entari-plugin-hyw 3.3.0__tar.gz → 3.3.1__tar.gz

{entari_plugin_hyw-3.3.0/src/entari_plugin_hyw.egg-info → entari_plugin_hyw-3.3.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: entari_plugin_hyw
-Version: 3.3.0
+Version: 3.3.1
 Summary: Use large language models to interpret chat messages
 Author-email: kumoSleeping <zjr2992@outlook.com>
 License: MIT
@@ -19,9 +19,9 @@ Description-Content-Type: text/markdown
 Requires-Dist: arclet-entari[full]>=0.16.5
 Requires-Dist: openai
 Requires-Dist: httpx
-Requires-Dist: markdown>=3.10
-Requires-Dist: crawl4ai>=0.7.8
-Requires-Dist: jinja2>=3.0
+Provides-Extra: playwright
+Requires-Dist: playwright>=1.56.0; extra == "playwright"
+Requires-Dist: trafilatura>=2.0.0; extra == "playwright"
 Provides-Extra: dev
 Requires-Dist: entari-plugin-server>=0.5.0; extra == "dev"
 Requires-Dist: satori-python-adapter-onebot11>=0.2.5; extra == "dev"
@@ -38,10 +38,6 @@ Requires-Dist: satori-python-adapter-onebot11>=0.2.5; extra == "dev"
 
 </div>
 
-# v3.3 迎来大幅度改动、现在图文不符
-
-
-
 ## 🎑 效果展示
 
 
@@ -51,9 +47,12 @@ Requires-Dist: satori-python-adapter-onebot11>=0.2.5; extra == "dev"
 </div>
 
 ## ✨ 功能特性
-- **关于搜索**：一次性触发 Bing 网页与图片搜索，组合结果后再回应。
-- 给予 `Alconna` 与 `MessageChain` 混合处理, 深度优化触发体验。
-- **网页获取**：使用 Playwright 进行实时页面获取。
+- **关于搜索**：
+  - 如果不设置 jina token, 模型会根据提示词优先使用 jina / playwright(成功率较低) 获取渲染 bing / google 混合搜索结果。
+  - 存在 jina token 时，模型会获得一个 web search 工具，~~但我没试过我喜欢白嫖~~。
+  - 也可以 OpenRouter 的 `:online` 参数，该参数会优先使用模型提供商的搜索、其次 `exa`(较贵) 进行网页搜索。
+- 给予 `Alconna` 与 `MessageChain` 混合处理, 深度优化触发体验`。
+- **网页获取**：支持通过 **Jina AI** 或 **Playwright** 进行实时页面获取。
 - **多模态理解**：支持图片视觉分析。
 - **上下文感知**：维护对话历史记录，支持连续的多轮对话。
 - `reaction` 表情, 表示任务开始。
@@ -68,8 +67,12 @@ Requires-Dist: satori-python-adapter-onebot11>=0.2.5; extra == "dev"
 pip install entari-plugin-hyw
 ```
 
-### 搜索
-默认通过 HTTP 请求搜索引擎（DuckDuckGo，可在配置中自定义完整搜索链接，如 `https://duckduckgo.com/?q={query}`）。
+### 启用 Playwright 支持
+如果你希望使用 Playwright 进行本地网页渲染（而非仅使用 Jina AI）：
+```bash
+pip install entari-plugin-hyw[playwright]
+playwright install chromium
+```
 
 ## ⚙️ 配置
 
@@ -83,6 +86,7 @@ plugins:
     command_name_list: ["zssm", "hyw"]
     
     # 主 LLM 模型配置（必需）, 如 x-ai/grok-4.1-fast:online、perplexity/sonar
+    # 如果模型不自带搜索 模型会根据提示词优先使用 jina / playwright(成功率较低) 获取渲染 bing / google 混合搜索结果
     model_name: "gx-ai/grok-4.1-fast:free"
     api_key: "your-api-key"
 
@@ -90,8 +94,19 @@ plugins:
     base_url: "openai-compatible-url"  
     
     # --- 浏览器与搜索 ---
+    # 网页浏览工具: "jina" (默认) 或 "playwright"
+    browser_tool: "jina"
+    
+    # 可选: Jina AI API Key (配置以获得更高限额)(免费方案20/min)
+    # 配置此项同时会启用 web search 工具
+    jina_api_key: "jina_..."
+    
+    # Playwright 设置
     headless: true
     
+    # 浏览器回退: 当首选 browser_tool 失败时，尝试使用备用 browser_tool (默认: false)
+    enable_browser_fallback: false
+    
     # --- 视觉配置 (可选) ---
     # 如果未设置，将回退使用主模型
     vision_model_name: "qwen-vl-plus"
@@ -103,10 +118,6 @@ plugins:
       reasoning:
         effort: low
     
-    # --- 交互体验 ---
-    # 是否开启表情反应 (默认: true)
-    reaction: true
-
     # --- 调试 ---
     save_conversation: false
 ```
@@ -137,6 +148,3 @@ hyw -t 一大段话。
 ### 引用回复
 支持引用消息进行追问，机器人会自动读取被引用的消息作为上下文：
 - **引用 + 命令**：机器人将理解被引用消息的内容（包括图片）通过 `MessageChain` 操作拼接 `Text`、`Image` 与部分 `Custom`。
-
-UncleCode. (2024). Crawl4AI: Open-source LLM Friendly Web Crawler & Scraper [Computer software]. 
-GitHub. https://github.com/unclecode/crawl4ai

{entari_plugin_hyw-3.3.0 → entari_plugin_hyw-3.3.1}/README.md

@@ -10,10 +10,6 @@
 
 </div>
 
-# v3.3 迎来大幅度改动、现在图文不符
-
-
-
 ## 🎑 效果展示
 
 
@@ -23,9 +19,12 @@
 </div>
 
 ## ✨ 功能特性
-- **关于搜索**：一次性触发 Bing 网页与图片搜索，组合结果后再回应。
-- 给予 `Alconna` 与 `MessageChain` 混合处理, 深度优化触发体验。
-- **网页获取**：使用 Playwright 进行实时页面获取。
+- **关于搜索**：
+  - 如果不设置 jina token, 模型会根据提示词优先使用 jina / playwright(成功率较低) 获取渲染 bing / google 混合搜索结果。
+  - 存在 jina token 时，模型会获得一个 web search 工具，~~但我没试过我喜欢白嫖~~。
+  - 也可以 OpenRouter 的 `:online` 参数，该参数会优先使用模型提供商的搜索、其次 `exa`(较贵) 进行网页搜索。
+- 给予 `Alconna` 与 `MessageChain` 混合处理, 深度优化触发体验`。
+- **网页获取**：支持通过 **Jina AI** 或 **Playwright** 进行实时页面获取。
 - **多模态理解**：支持图片视觉分析。
 - **上下文感知**：维护对话历史记录，支持连续的多轮对话。
 - `reaction` 表情, 表示任务开始。
@@ -40,8 +39,12 @@
 pip install entari-plugin-hyw
 ```
 
-### 搜索
-默认通过 HTTP 请求搜索引擎（DuckDuckGo，可在配置中自定义完整搜索链接，如 `https://duckduckgo.com/?q={query}`）。
+### 启用 Playwright 支持
+如果你希望使用 Playwright 进行本地网页渲染（而非仅使用 Jina AI）：
+```bash
+pip install entari-plugin-hyw[playwright]
+playwright install chromium
+```
 
 ## ⚙️ 配置
 
@@ -55,6 +58,7 @@ plugins:
     command_name_list: ["zssm", "hyw"]
     
     # 主 LLM 模型配置（必需）, 如 x-ai/grok-4.1-fast:online、perplexity/sonar
+    # 如果模型不自带搜索 模型会根据提示词优先使用 jina / playwright(成功率较低) 获取渲染 bing / google 混合搜索结果
     model_name: "gx-ai/grok-4.1-fast:free"
     api_key: "your-api-key"
 
@@ -62,8 +66,19 @@ plugins:
     base_url: "openai-compatible-url"  
     
     # --- 浏览器与搜索 ---
+    # 网页浏览工具: "jina" (默认) 或 "playwright"
+    browser_tool: "jina"
+    
+    # 可选: Jina AI API Key (配置以获得更高限额)(免费方案20/min)
+    # 配置此项同时会启用 web search 工具
+    jina_api_key: "jina_..."
+    
+    # Playwright 设置
     headless: true
     
+    # 浏览器回退: 当首选 browser_tool 失败时，尝试使用备用 browser_tool (默认: false)
+    enable_browser_fallback: false
+    
     # --- 视觉配置 (可选) ---
     # 如果未设置，将回退使用主模型
     vision_model_name: "qwen-vl-plus"
@@ -75,10 +90,6 @@ plugins:
       reasoning:
         effort: low
     
-    # --- 交互体验 ---
-    # 是否开启表情反应 (默认: true)
-    reaction: true
-
     # --- 调试 ---
     save_conversation: false
 ```
@@ -109,6 +120,3 @@ hyw -t 一大段话。
 ### 引用回复
 支持引用消息进行追问，机器人会自动读取被引用的消息作为上下文：
 - **引用 + 命令**：机器人将理解被引用消息的内容（包括图片）通过 `MessageChain` 操作拼接 `Text`、`Image` 与部分 `Custom`。
-
-UncleCode. (2024). Crawl4AI: Open-source LLM Friendly Web Crawler & Scraper [Computer software]. 
-GitHub. https://github.com/unclecode/crawl4ai

{entari_plugin_hyw-3.3.0 → entari_plugin_hyw-3.3.1}/pyproject.toml

@@ -4,16 +4,13 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "entari_plugin_hyw"
-version = "3.3.0"
+version = "3.3.1"
 description = "Use large language models to interpret chat messages"
 authors = [{name = "kumoSleeping", email = "zjr2992@outlook.com"}]
 dependencies = [
     "arclet-entari[full]>=0.16.5",
     "openai",
     "httpx",
-    "markdown>=3.10",
-    "crawl4ai>=0.7.8",
-    "jinja2>=3.0",
 ]
 requires-python = ">=3.10"
 readme = "README.md"
@@ -29,6 +26,7 @@ classifiers = [
 ]
 
 [project.optional-dependencies]
+playwright = ["playwright>=1.56.0", "trafilatura>=2.0.0"]
 dev = [
     "entari-plugin-server>=0.5.0",
     "satori-python-adapter-onebot11>=0.2.5",
@@ -40,8 +38,8 @@ Repository = "https://github.com/kumoSleeping/entari-plugin-hyw"
 "Issue Tracker" = "https://github.com/kumoSleeping/entari-plugin-hyw/issues"
 
 [tool.setuptools]
-include-package-data = true
+packages = ["entari_plugin_hyw"]
+package-dir = {"" = "src"}
 
-[tool.setuptools.packages.find]
-where = ["src"]
-include = ["entari_plugin_hyw*"]
+[tool.setuptools.package-data]
+entari_plugin_hyw = ["*.txt", "*.md"]

entari_plugin_hyw-3.3.1/src/entari_plugin_hyw/__init__.py

@@ -0,0 +1,364 @@
+from dataclasses import dataclass
+import html
+import time
+from collections import deque
+from typing import Any, Deque, Dict, List, Optional, Set, Text, Tuple, Union, TYPE_CHECKING, cast
+from typing_extensions import override
+from arclet.entari import metadata
+from arclet.entari import MessageChain, Session
+from arclet.entari.event.base import MessageEvent
+from loguru import logger
+from satori.exception import ActionFailed
+from arclet.entari import MessageChain, Image, Quote, Text
+import arclet.letoderea as leto
+from arclet.entari import MessageCreatedEvent, Session
+from arclet.entari import BasicConfModel, metadata, plugin_config
+import httpx
+import asyncio
+import json
+import re
+from arclet.alconna import (
+    Args,
+    Alconna,
+    AllParam,
+    MultiVar,
+    CommandMeta,
+    Option,
+)
+from arclet.entari import MessageChain, Session, command
+from arclet.entari import plugin, Ready, Cleanup, Startup
+from satori.element import Custom, E
+from .hyw_core import HYW, HYWConfig
+
+# 全局变量
+hyw_core = None
+
+class HistoryManager:
+    def __init__(self, max_records: int = 20):
+        self.max_records = max_records
+        self._order: Deque[str] = deque()
+        self._store: Dict[str, List[dict]] = {}
+        self._bindings: Dict[str, Set[str]] = {}
+        self._msg_map: Dict[str, str] = {}
+
+    def extract_message_id(self, message_like: Any) -> Optional[str]:
+        if message_like is None:
+            return None
+        if isinstance(message_like, (list, tuple)):
+            for item in message_like:
+                mid = self.extract_message_id(item)
+                if mid:
+                    return mid
+            return None
+        if isinstance(message_like, dict):
+            for key in ("message_id", "id"):
+                value = message_like.get(key)
+                if value:
+                    return str(value)
+        for attr in ("message_id", "id"):
+            value = getattr(message_like, attr, None)
+            if value:
+                return str(value)
+        nested = getattr(message_like, "message", None)
+        if nested is not None and nested is not message_like:
+            return self.extract_message_id(nested)
+        return None
+
+    def remove(self, conversation_id: Optional[str], *, remove_from_order: bool = True) -> None:
+        if not conversation_id:
+            return
+        cid = str(conversation_id)
+        if remove_from_order:
+            try:
+                self._order.remove(cid)
+            except ValueError:
+                pass
+        bindings = self._bindings.pop(cid, set())
+        for msg_id in bindings:
+            self._msg_map.pop(msg_id, None)
+        self._store.pop(cid, None)
+
+    def _enforce_limit(self) -> None:
+        while len(self._order) > self.max_records:
+            obsolete = self._order.popleft()
+            self.remove(obsolete, remove_from_order=False)
+
+    def remember(self, conversation_id: Optional[str], history: Optional[List[dict]], related_ids: List[Optional[str]]) -> None:
+        if not conversation_id or not history:
+            return
+        cid = str(conversation_id)
+        self._store[cid] = list(history)
+        binding_ids = {str(mid) for mid in related_ids if mid}
+        self._bindings[cid] = binding_ids
+        for mid in binding_ids:
+            self._msg_map[mid] = cid
+        self._order.append(cid)
+        self._enforce_limit()
+
+    def get_history(self, msg_id: str) -> Optional[List[dict]]:
+        cid = self._msg_map.get(msg_id)
+        if cid:
+            return list(self._store.get(cid, []))
+        return None
+    
+    def get_conversation_id(self, msg_id: str) -> Optional[str]:
+        return self._msg_map.get(msg_id)
+
+history_manager = HistoryManager()
+
+# Request lock for HYW agent
+_hyw_request_lock: Optional[asyncio.Lock] = None
+
+def _get_hyw_request_lock() -> asyncio.Lock:
+    global _hyw_request_lock
+    if _hyw_request_lock is None:
+        _hyw_request_lock = asyncio.Lock()
+    return _hyw_request_lock
+
+
+class HywConfig(BasicConfModel):
+    command_name_list: Union[str, List[str]] = "hyw"
+    model_name: str
+    api_key: str
+    base_url: str = "https://openrouter.ai/api/v1"
+    headless: bool = False
+    save_conversation: bool = False
+    
+    browser_tool: str = "jina"
+    jina_api_key: Optional[str] = None
+    
+    vision_model_name: Optional[str] = None
+    vision_base_url: Optional[str] = None
+    vision_api_key: Optional[str] = None
+    
+    extra_body: Optional[Dict[str, Any]] = None
+    
+    enable_browser_fallback: bool = False
+    # verbose: bool = False
+
+metadata(
+    "hyw",
+    author=[{"name": "kumoSleeping", "email": "zjr2992@outlook.com"}],
+    version="3.3.1",
+    description="",
+    config=HywConfig,
+)
+
+conf = plugin_config(HywConfig)
+alc = Alconna(
+    conf.command_name_list,
+    Option("-t|--text", dest="text_only", default=False, help_text="仅文本模式(禁用图片识别)"),
+    Args["all_param", AllParam],
+    # Option("-v|--verbose", dest="verbose", default=False, help_text="启用详细日志输出"),
+    meta=CommandMeta(compact=False)
+)
+
+# Create HYW configuration
+hyw_config = HYWConfig(
+    api_key=conf.api_key,
+    model_name=conf.model_name,
+    base_url=conf.base_url,
+    save_conversation=conf.save_conversation,
+    headless=conf.headless,
+    browser_tool=conf.browser_tool,
+    jina_api_key=conf.jina_api_key,
+    vision_model_name=conf.vision_model_name,
+    vision_base_url=conf.vision_base_url,
+    vision_api_key=conf.vision_api_key,
+    extra_body=conf.extra_body,
+    enable_browser_fallback=conf.enable_browser_fallback
+)
+
+hyw = HYW(config=hyw_config)
+
+
+
+# Emoji到代码的映射字典
+EMOJI_TO_CODE = {
+    "🐳": "128051",
+    "❌": "10060",
+    "🍧": "127847",
+    "✨": "10024",
+    "📫": "128235"
+}
+
+async def download_image(url: str) -> bytes:
+    """下载图片"""
+    try:
+        async with httpx.AsyncClient(timeout=30.0) as client:
+            resp = await client.get(url)
+            if resp.status_code == 200:
+                return resp.content
+            else:
+                raise ActionFailed(f"下载图片失败，状态码: {resp.status_code}")
+    except Exception as e:
+        raise ActionFailed(f"下载图片失败: {url}, 错误: {str(e)}")
+
+
+def process_onebot_json(json_data_str: str) -> str:
+    try:
+        # 解码HTML实体
+        json_str = html.unescape(json_data_str)
+        return json_str
+    except Exception as e:
+        return json_data_str
+
+
+async def react(session: Session, emoji: str):
+    try:
+        if session.event.login.platform == "onebot":
+            code = EMOJI_TO_CODE.get(emoji, "10024")
+            await session.account.protocol.call_api("internal/set_group_reaction", {"group_id": int(session.guild.id), "message_id": int(session.event.message.id), "code": code, "is_add": True})
+        else:
+            await session.reaction_create(emoji=emoji)
+    except ActionFailed:
+        pass
+
+def handle_shortcut(message_chain: MessageChain) -> Tuple[bool, str]:
+    current_msg_text = str(message_chain.get(Text)) if message_chain.get(Text) else ""
+    is_shortcut = False
+    shortcut_replacement = ""
+    if current_msg_text.strip().startswith("/"):
+        is_shortcut = True
+        shortcut_replacement = current_msg_text.strip()[1:]
+    return is_shortcut, shortcut_replacement
+
+async def process_images(mc: MessageChain, parse_result: Any) -> Tuple[List[str], Optional[str]]:
+    is_text_only = False
+    if parse_result.matched:
+        def get_bool_value(val):
+            if hasattr(val, 'value'):
+                return bool(val.value)
+            return bool(val)
+        is_text_only = get_bool_value(getattr(parse_result, 'text_only', False))
+    
+    text_str = str(mc.get(Text) or "")
+    if not is_text_only and re.search(r'(?:^|\s)(-t|--text)(?:$|\s)', text_str):
+        is_text_only = True
+    
+    if is_text_only:
+        logger.info("检测到仅文本模式参数，跳过图片分析")
+        return [], None
+
+    has_images = bool(mc.get(Image))
+    images = []
+    if has_images:
+        urls = mc[Image].map(lambda x: x.src)
+        tasks = [download_image(url) for url in urls]
+        raw_images = await asyncio.gather(*tasks)
+        import base64
+        images = [base64.b64encode(img).decode('utf-8') for img in raw_images]
+    
+    return images, None
+
+@leto.on(MessageCreatedEvent)
+async def on_message_created(message_chain: MessageChain, session: Session[MessageEvent]):
+    # Skip if no substantial content in original message
+    original_text = str(message_chain.get(Text)).strip()
+    has_images = bool(message_chain.get(Image))
+    has_custom = bool(message_chain.get(Custom))
+    if not original_text and not has_images and not has_custom:
+        return
+    
+    if session.reply:
+        try:
+            message_chain.extend(MessageChain(" ") + session.reply.origin.message)
+        except Exception:
+            pass
+            
+    message_chain = message_chain.get(Text) + message_chain.get(Image) + message_chain.get(Custom)
+    
+    quoted_message_id: Optional[str] = None
+    conversation_history_key: Optional[str] = None
+    conversation_history_payload: List[dict] = []
+    
+    if session.reply:
+        try:
+            quoted_message_id = str(session.reply.origin.id) if hasattr(session.reply.origin, 'id') else None
+        except Exception as e:
+            logger.warning(f"提取引用消息ID失败: {e}")
+            quoted_message_id = None
+            
+        if quoted_message_id:
+            conversation_history_key = history_manager.get_conversation_id(quoted_message_id)
+            if conversation_history_key:
+                conversation_history_payload = history_manager.get_history(quoted_message_id) or []
+                logger.info(f"继续对话模式触发, 引用消息ID: {quoted_message_id}, 历史长度: {len(conversation_history_payload)}")
+
+    parse_result = alc.parse(message_chain)
+    is_shortcut, shortcut_replacement = handle_shortcut(message_chain)
+    
+    should_process = parse_result.matched or (bool(conversation_history_key) and is_shortcut)
+    
+    if not should_process:
+        return
+
+    raw_param_chain: MessageChain = parse_result.all_param if parse_result.matched else message_chain  # type: ignore
+    if not parse_result.matched and is_shortcut:
+        logger.debug(f"触发快捷指令，替换内容: {shortcut_replacement}")
+        
+    mc = MessageChain(raw_param_chain)
+    
+    async def process_request() -> None:
+        await react(session, "✨")
+        try:
+            if is_shortcut and not parse_result.matched:
+                msg = shortcut_replacement
+            else:
+                msg = mc.get(Text).strip() if mc.get(Text) else ""
+            
+            if mc.get(Custom): # type: ignore
+                custom_elements = [e for e in mc if isinstance(e, Custom)]
+                for custom in custom_elements:
+                    if custom.tag == 'onebot:json':
+                        decoded_json = process_onebot_json(custom.attributes())
+                        msg += decoded_json
+                        break
+            
+            time_start = time.perf_counter()
+            images, error_msg = await process_images(mc, parse_result)
+            
+            if error_msg:
+                await session.send(error_msg)
+                return
+
+            lock = _get_hyw_request_lock()
+            async with lock:
+                response = await hyw.agent(str(msg), conversation_history=conversation_history_payload, images=images)
+            
+            response_content = response.get("llm_response", "") if isinstance(response, dict) else ""
+            new_history = response.get("conversation_history", []) if isinstance(response, dict) else []
+            
+            try:
+                send_result = await session.send([Quote(session.event.message.id), response_content])
+            except ActionFailed as e:
+                if "9057" in str(e):
+                    logger.warning(f"发送消息失败(9057)，尝试截断发送: {e}")
+                    truncated_content = response_content[:1000] + "\n\n[...内容过长，已大幅截断...]"
+                    send_result = await session.send([Quote(session.event.message.id), truncated_content])
+                else:
+                    raise e
+            
+            sent_message_id = history_manager.extract_message_id(send_result)
+            current_user_message_id = str(session.event.message.id)
+            related_ids: List[Optional[str]] = [current_user_message_id, sent_message_id]
+            
+            if conversation_history_key:
+                history_manager.remove(conversation_history_key)
+                related_ids.append(quoted_message_id)
+            
+            # Check turn limit
+            user_turns = len([m for m in new_history if m.get("role") == "user"])
+            if user_turns < 5:
+                history_manager.remember(sent_message_id, new_history, related_ids)
+            else:
+                logger.info(f"对话轮数达到上限 ({user_turns})，停止记录历史")
+            
+        except Exception as exc:
+            await react(session, "❌")
+            logger.exception("处理HYW消息失败: {}", exc)
+
+    asyncio.create_task(process_request())
+    return
+
+