Merge branch 'main' into py-test

This commit is contained in:
hrz
2025-05-09 18:08:48 +08:00
committed by GitHub
22 changed files with 770 additions and 165 deletions
+12 -3
View File
@@ -104,12 +104,13 @@ plugins:
get_weather: { "api_key": "a861d0d5e7bf4ee1a83d9a9e4f96d4da", "default_location": "广州" }
# 获取新闻插件的配置,这里根据需要的新闻类型传入对应的url链接,默认支持社会、科技、财经新闻
# 更多类型的新闻列表查看 https://www.chinanews.com.cn/rss/
get_news:
get_news_from_chinanews:
default_rss_url: "https://www.chinanews.com.cn/rss/society.xml"
category_urls:
society: "https://www.chinanews.com.cn/rss/society.xml"
world: "https://www.chinanews.com.cn/rss/world.xml"
finance: "https://www.chinanews.com.cn/rss/finance.xml"
get_news_from_newsnow: {"url": "https://newsnow.busiyi.world/api/s?id="}
home_assistant:
devices:
- 客厅,玩具灯,switch.cuco_cn_460494544_cp1_on_p_2_1
@@ -156,7 +157,7 @@ selected_module:
Memory: nomem
# 意图识别模块开启后,可以播放音乐、控制音量、识别退出指令。
# 不想开通意图识别,就设置成:nointent
# 意图识别可使用intent_llm。优点:通用性强,缺点:增加串行前置意图识别模块,会增加处理时间,这个意图识别暂时不支持控制音量大小等iot操作
# 意图识别可使用intent_llm。优点:通用性强,缺点:增加串行前置意图识别模块,会增加处理时间,支持控制音量大小等iot操作
# 意图识别可使用function_call,缺点:需要所选择的LLM支持function_call,优点:按需调用工具、速度快,理论上能全部操作所有iot指令
# 默认免费的ChatGLMLLM就已经支持function_call,但是如果像追求稳定建议把LLM设置成:DoubaoLLM,使用的具体model_name是:doubao-1-5-pro-32k-250115
Intent: function_call
@@ -174,6 +175,13 @@ Intent:
# 如果这里不填,则会默认使用selected_module.LLM的模型作为意图识别的思考模型
# 如果你的不想使用selected_module.LLM意图识别,这里最好使用独立的LLM作为意图识别,例如使用免费的ChatGLMLLM
llm: ChatGLMLLM
# plugins_func/functions下的模块,可以通过配置,选择加载哪个模块,加载后对话支持相应的function调用
# 系统默认已经记载“handle_exit_intent(退出识别)”、“play_music(音乐播放)”插件,请勿重复加载
# 下面是加载查天气、角色切换、加载查新闻的插件示例
functions:
- get_weather
- get_news_from_newsnow
- play_music
function_call:
# 不需要动type
type: function_call
@@ -183,7 +191,8 @@ Intent:
functions:
- change_role
- get_weather
- get_news
# - get_news_from_chinanews
- get_news_from_newsnow
# play_music是服务器自带的音乐播放,hass_play_music是通过home assistant控制的独立外部程序音乐播放
# 如果用了hass_play_music,就不要开启play_music,两者只留一个
- play_music
+17 -9
View File
@@ -60,9 +60,9 @@ class ConnectionHandler:
self.config = copy.deepcopy(config)
self.session_id = str(uuid.uuid4())
self.logger = setup_logging()
self.auth = AuthMiddleware(config)
self.server = server # 保存server实例的引用
self.auth = AuthMiddleware(config)
self.need_bind = False
self.bind_code = None
self.read_config_from_api = self.config.get("read_config_from_api", False)
@@ -130,8 +130,10 @@ class ConnectionHandler:
if len(cmd) > self.max_cmd_length:
self.max_cmd_length = len(cmd)
self.close_after_chat = False # 是否在聊天结束后关闭连接
self.use_function_call_mode = False
# 是否在聊天结束后关闭连接
self.close_after_chat = False
self.load_function_plugin = False
self.intent_type = "nointent"
self.timeout_task = None
self.timeout_seconds = (
@@ -410,11 +412,11 @@ class ConnectionHandler:
self.memory.init_memory(self.device_id, self.llm)
def _initialize_intent(self):
if (
self.config["Intent"][self.config["selected_module"]["Intent"]]["type"]
== "function_call"
):
self.use_function_call_mode = True
self.intent_type = self.config["Intent"][
self.config["selected_module"]["Intent"]
]["type"]
if self.intent_type == "function_call" or self.intent_type == "intent_llm":
self.load_function_plugin = True
"""初始化意图识别模块"""
# 获取意图识别配置
intent_config = self.config["Intent"]
@@ -802,7 +804,13 @@ class ConnectionHandler:
)
self.dialogue.put(
Message(role="tool", tool_call_id=function_id, content=text)
Message(
role="tool",
tool_call_id=(
str(uuid.uuid4()) if function_id is None else function_id
),
content=text,
)
)
self.chat_with_function_calling(text, tool_call=True)
elif result.action == Action.NOTFOUND or result.action == Action.ERROR:
@@ -5,6 +5,8 @@ from core.handle.sendAudioHandle import send_stt_message
from core.handle.helloHandle import checkWakeupWords
from core.utils.util import remove_punctuation_and_length
from core.utils.dialogue import Message
from plugins_func.register import Action
from loguru import logger
TAG = __name__
@@ -17,7 +19,7 @@ async def handle_user_intent(conn, text):
if await checkWakeupWords(conn, text):
return True
if conn.use_function_call_mode:
if conn.intent_type == "function_call":
# 使用支持function calling的聊天方法,不再进行意图分析
return False
# 使用LLM进行意图分析
@@ -100,24 +102,35 @@ async def process_intent_result(conn, intent_result, original_text):
result = conn.func_handler.handle_llm_function_call(
conn, function_call_data
)
if result and function_name != "play_music":
# 获取当前最新的文本索引
text = result.response
if text is None:
logger.bind(tag=TAG).debug(f"检测到Action : {result.action}")
if result:
if result.action == Action.RESPONSE: # 直接回复前端
text = result.response
if text is not None:
speak_and_play(conn, text)
elif result.action == Action.REQLLM: # 调用函数后再请求llm生成回复
text = result.result
if text is not None:
text_index = (
conn.tts_last_text_index + 1
if hasattr(conn, "tts_last_text_index")
else 0
)
conn.recode_first_last_text(text, text_index)
future = conn.executor.submit(
conn.speak_and_play, text, text_index
)
conn.llm_finish_task = True
conn.tts_queue.put((future, text_index))
conn.dialogue.put(Message(role="assistant", content=text))
conn.dialogue.put(Message(role="tool", content=text))
llm_result = conn.intent.replyResult(text, original_text)
if llm_result is None:
llm_result = text
speak_and_play(conn, llm_result)
elif (
result.action == Action.NOTFOUND
or result.action == Action.ERROR
):
text = result.result
if text is not None:
speak_and_play(conn, text)
elif function_name != "play_music":
# For backward compatibility with original code
# 获取当前最新的文本索引
text = result.response
if text is None:
text = result.result
if text is not None:
speak_and_play(conn, text)
# 将函数执行放在线程池中
conn.executor.submit(process_function_call)
@@ -128,21 +141,12 @@ async def process_intent_result(conn, intent_result, original_text):
return False
def extract_text_in_brackets(s):
"""
从字符串中提取中括号内的文字
:param s: 输入字符串
:return: 中括号内的文字,如果不存在则返回空字符串
"""
left_bracket_index = s.find("[")
right_bracket_index = s.find("]")
if (
left_bracket_index != -1
and right_bracket_index != -1
and left_bracket_index < right_bracket_index
):
return s[left_bracket_index + 1 : right_bracket_index]
else:
return ""
def speak_and_play(conn, text):
text_index = (
conn.tts_last_text_index + 1 if hasattr(conn, "tts_last_text_index") else 0
)
conn.recode_first_last_text(text, text_index)
future = conn.executor.submit(conn.speak_and_play, text, text_index)
conn.llm_finish_task = True
conn.tts_queue.put((future, text_index))
conn.dialogue.put(Message(role="assistant", content=text))
+1 -1
View File
@@ -317,7 +317,7 @@ async def handleIotDescriptors(conn, descriptors):
)
conn.iot_descriptors[descriptor["name"]] = iot_descriptor
if conn.use_function_call_mode:
if conn.load_function_plugin:
# 注册或获取设备类型
type_id = register_device_type(descriptor)
device_functions = device_type_registry.get_device_functions(type_id)
@@ -76,7 +76,7 @@ async def startToChat(conn, text):
# 意图未被处理,继续常规聊天流程
await send_stt_message(conn, text)
if conn.use_function_call_mode:
if conn.intent_type == "function_call":
# 使用支持function calling的聊天方法
conn.executor.submit(conn.chat_with_function_calling, text)
else:
@@ -9,18 +9,6 @@ logger = setup_logging()
class IntentProviderBase(ABC):
def __init__(self, config):
self.config = config
self.intent_options = [
{
"name": "handle_exit_intent",
"desc": "结束聊天, 用户发来如再见之类的表示结束的话, 不想再进行对话的时候",
},
{
"name": "play_music",
"desc": "播放音乐, 用户希望你可以播放音乐, 只用于播放音乐的意图",
},
{"name": "get_time", "desc": "获取今天日期或者当前时间信息"},
{"name": "continue_chat", "desc": "继续聊天"},
]
def set_llm(self, llm):
self.llm = llm
@@ -15,57 +15,72 @@ class IntentProvider(IntentProviderBase):
def __init__(self, config):
super().__init__(config)
self.llm = None
self.promot = self.get_intent_system_prompt()
self.promot = ""
# 添加缓存管理
self.intent_cache = {} # 缓存意图识别结果
self.cache_expiry = 600 # 缓存有效期10分钟
self.cache_max_size = 100 # 最多缓存100个意图
self.history_count = 4 # 默认使用最近4条对话记录
def get_intent_system_prompt(self) -> str:
def get_intent_system_prompt(self, functions_list: str) -> str:
"""
根据配置的意图选项动态生成系统提示词
根据配置的意图选项和可用函数动态生成系统提示词
Args:
functions: 可用的函数列表,JSON格式字符串
Returns:
格式化后的系统提示词
"""
# 构建函数说明部分
functions_desc = "可用的函数列表:\n"
for func in functions_list:
func_info = func.get("function", {})
name = func_info.get("name", "")
desc = func_info.get("description", "")
params = func_info.get("parameters", {})
functions_desc += f"\n函数名: {name}\n"
functions_desc += f"描述: {desc}\n"
if params:
functions_desc += "参数:\n"
for param_name, param_info in params.get("properties", {}).items():
param_desc = param_info.get("description", "")
param_type = param_info.get("type", "")
functions_desc += f"- {param_name} ({param_type}): {param_desc}\n"
functions_desc += "---\n"
prompt = (
"你是一个意图识别助手。请分析用户的最后一句话,判断用户意图属于以下哪一类:\n"
"<start>"
f"{str(self.intent_options)}"
"<end>\n"
"处理步骤:"
"1. 思考意图类型,生成function_call格式"
"\n\n"
"返回格式示例\n"
'1. 播放音乐意图: {"function_call": {"name": "play_music", "arguments": {"song_name": "音乐名称"}}}\n'
'2. 结束对话意图: {"function_call": {"name": "handle_exit_intent", "arguments": {"say_goodbye": "goodbye"}}}\n'
'3. 获取当天日期时间: {"function_call": {"name": "get_time"}}\n'
'4. 继续聊天意图: {"function_call": {"name": "continue_chat"}}\n'
"\n"
"注意:\n"
'- 播放音乐:无歌名时,song_name设为"random"\n'
"- 如果没有明显的意图,应按照继续聊天意图处理\n"
"- 只返回纯JSON,不要任何其他内容\n"
"\n"
"示例分析:\n"
"你是一个意图识别助手。请分析用户的最后一句话,判断用户意图并调用相应的函数。\n\n"
f"{functions_desc}\n"
"处理步骤:\n"
"1. 分析用户输入,确定用户意图\n"
"2. 从可用函数列表中选择最匹配的函数\n"
"3. 如果找到匹配的函数,生成对应的function_call 格式\n"
'4. 如果没有找到匹配的函数,返回{"function_call": {"name": "continue_chat"}}\n\n'
"返回格式要求\n"
"1. 必须返回纯JSON格式\n"
"2. 必须包含function_call字段\n"
"3. function_call必须包含name字段\n"
"4. 如果函数需要参数,必须包含arguments字段\n\n"
"示例:\n"
"```\n"
"用户: 你也太搞笑了\n"
'返回: {"function_call": {"name": "continue_chat"}}\n'
"```\n"
"```\n"
"用户: 现在是几号了?现在几点了?\n"
"用户: 现在几点了?\n"
'返回: {"function_call": {"name": "get_time"}}\n'
"```\n"
"```\n"
"用户: 我们明天再聊吧\n"
'返回: {"function_call": {"name": "handle_exit_intent"}}\n'
"用户: 我想结束对话\n"
'返回: {"function_call": {"name": "handle_exit_intent", "arguments": {"say_goodbye": "goodbye"}}}\n'
"```\n"
"```\n"
"用户: 播放中秋月\n"
'返回: {"function_call": {"name": "play_music", "arguments": {"song_name": "中秋月"}}}\n'
"```\n"
"```\n"
"可用的音乐名称:\n"
"用户: 你好啊\n"
'返回: {"function_call": {"name": "continue_chat"}}\n'
"```\n\n"
"注意:\n"
"1. 只返回JSON格式,不要包含任何其他文字\n"
'2. 如果没有找到匹配的函数,返回{"function_call": {"name": "continue_chat"}}\n'
"3. 确保返回的JSON格式正确,包含所有必要的字段\n"
)
return prompt
@@ -90,6 +105,14 @@ class IntentProvider(IntentProviderBase):
for key, _ in sorted_items[: len(sorted_items) - self.cache_max_size]:
del self.intent_cache[key]
def replyResult(self, text: str, original_text: str):
llm_result = self.llm.response_no_stream(
system_prompt=text,
user_prompt="请根据以上内容,像人类一样说话的口吻回复用户,要求简洁,请直接返回结果。用户现在说:"
+ original_text,
)
return llm_result
async def detect_intent(self, conn, dialogue_history: List[Dict], text: str) -> str:
if not self.llm:
raise ValueError("LLM provider not set")
@@ -118,22 +141,35 @@ class IntentProvider(IntentProviderBase):
# 清理缓存
self.clean_cache()
# 构建用户最后一句话的提示
msgStr = ""
if self.promot == "":
if hasattr(conn, "func_handler"):
functions = conn.func_handler.get_functions()
self.promot = self.get_intent_system_prompt(functions)
# 只使用最后两句即可
if len(dialogue_history) >= 2:
# 保证最少有两句话的时候处理
msgStr += f"{dialogue_history[-2].role}: {dialogue_history[-2].content}\n"
msgStr += f"{dialogue_history[-1].role}: {dialogue_history[-1].content}\n"
msgStr += f"User: {text}\n"
user_prompt = f"当前的对话如下:\n{msgStr}"
music_config = initialize_music_handler(conn)
music_file_names = music_config["music_file_names"]
prompt_music = f"{self.promot}\n<start>{music_file_names}\n<end>"
prompt_music = f"{self.promot}\n<musicNames>{music_file_names}\n</musicNames>"
devices = conn.config["plugins"]["home_assistant"].get("devices", [])
if len(devices) > 0:
hass_prompt = "\n下面是我家智能设备列表(位置,设备名,entity_id),可以通过homeassistant控制\n"
for device in devices:
hass_prompt += device + "\n"
prompt_music += hass_prompt
logger.bind(tag=TAG).debug(f"User prompt: {prompt_music}")
# 构建用户对话历史的提示
msgStr = ""
# 获取最近的对话历史
start_idx = max(0, len(dialogue_history) - self.history_count)
for i in range(start_idx, len(dialogue_history)):
msgStr += f"{dialogue_history[i].role}: {dialogue_history[i].content}\n"
msgStr += f"User: {text}\n"
user_prompt = f"current dialogue:\n{msgStr}"
# 记录预处理完成时间
preprocess_time = time.time() - total_start_time
logger.bind(tag=TAG).debug(f"意图识别预处理耗时: {preprocess_time:.4f}")
+7 -1
View File
@@ -33,7 +33,13 @@ class Dialogue:
dialogue.append({"role": m.role, "tool_calls": m.tool_calls})
elif m.role == "tool":
dialogue.append(
{"role": m.role, "tool_call_id": m.tool_call_id, "content": m.content}
{
"role": m.role,
"tool_call_id": (
str(uuid.uuid4()) if m.tool_call_id is None else m.tool_call_id
),
"content": m.content,
}
)
else:
dialogue.append({"role": m.role, "content": m.content})
@@ -8,10 +8,10 @@ from plugins_func.register import register_function, ToolType, ActionResponse, A
TAG = __name__
logger = setup_logging()
GET_NEWS_FUNCTION_DESC = {
GET_NEWS_FROM_CHINANEWS_FUNCTION_DESC = {
"type": "function",
"function": {
"name": "get_news",
"name": "get_news_from_chinanews",
"description": (
"获取最新新闻,随机选择一条新闻进行播报。"
"用户可以指定新闻类型,如社会新闻、科技新闻、国际新闻等。"
@@ -23,20 +23,20 @@ GET_NEWS_FUNCTION_DESC = {
"properties": {
"category": {
"type": "string",
"description": "新闻类别,例如社会、科技、国际。可选参数,如果不提供则使用默认类别"
"description": "新闻类别,例如社会、科技、国际。可选参数,如果不提供则使用默认类别",
},
"detail": {
"type": "boolean",
"description": "是否获取详细内容,默认为false。如果为true,则获取上一条新闻的详细内容"
"description": "是否获取详细内容,默认为false。如果为true,则获取上一条新闻的详细内容",
},
"lang": {
"type": "string",
"description": "返回用户使用的语言code,例如zh_CN/zh_HK/en_US/ja_JP等,默认zh_CN"
}
"description": "返回用户使用的语言code,例如zh_CN/zh_HK/en_US/ja_JP等,默认zh_CN",
},
},
"required": ["lang"]
}
}
"required": ["lang"],
},
},
}
@@ -51,18 +51,30 @@ def fetch_news_from_rss(rss_url):
# 查找所有item元素(新闻条目)
news_items = []
for item in root.findall('.//item'):
title = item.find('title').text if item.find('title') is not None else "无标题"
link = item.find('link').text if item.find('link') is not None else "#"
description = item.find('description').text if item.find('description') is not None else "无描述"
pubDate = item.find('pubDate').text if item.find('pubDate') is not None else "未知时间"
for item in root.findall(".//item"):
title = (
item.find("title").text if item.find("title") is not None else "无标题"
)
link = item.find("link").text if item.find("link") is not None else "#"
description = (
item.find("description").text
if item.find("description") is not None
else "无描述"
)
pubDate = (
item.find("pubDate").text
if item.find("pubDate") is not None
else "未知时间"
)
news_items.append({
'title': title,
'link': link,
'description': description,
'pubDate': pubDate
})
news_items.append(
{
"title": title,
"link": link,
"description": description,
"pubDate": pubDate,
}
)
return news_items
except Exception as e:
@@ -76,18 +88,24 @@ def fetch_news_detail(url):
response = requests.get(url)
response.raise_for_status()
soup = BeautifulSoup(response.content, 'html.parser')
soup = BeautifulSoup(response.content, "html.parser")
# 尝试提取正文内容 (这里的选择器需要根据实际网站结构调整)
content_div = soup.select_one('.content_desc, .content, article, .article-content')
content_div = soup.select_one(
".content_desc, .content, article, .article-content"
)
if content_div:
paragraphs = content_div.find_all('p')
content = '\n'.join([p.get_text().strip() for p in paragraphs if p.get_text().strip()])
paragraphs = content_div.find_all("p")
content = "\n".join(
[p.get_text().strip() for p in paragraphs if p.get_text().strip()]
)
return content
else:
# 如果找不到特定的内容区域,尝试获取所有段落
paragraphs = soup.find_all('p')
content = '\n'.join([p.get_text().strip() for p in paragraphs if p.get_text().strip()])
paragraphs = soup.find_all("p")
content = "\n".join(
[p.get_text().strip() for p in paragraphs if p.get_text().strip()]
)
return content[:2000] # 限制长度
except Exception as e:
logger.bind(tag=TAG).error(f"获取新闻详情失败: {e}")
@@ -111,7 +129,7 @@ def map_category(category_text):
"财经": "finance",
"财经新闻": "finance",
"金融": "finance",
"经济": "finance"
"经济": "finance",
}
# 转换为小写并去除空格
@@ -121,20 +139,36 @@ def map_category(category_text):
return category_map.get(normalized_category, category_text)
@register_function('get_news', GET_NEWS_FUNCTION_DESC, ToolType.SYSTEM_CTL)
def get_news(conn, category: str = None, detail: bool = False, lang: str = "zh_CN"):
@register_function(
"get_news_from_chinanews",
GET_NEWS_FROM_CHINANEWS_FUNCTION_DESC,
ToolType.SYSTEM_CTL,
)
def get_news_from_chinanews(
conn, category: str = None, detail: bool = False, lang: str = "zh_CN"
):
"""获取新闻并随机选择一条进行播报,或获取上一条新闻的详细内容"""
try:
# 如果detail为True,获取上一条新闻的详细内容
if detail:
if not hasattr(conn, 'last_news_link') or not conn.last_news_link or 'link' not in conn.last_news_link:
return ActionResponse(Action.REQLLM, "抱歉,没有找到最近查询的新闻,请先获取一条新闻。", None)
if (
not hasattr(conn, "last_news_link")
or not conn.last_news_link
or "link" not in conn.last_news_link
):
return ActionResponse(
Action.REQLLM,
"抱歉,没有找到最近查询的新闻,请先获取一条新闻。",
None,
)
link = conn.last_news_link.get('link')
title = conn.last_news_link.get('title', '未知标题')
link = conn.last_news_link.get("link")
title = conn.last_news_link.get("title", "未知标题")
if link == '#':
return ActionResponse(Action.REQLLM, "抱歉,该新闻没有可用的链接获取详细内容。", None)
if link == "#":
return ActionResponse(
Action.REQLLM, "抱歉,该新闻没有可用的链接获取详细内容。", None
)
logger.bind(tag=TAG).debug(f"获取新闻详情: {title}, URL={link}")
@@ -142,8 +176,11 @@ def get_news(conn, category: str = None, detail: bool = False, lang: str = "zh_C
detail_content = fetch_news_detail(link)
if not detail_content or detail_content == "无法获取详细内容":
return ActionResponse(Action.REQLLM,
f"抱歉,无法获取《{title}》的详细内容,可能是链接已失效或网站结构发生变化。", None)
return ActionResponse(
Action.REQLLM,
f"抱歉,无法获取《{title}》的详细内容,可能是链接已失效或网站结构发生变化。",
None,
)
# 构建详情报告
detail_report = (
@@ -158,8 +195,10 @@ def get_news(conn, category: str = None, detail: bool = False, lang: str = "zh_C
# 否则,获取新闻列表并随机选择一条
# 从配置中获取RSS URL
rss_config = conn.config["plugins"]["get_news"]
default_rss_url = rss_config.get("default_rss_url", "https://www.chinanews.com.cn/rss/society.xml")
rss_config = conn.config["plugins"]["get_news_from_chinanews"]
default_rss_url = rss_config.get(
"default_rss_url", "https://www.chinanews.com.cn/rss/society.xml"
)
# 将用户输入的类别映射到配置中的类别键
mapped_category = map_category(category)
@@ -169,23 +208,27 @@ def get_news(conn, category: str = None, detail: bool = False, lang: str = "zh_C
if mapped_category and mapped_category in rss_config.get("category_urls", {}):
rss_url = rss_config["category_urls"][mapped_category]
logger.bind(tag=TAG).info(f"获取新闻: 原始类别={category}, 映射类别={mapped_category}, URL={rss_url}")
logger.bind(tag=TAG).info(
f"获取新闻: 原始类别={category}, 映射类别={mapped_category}, URL={rss_url}"
)
# 获取新闻列表
news_items = fetch_news_from_rss(rss_url)
if not news_items:
return ActionResponse(Action.REQLLM, "抱歉,未能获取到新闻信息,请稍后再试。", None)
return ActionResponse(
Action.REQLLM, "抱歉,未能获取到新闻信息,请稍后再试。", None
)
# 随机选择一条新闻
selected_news = random.choice(news_items)
# 保存当前新闻链接到连接对象,以便后续查询详情
if not hasattr(conn, 'last_news_link'):
if not hasattr(conn, "last_news_link"):
conn.last_news_link = {}
conn.last_news_link = {
'link': selected_news.get('link', '#'),
'title': selected_news.get('title', '未知标题')
"link": selected_news.get("link", "#"),
"title": selected_news.get("title", "未知标题"),
}
# 构建新闻报告
@@ -203,4 +246,6 @@ def get_news(conn, category: str = None, detail: bool = False, lang: str = "zh_C
except Exception as e:
logger.bind(tag=TAG).error(f"获取新闻出错: {e}")
return ActionResponse(Action.REQLLM, "抱歉,获取新闻时发生错误,请稍后再试。", None)
return ActionResponse(
Action.REQLLM, "抱歉,获取新闻时发生错误,请稍后再试。", None
)
@@ -0,0 +1,214 @@
import random
import requests
import json
from config.logger import setup_logging
from plugins_func.register import register_function, ToolType, ActionResponse, Action
from markitdown import MarkItDown
TAG = __name__
logger = setup_logging()
# 新闻来源字典,包含名称和对应的API ID
NEWS_SOURCES = {
"thepaper": "澎湃新闻",
"baidu": "百度热搜",
"cls-depth": "财联社",
}
# 动态生成新闻源描述
def generate_news_sources_description():
sources_desc = []
for source_id, source_name in NEWS_SOURCES.items():
sources_desc.append(f"{source_name}({source_id})")
return "".join(sources_desc)
GET_NEWS_FROM_NEWSNOW_FUNCTION_DESC = {
"type": "function",
"function": {
"name": "get_news_from_newsnow",
"description": (
"获取最新新闻,随机选择一条新闻进行播报。"
f"用户可以选择不同的新闻源,如{generate_news_sources_description()}等。"
"如果没有指定,默认从澎湃新闻获取。"
"用户可以要求获取详细内容,此时会获取新闻的详细内容。"
),
"parameters": {
"type": "object",
"properties": {
"source": {
"type": "string",
"description": f"新闻源,例如{generate_news_sources_description()}等。可选参数,如果不提供则使用默认新闻源",
},
"detail": {
"type": "boolean",
"description": "是否获取详细内容,默认为false。如果为true,则获取上一条新闻的详细内容",
},
"lang": {
"type": "string",
"description": "返回用户使用的语言code,例如zh_CN/zh_HK/en_US/ja_JP等,默认zh_CN",
},
},
"required": ["lang"],
},
},
}
def fetch_news_from_api(conn, source="thepaper"):
"""从API获取新闻列表"""
try:
api_url = f"https://newsnow.busiyi.world/api/s?id={source}"
if conn.config["plugins"].get("get_news_from_newsnow") and conn.config[
"plugins"
]["get_news_from_newsnow"].get("url"):
api_url = conn.config["plugins"]["get_news_from_newsnow"]["url"] + source
response = requests.get(api_url, timeout=10)
response.raise_for_status()
data = response.json()
if "items" in data:
return data["items"]
else:
logger.bind(tag=TAG).error(f"获取新闻API响应格式错误: {data}")
return []
except Exception as e:
logger.bind(tag=TAG).error(f"获取新闻API失败: {e}")
return []
def fetch_news_detail(url):
"""获取新闻详情页内容并使用MarkItDown清理HTML"""
try:
response = requests.get(url, timeout=10)
response.raise_for_status()
# 使用MarkItDown清理HTML内容
md = MarkItDown(enable_plugins=False)
result = md.convert(response)
# 获取清理后的文本内容
clean_text = result.text_content
# 如果清理后的内容为空,返回提示信息
if not clean_text or len(clean_text.strip()) == 0:
logger.bind(tag=TAG).warning(f"清理后的新闻内容为空: {url}")
return "无法解析新闻详情内容,可能是网站结构特殊或内容受限。"
return clean_text
except Exception as e:
logger.bind(tag=TAG).error(f"获取新闻详情失败: {e}")
return "无法获取详细内容"
@register_function(
"get_news_from_newsnow",
GET_NEWS_FROM_NEWSNOW_FUNCTION_DESC,
ToolType.SYSTEM_CTL,
)
def get_news_from_newsnow(
conn, source: str = "thepaper", detail: bool = False, lang: str = "zh_CN"
):
"""获取新闻并随机选择一条进行播报,或获取上一条新闻的详细内容"""
try:
# 如果detail为True,获取上一条新闻的详细内容
detail = str(detail).lower() == "true"
if detail:
if (
not hasattr(conn, "last_newsnow_link")
or not conn.last_newsnow_link
or "url" not in conn.last_newsnow_link
):
return ActionResponse(
Action.REQLLM,
"抱歉,没有找到最近查询的新闻,请先获取一条新闻。",
None,
)
url = conn.last_newsnow_link.get("url")
title = conn.last_newsnow_link.get("title", "未知标题")
source_id = conn.last_newsnow_link.get("source_id", "thepaper")
source_name = NEWS_SOURCES.get(source_id, "未知来源")
if not url or url == "#":
return ActionResponse(
Action.REQLLM, "抱歉,该新闻没有可用的链接获取详细内容。", None
)
logger.bind(tag=TAG).debug(
f"获取新闻详情: {title}, 来源: {source_name}, URL={url}"
)
# 获取新闻详情
detail_content = fetch_news_detail(url)
if not detail_content or detail_content == "无法获取详细内容":
return ActionResponse(
Action.REQLLM,
f"抱歉,无法获取《{title}》的详细内容,可能是链接已失效或网站结构发生变化。",
None,
)
# 构建详情报告
detail_report = (
f"根据下列数据,用{lang}回应用户的新闻详情查询请求:\n\n"
f"新闻标题: {title}\n"
# f"新闻来源: {source_name}\n"
f"详细内容: {detail_content}\n\n"
f"(请对上述新闻内容进行总结,提取关键信息,以自然、流畅的方式向用户播报,"
f"不要提及这是总结,就像是在讲述一个完整的新闻故事)"
)
return ActionResponse(Action.REQLLM, detail_report, None)
# 否则,获取新闻列表并随机选择一条
# 验证新闻源是否有效,如果无效则使用默认源
if source not in NEWS_SOURCES:
logger.bind(tag=TAG).warning(f"无效的新闻源: {source},使用默认源thepaper")
source = "thepaper"
source_name = NEWS_SOURCES.get(source, "澎湃新闻")
logger.bind(tag=TAG).info(f"获取新闻: 新闻源={source}({source_name})")
# 获取新闻列表
news_items = fetch_news_from_api(conn, source)
if not news_items:
return ActionResponse(
Action.REQLLM,
f"抱歉,未能从{source_name}获取到新闻信息,请稍后再试或尝试其他新闻源。",
None,
)
# 随机选择一条新闻
selected_news = random.choice(news_items)
# 保存当前新闻链接到连接对象,以便后续查询详情
if not hasattr(conn, "last_newsnow_link"):
conn.last_newsnow_link = {}
conn.last_newsnow_link = {
"url": selected_news.get("url", "#"),
"title": selected_news.get("title", "未知标题"),
"source_id": source,
}
# 构建新闻报告
news_report = (
f"根据下列数据,用{lang}回应用户的新闻查询请求:\n\n"
f"新闻标题: {selected_news['title']}\n"
# f"新闻来源: {source_name}\n"
f"(请以自然、流畅的方式向用户播报这条新闻标题,"
f"提示用户可以要求获取详细内容,此时会获取新闻的详细内容。)"
)
return ActionResponse(Action.REQLLM, news_report, None)
except Exception as e:
logger.bind(tag=TAG).error(f"获取新闻出错: {e}")
return ActionResponse(
Action.REQLLM, "抱歉,获取新闻时发生错误,请稍后再试。", None
)
@@ -8,7 +8,7 @@ HASS_CACHE = {}
def append_devices_to_prompt(conn):
if conn.use_function_call_mode:
if conn.intent_type == "function_call":
funcs = conn.config["Intent"][conn.config["selected_module"]["Intent"]].get(
"functions", []
)
@@ -27,7 +27,7 @@ def append_devices_to_prompt(conn):
def initialize_hass_handler(conn):
global HASS_CACHE
if HASS_CACHE == {}:
if conn.use_function_call_mode:
if conn.load_function_plugin:
funcs = conn.config["Intent"][conn.config["selected_module"]["Intent"]].get(
"functions", []
)
+2 -1
View File
@@ -28,4 +28,5 @@ PySocks==1.7.1
dashscope==1.23.1
baidu-aip==4.16.13
chardet==5.2.0
aioconsole==0.8.1
aioconsole==0.8.1
markitdown==0.1.1