", "canvas_export_quality": 0.88, "src_allow_patterns": [], "mode": "all", "audio_capture_enabled": True, "audio_capture_mute_playback": True, "audio_capture_preload_enabled": True, "audio_capture_reload_before_workflow": False, "audio_capture_preserve_graph": True, "audio_capture_terminal_settle_seconds": 0.35, "audio_trigger_selector": "", "audio_trigger_labels": ["朗读", "语音朗读", "收听", "read aloud", "listen", "tts", "voice"], "audio_capture_max_wait_seconds": 12.0, "audio_capture_min_wait_seconds": 2.0, "audio_capture_hard_max_wait_seconds": 45.0, "audio_capture_estimated_chars_per_second": 4.8, "audio_capture_wait_padding_seconds": 1.2, "audio_network_capture": { "enabled": False, "timeout_seconds": 2.5, "transport": "page_websocket_probe", "url_patterns": ["voicegenie", "speech", "audio", "tts"], "extractor": "voicegenie_binary_stream", "settle_seconds": 0.35, "max_payload_bytes": 10 * 1024 * 1024, }, "audio_browser_tts_fallback": { "enabled": False, "provider": "doubao_samantha", "speaker": "2", "speech_rate": 0, "pitch": 0, "format": "aac", "timeout_seconds": 30.0, "pc_version": "3.20.2", "aid": "497858", "real_aid": "497858", "language": "zh", "device_platform": "web", "pkg_type": "release_version", "region": "CN", "sys_region": "CN", "use_olympus_account": "1", "samantha_web": "1", }, "audio_capture_poll_seconds": 0.25, "audio_capture_silence_seconds": 1.2, "audio_capture_activity_threshold": 0.006, "audio_capture_activity_silence_seconds": 0.65, } def validate_site_config(config: Dict[str, Any]) -> bool: """验证站点配置是否有效""" if "selectors" not in config or "workflow" not in config: return False if not isinstance(config["selectors"], dict): return False if not isinstance(config["workflow"], list): return False for step in config["workflow"]: if not validate_workflow_step(step): return False if "stream_config" in config: if not isinstance(config["stream_config"], dict): return False stream_config = config["stream_config"] if "mode" in stream_config: mode = str(stream_config["mode"]).strip().lower() if mode not in {"dom", "network"}: return False if "hard_timeout" in stream_config and not isinstance(stream_config["hard_timeout"], (int, float)): return False if "request_transport" in stream_config: request_transport = stream_config["request_transport"] if not isinstance(request_transport, dict): return False if "mode" in request_transport and not isinstance(request_transport["mode"], str): return False if "profile" in request_transport and not isinstance(request_transport["profile"], str): return False if "options" in request_transport and not isinstance(request_transport["options"], dict): return False if "network" in stream_config: if not isinstance(stream_config["network"], dict): return False network_config = stream_config["network"] string_fields = [ "listen_pattern", "stream_match_pattern", "parser", ] numeric_fields = [ "first_response_timeout", "silence_threshold", "response_interval", ] for key in string_fields: if key in network_config and not isinstance(network_config[key], str): return False for key in numeric_fields: if key in network_config and not isinstance(network_config[key], (int, float)): return False if "stream_match_mode" in network_config: mode = str(network_config["stream_match_mode"]).strip().lower() if mode not in {"keyword", "regex"}: return False if "send_confirmation" in stream_config: if not isinstance(stream_config["send_confirmation"], dict): return False send_confirmation = stream_config["send_confirmation"] numeric_fields = [ "post_click_observe_window", "pre_retry_probe_window", "retry_observe_window", "attachment_observe_window", "retry_interval", "retry_cooldown_window", ] int_fields = [ "max_retry_count", ] string_fields = [ "retry_key_combo", ] bool_fields = [ "retry_on_unconfirmed_send", "accept_attachment_change", "accept_attachment_disappear", "accept_probe_confirmation", "retry_block_on_stop_button", "retry_block_if_generating", "trust_network_activity", "trust_generating_indicator", "trust_send_disabled_with_input_shrink", ] enum_fields = { "attachment_sensitivity": {"low", "medium", "high"}, "retry_action": {"click_send_btn", "key_press"}, } for key in numeric_fields: if key in send_confirmation and not isinstance(send_confirmation[key], (int, float)): return False for key in int_fields: if key in send_confirmation and not isinstance(send_confirmation[key], int): return False for key in string_fields: if key in send_confirmation and not isinstance(send_confirmation[key], str): return False for key in bool_fields: if key in send_confirmation and not isinstance(send_confirmation[key], bool): return False for key, allowed_values in enum_fields.items(): if key not in send_confirmation: continue value = str(send_confirmation[key]).strip().lower() if value not in allowed_values: return False if "attachment_monitor" in stream_config: if not isinstance(stream_config["attachment_monitor"], dict): return False attachment_monitor = stream_config["attachment_monitor"] numeric_fields = [ "idle_timeout", "hard_max_wait", ] bool_fields = [ "require_attachment_present", "require_upload_signal_before_ready", "continue_once_on_unconfirmed_send", ] list_fields = [ "root_selectors", "attachment_selectors", "pending_selectors", "status_selectors", "content_exclusion_selectors", "busy_text_markers", "ignored_busy_text_markers", "send_button_disabled_markers", ] for key in numeric_fields: if key in attachment_monitor and not isinstance(attachment_monitor[key], (int, float)): return False for key in bool_fields: if key in attachment_monitor and not isinstance(attachment_monitor[key], bool): return False for key in list_fields: if key not in attachment_monitor: continue value = attachment_monitor[key] if not isinstance(value, list): return False if any(not isinstance(item, str) for item in value): return False if "file_paste" in config: if not isinstance(config["file_paste"], dict): return False file_paste = config["file_paste"] if "enabled" in file_paste and not isinstance(file_paste["enabled"], bool): return False if "threshold" in file_paste and not isinstance(file_paste["threshold"], int): return False if "temp_file_type" in file_paste: temp_file_type = str(file_paste["temp_file_type"]).strip().lower() if temp_file_type not in {"txt", "pdf", "chunk", "error"}: return False if "hint_text" in file_paste and not isinstance(file_paste["hint_text"], str): return False if "txt_hint_text" in file_paste and not isinstance(file_paste["txt_hint_text"], str): return False if "pdf_hint_text" in file_paste and not isinstance(file_paste["pdf_hint_text"], str): return False if "error_hint_text" in file_paste and not isinstance(file_paste["error_hint_text"], str): return False if "reacquire_input_after_upload" in file_paste and not isinstance(file_paste["reacquire_input_after_upload"], bool): return False if "post_upload_input_selector" in file_paste and not isinstance(file_paste["post_upload_input_selector"], str): return False if "post_upload_settle" in file_paste and not isinstance(file_paste["post_upload_settle"], (int, float)): return False if "upload_signal_timeout" in file_paste and not isinstance(file_paste["upload_signal_timeout"], (int, float)): return False if "upload_signal_grace" in file_paste and not isinstance(file_paste["upload_signal_grace"], (int, float)): return False if "state_probe" in file_paste: state_probe = file_paste["state_probe"] if not isinstance(state_probe, dict): return False if "enabled" in state_probe and not isinstance(state_probe["enabled"], bool): return False if "code" in state_probe and not isinstance(state_probe["code"], str): return False if "send_confirmation" in file_paste: send_confirmation = file_paste["send_confirmation"] if not isinstance(send_confirmation, dict): return False numeric_fields = [ "post_click_observe_window", "pre_retry_probe_window", "retry_observe_window", "attachment_observe_window", "retry_interval", "retry_cooldown_window", ] int_fields = [ "max_retry_count", ] string_fields = [ "retry_key_combo", ] bool_fields = [ "retry_on_unconfirmed_send", "accept_attachment_change", "accept_attachment_disappear", "accept_probe_confirmation", "retry_block_on_stop_button", "retry_block_if_generating", "trust_network_activity", "trust_generating_indicator", "trust_send_disabled_with_input_shrink", ] enum_fields = { "attachment_sensitivity": {"low", "medium", "high"}, "retry_action": {"click_send_btn", "key_press"}, } for key in numeric_fields: if key in send_confirmation and not isinstance(send_confirmation[key], (int, float)): return False for key in int_fields: if key in send_confirmation and not isinstance(send_confirmation[key], int): return False for key in string_fields: if key in send_confirmation and not isinstance(send_confirmation[key], str): return False for key in bool_fields: if key in send_confirmation and not isinstance(send_confirmation[key], bool): return False for key, allowed_values in enum_fields.items(): if key not in send_confirmation: continue value = str(send_confirmation[key]).strip().lower() if value not in allowed_values: return False if "attachment_monitor" in file_paste: attachment_monitor = file_paste["attachment_monitor"] if not isinstance(attachment_monitor, dict): return False numeric_fields = [ "idle_timeout", "hard_max_wait", ] bool_fields = [ "require_attachment_present", "require_upload_signal_before_ready", "continue_once_on_unconfirmed_send", ] list_fields = [ "root_selectors", "attachment_selectors", "pending_selectors", "status_selectors", "content_exclusion_selectors", "busy_text_markers", "ignored_busy_text_markers", "send_button_disabled_markers", ] for key in numeric_fields: if key in attachment_monitor and not isinstance(attachment_monitor[key], (int, float)): return False for key in bool_fields: if key in attachment_monitor and not isinstance(attachment_monitor[key], bool): return False for key in list_fields: if key not in attachment_monitor: continue value = attachment_monitor[key] if not isinstance(value, list): return False if any(not isinstance(item, str) for item in value): return False if "prompt_padding" in config: if not isinstance(config["prompt_padding"], dict): return False prompt_padding = config["prompt_padding"] if "enabled" in prompt_padding and not isinstance(prompt_padding["enabled"], bool): return False if "marker_text" in prompt_padding and not isinstance(prompt_padding["marker_text"], str): return False if "segments_per_side" in prompt_padding and not isinstance(prompt_padding["segments_per_side"], int): return False if "random_insert_enabled" in prompt_padding and not isinstance(prompt_padding["random_insert_enabled"], bool): return False if "random_insert_chars" in prompt_padding and not isinstance(prompt_padding["random_insert_chars"], str): return False return True def get_default_send_confirmation_config() -> SendConfirmationConfig: """Get the default send confirmation strategy.""" return { "attachment_sensitivity": "medium", "post_click_observe_window": 1.8, "pre_retry_probe_window": 0.12, "retry_observe_window": 0.9, "attachment_observe_window": 6.0, "max_retry_count": 2, "retry_interval": 0.6, "retry_cooldown_window": 1.5, "retry_action": "click_send_btn", "retry_key_combo": "Enter", "retry_on_unconfirmed_send": True, "accept_attachment_change": False, "accept_attachment_disappear": False, "accept_probe_confirmation": True, "retry_block_on_stop_button": True, "retry_block_if_generating": True, "trust_network_activity": True, "trust_generating_indicator": True, "trust_send_disabled_with_input_shrink": True, } def get_default_attachment_monitor_config() -> AttachmentMonitorConfig: """Get default per-site attachment monitor rules.""" return { "root_selectors": [], "attachment_selectors": [], "pending_selectors": [], "busy_text_markers": [], "send_button_disabled_markers": [], "require_attachment_present": False, "require_upload_signal_before_ready": False, "continue_once_on_unconfirmed_send": True, "idle_timeout": 8.0, "hard_max_wait": 90.0, } def get_default_stream_config() -> StreamConfig: """获取默认的流式监控配置""" return { "request_transport": { "mode": "workflow", "profile": "", "options": {}, }, "send_confirmation": get_default_send_confirmation_config(), "attachment_monitor": get_default_attachment_monitor_config(), } def merge_stream_config( site_config: Optional[StreamConfig], defaults: Optional[StreamConfig] = None ) -> StreamConfig: """合并流式监控配置""" if defaults is None: defaults = get_default_stream_config() if site_config is None: return defaults.copy() result = defaults.copy() result.update(site_config) default_request_transport = defaults.get("request_transport") site_request_transport = site_config.get("request_transport") if isinstance(default_request_transport, dict): result["request_transport"] = copy.deepcopy(default_request_transport) if isinstance(site_request_transport, dict): result["request_transport"].update(site_request_transport) if isinstance(default_request_transport.get("options"), dict): result["request_transport"]["options"] = copy.deepcopy(default_request_transport.get("options") or {}) if isinstance(site_request_transport.get("options"), dict): result["request_transport"]["options"].update(site_request_transport["options"]) elif isinstance(site_request_transport, dict): result["request_transport"] = copy.deepcopy(site_request_transport) default_send_confirmation = defaults.get("send_confirmation") site_send_confirmation = site_config.get("send_confirmation") if isinstance(default_send_confirmation, dict): result["send_confirmation"] = default_send_confirmation.copy() if isinstance(site_send_confirmation, dict): result["send_confirmation"].update(site_send_confirmation) elif isinstance(site_send_confirmation, dict): result["send_confirmation"] = site_send_confirmation.copy() default_attachment_monitor = defaults.get("attachment_monitor") site_attachment_monitor = site_config.get("attachment_monitor") if isinstance(default_attachment_monitor, dict): result["attachment_monitor"] = default_attachment_monitor.copy() if isinstance(site_attachment_monitor, dict): result["attachment_monitor"].update(site_attachment_monitor) elif isinstance(site_attachment_monitor, dict): result["attachment_monitor"] = site_attachment_monitor.copy() return result # ================= 导出列表 ================= __all__ = [ # 类型定义 'ActionType', 'WorkflowStep', 'SelectorDefinition', 'GlobalConfig', 'StreamConfig', 'NetworkConfig', 'SiteConfig', 'SelectorValidationResult', 'AIAnalysisResult', 'HealthCheckResult', 'PageStatusResult', 'ChatMessage', 'ChatCompletionRequest', 'DeltaContent', 'StreamChoice', 'StreamResponse', 'MessageContent', 'NonStreamChoice', 'UsageInfo', 'NonStreamResponse', 'ErrorDetail', 'ErrorResponse', 'ModelInfo', 'ModelsResponse', # 提取器相关 'ExtractorConfigDict', 'ExtractorDefinition', 'ExtractorListResponse', 'ExtractorTestRequest', 'ExtractorVerifyRequest', 'ExtractorVerifyResponse', 'ExtractorAssignRequest', # 常量 'REQUIRED_SELECTOR_KEYS', 'OPTIONAL_SELECTOR_KEYS', 'ALL_SELECTOR_KEYS', 'DEFAULT_SELECTOR_DEFINITIONS', 'SITE_ADVANCED_FIELDS', 'PRESET_ADVANCED_FIELDS', 'ADVANCED_FIELDS', # 工具函数 'get_default_selector_definitions', 'validate_workflow_step', 'validate_site_config', 'SendConfirmationConfig', 'AttachmentMonitorConfig', 'get_default_send_confirmation_config', 'get_default_attachment_monitor_config', 'get_default_stream_config', 'merge_stream_config', 'ImageData', 'ImageExtractionConfig', 'ModalityPolicyConfig', 'ModalityRunPolicy', 'normalize_modalities_config', 'normalize_modality_policy', 'get_modality_policy', 'is_modality_enabled', 'get_modality_run_policy', 'get_enabled_modalities', 'get_default_image_extraction_config', 'FilePasteConfig', 'get_default_file_paste_config', 'PromptPaddingConfig', 'get_default_prompt_padding_config', ] # ================= 测试 ================= if __name__ == "__main__": print("=" * 50) print("Schema 模型测试") print("=" * 50) # 测试 Pydantic 模型 test_request = ExtractorTestRequest( site_id="chatgpt.com", extractor_id="deep_mode_v1", test_prompt="Hello!" ) print(f"\n✅ ExtractorTestRequest: {test_request.model_dump()}") test_response = ExtractorVerifyResponse( similarity=0.98, passed=True, message="验证通过" ) print(f"✅ ExtractorVerifyResponse: {test_response.model_dump()}") print("\n" + "=" * 50) print("所有测试通过!") print("=" * 50)