import asyncio import json import urllib.request import urllib.error import base64 import traceback from typing import List, Tuple, Type, Optional # 导入新插件系统 from src.plugin_system.apis import chat_api from src.plugin_system.apis import send_api from src.plugin_system.apis import config_api from src.plugin_system.base.base_plugin import BasePlugin from src.plugin_system.base.base_plugin import register_plugin from src.plugin_system.base.base_action import BaseAction from src.plugin_system.base.component_types import ComponentInfo, ActionActivationType, ChatMode from src.plugin_system.base.config_types import ConfigField from src.common.logger import get_logger logger = get_logger("pic_action") # 当此模块被加载时,尝试生成配置文件(如果它不存在) # 注意:在某些插件加载机制下,这可能会在每次机器人启动或插件重载时执行 # 考虑是否需要更复杂的逻辑来决定何时运行 (例如,仅在首次安装时) # ===== Action组件 ===== class Custom_Pic_Action(BaseAction): """生成一张图片并发送""" # 激活设置 focus_activation_type = ActionActivationType.LLM_JUDGE # Focus模式使用LLM判定,精确理解需求 normal_activation_type = ActionActivationType.KEYWORD # Normal模式使用关键词激活,快速响应 mode_enable = ChatMode.ALL parallel_action = True # 动作基本信息 action_name = "draw_picture" action_description = ( "可以根据特定的描述,生成并发送一张图片,如果没提供描述,就根据聊天内容生成,你可以立刻画好,不用等待" ) # 关键词设置(用于Normal模式) activation_keywords = ["画", "绘制", "生成图片", "画图", "draw", "paint", "图片生成"] # LLM判定提示词(用于Focus模式) llm_judge_prompt = """ 判定是否需要使用图片生成动作的条件: 1. 用户明确要求画图、生成图片或创作图像 2. 用户描述了想要看到的画面或场景 3. 对话中提到需要视觉化展示某些概念 4. 用户想要创意图片或艺术作品 适合使用的情况: - "画一张..."、"画个..."、"生成图片" - "我想看看...的样子" - "能画出...吗" - "创作一幅..." 绝对不要使用的情况: 1. 纯文字聊天和问答 2. 只是提到"图片"、"画"等词但不是要求生成 3. 谈论已存在的图片或照片 4. 技术讨论中提到绘图概念但无生成需求 5. 用户明确表示不需要图片时 """ keyword_case_sensitive = False # 动作参数定义 action_parameters = { "description": "图片描述,输入你想要生成并发送的图片的描述,将描述翻译为英文单词组合,并用‘,‘分隔,描述中不要出现中文,必填", "size": "图片尺寸,例如 '512x512' (可选, 默认从配置或 '1024x1024')", } # 动作使用场景 action_require = [ #"当有人让你画东西时使用,你可以立刻画好,不用等待", "当有人要求你生成并发送一张图片时使用,不要频率太高", #"当有人让你画一张图时使用", "当你想要通过自画像表情包来表达自己情感时使用,不要频率太高", "重点:不要连续发,如果你在前10句内已经发送过[图片]或者[表情包]或记录出现过类似描述的[图片],就不要不选择此动作", ] associated_types = ["text", "image"] # 简单的请求缓存,避免短时间内重复请求 _request_cache = {} _cache_max_size = 10 async def execute(self) -> Tuple[bool, Optional[str]]: """执行图片生成动作""" logger.info(f"{self.log_prefix} 执行绘图模型图片生成动作") # 配置验证 http_base_url = self.get_config("api.base_url") http_api_key = self.get_config("api.api_key") if not (http_base_url and http_api_key): error_msg = "抱歉,图片生成功能所需的HTTP配置(如API地址或密钥)不完整,无法提供服务。" await self.send_text(error_msg) logger.error(f"{self.log_prefix} HTTP调用配置缺失: base_url 或 api_key.") return False, "HTTP配置不完整" # API密钥验证 if http_api_key == "YOUR_DOUBAO_API_KEY_HERE": error_msg = "图片生成功能尚未配置,请设置正确的API密钥。" await self.send_text(error_msg) logger.error(f"{self.log_prefix} API密钥未配置") return False, "API密钥未配置" # 参数验证 description = self.action_data.get("description") if not description or not description.strip(): logger.warning(f"{self.log_prefix} 图片描述为空,无法生成图片。") await self.send_text("你需要告诉我想要画什么样的图片哦~ 比如说'画一只可爱的小猫'") return False, "图片描述为空" # 清理和验证描述 description = description.strip() if len(description) > 1000: # 限制描述长度 description = description[:1000] logger.info(f"{self.log_prefix} 图片描述过长,已截断") # 获取配置 default_model = self.get_config("generation.default_model", "doubao-seedream-3-0-t2i-250415") image_size = self.action_data.get("size", self.get_config("generation.default_size", "1024x1024")) # 验证图片尺寸格式 if not self._validate_image_size(image_size): logger.warning(f"{self.log_prefix} 无效的图片尺寸: {image_size},使用默认值") image_size = "1024x1024" # 检查缓存 cache_key = self._get_cache_key(description, default_model, image_size) if cache_key in self._request_cache: cached_result = self._request_cache[cache_key] logger.info(f"{self.log_prefix} 使用缓存的图片结果") await self.send_text("我之前画过类似的图片,用之前的结果~") # 直接发送缓存的结果 send_success = await self.send_image(cached_result) if send_success: await self.send_text("图片已发送!") return True, "图片已发送(缓存)" else: # 缓存失败,清除这个缓存项并继续正常流程 del self._request_cache[cache_key] # 获取其他配置参数 seed_val = self.get_config("generation.default_seed",42)#种子 guidance_scale_val=self.get_config("default_guidance_scale",2.5)#强度 watermark_val = self.get_config("generation.default_watermark", True)#水印 await self.send_text( f"收到!正在为您生成关于 '{description}' 的图片,请稍候...(模型: {default_model}, 尺寸: {image_size})" ) try: success, result = await asyncio.to_thread( self._make_http_image_request, prompt=description, model=default_model, size=image_size, seed=seed_val, guidance_scale=guidance_scale_val, watermark=watermark_val, ) except Exception as e: logger.error(f"{self.log_prefix} (HTTP) 异步请求执行失败: {e!r}", exc_info=True) traceback.print_exc() success = False result = f"图片生成服务遇到意外问题: {str(e)[:100]}" if success: # 如果返回的是Base64数据(以"iVBORw"等开头),直接使用 if result.startswith(("iVBORw", "/9j/", "UklGR", "R0lGOD")): # 常见图片格式的Base64前缀 #logger.info(f"{self.log_prefix} 获取到Base64图片数据,直接发送") #logger.info(f"{self.log_prefix} (HTTP) 获取到Base64图片数据,长度: {len(self.log_prefix)}") send_success = await self.send_image(result) if send_success: await self.send_text("图片表情已发送!") return True, "图片表情已发送(Base64)" else: await self.send_text("图片已处理为Base64,但作为表情发送失败了")#("图片已处理为Base64,但作为表情发送失败了。") return False, "图片表情发送失败 (Base64)" else: # 否则认为是URL image_url = result # print(f"image_url: {image_url}") # print(f"result: {result}") logger.info(f"{self.log_prefix} 图片URL获取成功: {image_url[:70]}... 下载并编码.") try: encode_success, encode_result = await asyncio.to_thread(self._download_and_encode_base64, image_url) except Exception as e: logger.error(f"{self.log_prefix} (B64) 异步下载/编码失败: {e!r}", exc_info=True) traceback.print_exc() encode_success = False encode_result = f"图片下载或编码时发生内部错误: {str(e)[:100]}" if encode_success: base64_image_string = encode_result send_success = await self.send_image(base64_image_string) if send_success: # 缓存成功的结果 self._request_cache[cache_key] = base64_image_string self._cleanup_cache() await self.send_text("图片已发送!") return True, "图片已成功生成并发送" else: print(f"send_success: {send_success}") await self.send_text("图片已处理为Base64,但发送失败了。") return False, "图片发送失败 (Base64)" else: await self.send_text(f"获取到图片URL,但在处理图片时失败了:{encode_result}") return False, f"图片处理失败(Base64): {encode_result}" else: error_message = result await self.send_text(f"哎呀,生成图片时遇到问题:{error_message}") return False, f"图片生成失败: {error_message}" def _download_and_encode_base64(self, image_url: str) -> Tuple[bool, str]: """下载图片并将其编码为Base64字符串""" logger.info(f"{self.log_prefix} (B64) 下载并编码图片: {image_url[:70]}...") try: with urllib.request.urlopen(image_url, timeout=60) as response: if response.status == 200: image_bytes = response.read() base64_encoded_image = base64.b64encode(image_bytes).decode("utf-8") logger.info(f"{self.log_prefix} (B64) 图片下载编码完成. Base64长度: {len(base64_encoded_image)}") return True, base64_encoded_image else: error_msg = f"下载图片失败 (状态: {response.status})" logger.error(f"{self.log_prefix} (B64) {error_msg} URL: {image_url}") return False, error_msg except Exception as e: logger.error(f"{self.log_prefix} (B64) 下载或编码时错误: {e!r}", exc_info=True) traceback.print_exc() return False, f"下载或编码图片时发生错误: {str(e)[:100]}" @classmethod def _get_cache_key(cls, description: str, model: str, size: str) -> str: """生成缓存键""" return f"{description[:100]}|{model}|{size}" @classmethod def _cleanup_cache(cls): """清理缓存,保持大小在限制内""" if len(cls._request_cache) > cls._cache_max_size: keys_to_remove = list(cls._request_cache.keys())[: -cls._cache_max_size // 2] for key in keys_to_remove: del cls._request_cache[key] def _validate_image_size(self, image_size: str) -> bool: """验证图片尺寸格式""" try: width, height = map(int, image_size.split("x")) return 100 <= width <= 10000 and 100 <= height <= 10000 except (ValueError, TypeError): return False def _make_http_image_request( self, prompt: str, model: str, size: str, seed: int | None, guidance_scale: float, watermark: bool ) -> Tuple[bool, str]: """发送HTTP请求生成图片""" base_url = self.get_config("api.base_url") generate_api_key = self.get_config("api.api_key") endpoint = f"{base_url.rstrip('/')}/images/generations" # 获取配置参数 - 使用字符串键名 custom_prompt_add = self.get_config("generation.custom_prompt_add", "")#附加正面提示词参数 negative_prompt_add = self.get_config("generation.negative_prompt_add", "")#附加负面提示词参数 prompt_add= prompt + ", " + custom_prompt_add negative_prompt = negative_prompt_add payload_dict = { "model": model, "prompt": prompt_add, # 使用附加的正面提示词 "negative_prompt":negative_prompt, #"response_format": "b64_json",# gpt-image-1 无法使用 url 返回为 “b64_json",豆包默认返回为 "url" "size": size, "guidance_scale": guidance_scale, "watermark": watermark, "seed": seed, # seed is now always an int from process() "api-key": generate_api_key, } # if seed is not None: # No longer needed, seed is always an int # payload_dict["seed"] = seed data = json.dumps(payload_dict).encode("utf-8") headers = { "Content-Type": "application/json", "Accept": "application/json", "Authorization": f"{generate_api_key}", }# gpt 或其他模型密钥要删除‘Bearer ’前缀 logger.info(f"{self.log_prefix} (HTTP) 发起图片请求: {model}, custom_prompt_add:{custom_prompt_add[:30]}...,Prompt: {prompt_add[:30]}...,NegativePrompt: {negative_prompt[:30]}... To: {endpoint}") logger.debug( f"{self.log_prefix} (HTTP) Request Headers: {{...Authorization: Bearer {generate_api_key[:10]}...}}" ) logger.debug( f"{self.log_prefix} (HTTP) Request Body (api-key omitted): {json.dumps({k: v for k, v in payload_dict.items() if k != 'api-key'})}" ) req = urllib.request.Request(endpoint, data=data, headers=headers, method="POST") try: with urllib.request.urlopen(req, timeout=60) as response: response_status = response.status response_body_bytes = response.read() response_body_str = response_body_bytes.decode("utf-8") logger.info(f"{self.log_prefix} (HTTP) 响应: {response_status}. Preview: {response_body_str[:150]}...") if 200 <= response_status < 300: response_data = json.loads(response_body_str) b64_data = None image_url = None #清理缓存 # 优先检查Base64数据 if ( isinstance(response_data.get("data"), list) and response_data["data"] and isinstance(response_data["data"][0], dict) and "b64_json" in response_data["data"][0] ): b64_data = response_data["data"][0]["b64_json"] logger.info(f"{self.log_prefix} (HTTP) 获取到Base64图片数据,长度: {len(b64_data)}") return True, b64_data # 直接返回Base64字符串 elif ( isinstance(response_data.get("data"), list) and response_data["data"] and isinstance(response_data["data"][0], dict) ): image_url = response_data["data"][0].get("url") elif(#魔搭社区返回的 json isinstance(response_data.get("images"), list) and response_data["images"] and isinstance(response_data["images"][0], dict) ): image_url = response_data["images"][0].get("url") elif response_data.get("url"): image_url = response_data.get("url") if image_url: logger.info(f"{self.log_prefix} (HTTP) 图片生成成功,URL: {image_url[:70]}...") return True, image_url else: logger.error( f"{self.log_prefix} (HTTP) API成功但无图片URL. 响应预览: {response_body_str[:300]}..." ) return False, "图片生成API响应成功但未找到图片URL" else: logger.error( f"{self.log_prefix} (HTTP) API请求失败. 状态: {response.status}. 正文: {response_body_str[:300]}..." ) return False, f"图片API请求失败(状态码 {response.status})" except Exception as e: logger.error(f"{self.log_prefix} (HTTP) 图片生成时意外错误: {e!r}", exc_info=True) traceback.print_exc() return False, f"图片生成HTTP请求时发生意外错误: {str(e)[:100]}" # ===== 插件注册 ===== @register_plugin class CustomPicPlugin(BasePlugin): """根据描述使用不同的 绘图 API生成图片的动作处理类""" # 插件基本信息 plugin_name = "custom_pic_plugin"# 内部标识符 plugin_version = "1.0.2" plugin_author = "Ptrel" enable_plugin = True config_file_name = "config.toml" # 步骤1: 定义配置节的描述 config_section_descriptions = { "plugin": "插件启用配置", "api": "API的基础url", "generation": "图片生成参数配置,控制生成图片的各种参数", "cache": "结果缓存配置", "components": "组件启用配置", } # 步骤2: 使用ConfigField定义详细的配置Schema config_schema = { "plugin": { "name": ConfigField(type=str, default="custom_pic_plugin", description="自定义提示词绘图", required=True), "version": ConfigField(type=str, default="1.1.2", description="插件版本号"), "enabled": ConfigField(type=bool, default=False, description="是否启用插件") }, "api": { "base_url": ConfigField( type=str, default="https://api-inference.modelscope.cn/v1", description="API的基础url", example="https://api.example.com/v1", choices=[ "https://api-inference.modelscope.cn/v1\"#魔搭可自选 lora,对应模型网址:https://modelscope.cn/models?page=1&tabKey=task&tasks=hotTask:text-to-image-synthesis&type=tasks", "\n#https://ark.cn-beijing.volces.com/api/v3\"#豆包火山方舟 API,对应网址:https://console.volcengine.com/auth/login?redirectURI=%2Fmessage%2Finnermsg", "\n#https://api.chatanywhere.tech/v1\"#chatany API,对应文档网址:https://chatanywhere.apifox.cn/",] ), "api_key": ConfigField( type=str, default="Bearer xxxxxxxxxxxxxxxxxxxxxx", description="API 的 api 密钥,需要添加‘Bearer ’前缀,chatany 不需要前缀,根据不同 api 文档进行选择,如 chatanywhere 的 key 不需要 Berarer,即直接输入密钥‘xxxxxxxxxxxxxxxxxxxxxxxx’", required=True ), }, "generation": { "default_model": ConfigField( type=str, default="cancel13/liaocao\"#潦草 lora 图片生成模型(魔搭)", description="模型选择,可自定义,以下为示例", choices=[ "MusePublic/14_ckpt_SD_XL\"#万象熔炉 | Anything XL,社区模型(魔搭)","\n#cancel13/liaocao\"#潦草 lora 图片生成模型,社区模型(魔搭)","\n#doubao-seedream-3-0-t2i-250415\"#豆包图片生成模型约 0.3¥ 一张图(火山)", "\n#gpt-image-1\"#GPT 生图,约 1.3¥ 一张图(chatany)" ] ), "default_size": ConfigField( type=str, default="1024x1024", description="默认图片尺寸", example="1024x1024", choices=["1024x1024", "1024x1280", "1280x1024", "1024x1536", "1536x1024"], ), "default_watermark": ConfigField( type=bool, default=True, description="是否默认添加水印"), "default_guidance_scale": ConfigField( type=float, default=2.5, description="模型指导强度,影响图片与提示的关联性", example="2.0" ), "default_seed": ConfigField( type=int, default=42, description="随机种子,用于复现图片"), "custom_prompt_add": ConfigField( type=str, default="Nordic picture book art style, minimalist flat design, soft rounded lines, high saturation color blocks collision, dominant forest green and warm orange palette, low contrast lighting, hand-drawn pencil texture, healing fairy-tale atmosphere, geometric natural forms, ample white space composition, warm and clean aesthetic,liaocao\"#北欧绘本艺术风格,简约扁平设计,柔和圆润线条,高饱和度色块碰撞,森林绿与暖橙主色调,低对比度光影,手绘铅笔质感,治愈系童话氛围,几何化自然形态,留白构图,温暖干净画面", description="正面附加提示词(尽量使用英文,且用词语和逗号的形式,豆包可以使用中文句子提示词)" ), "negative_prompt_add": ConfigField( type=str, default="Pornography,nudity,lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry\"#色情,裸体,低分辨率,糟糕的解剖结构,糟糕的手,文字,错误,缺失的手指,多余的数字,更少的数字,裁剪,最差的质量,低质量,正常质量,jpeg文物,签名,水印,用户名,模糊色情,裸体,低分辨率,糟糕的解剖结构,糟糕的手,文字,错误,缺失的手指,多余的数字,更少的数字,裁剪,最差的质量,低质量,正常质量,jpeg文物,签名,水印,用户名,模糊", description="负面附加提示词,保持默认或使用豆包时可留空" ), }, "cache": { "enabled": ConfigField(type=bool, default=True, description="是否启用请求缓存"), "max_size": ConfigField(type=int, default=10, description="最大缓存数量"), }, "components": { "enable_image_generation": ConfigField(type=bool, default=True, description="是否启用图片生成Action") }, "logging": { "level": ConfigField( type=str, default="INFO", description="日志记录级别", choices=["DEBUG", "INFO", "WARNING", "ERROR"] ), "prefix": ConfigField(type=str, default="[custom_pic_Plugin]", description="日志记录前缀", example="[custom_pic_Plugin]") } } def get_plugin_components(self) -> List[Tuple[ComponentInfo, Type]]: """返回插件包含的组件列表""" # 从配置获取组件启用状态 enable_image_generation = self.get_config("components.enable_image_generation", True) components = [] if enable_image_generation: # 添加我们的Action components.append((Custom_Pic_Action.get_action_info(), Custom_Pic_Action)) return components