3.3

1. Add more translations
2. Improve AI

.gitignore

@@ -11,6 +11,7 @@ tests/tools/*
 tests/rss/*
 tests/debug_mail/*
 tests/cov_html/*
+tests/pobackup/*
 .idea/
 datastore/
 .coverage

application/lib/ebook_summarizer/html_summarizer.py

@@ -7,21 +7,21 @@
 from application.ke_utils import loc_exc_pos
 
 def get_summarizer_engines():
-    return simple_ai_provider._PROV_AI_LIST
+    return simple_ai_provider.AI_LIST
 
 class HtmlSummarizer:
     def __init__(self, params: dict):
         self.params = params
         name = self.params.get('engine')
-        if name not in simple_ai_provider._PROV_AI_LIST:
+        if name not in simple_ai_provider.AI_LIST:
             default_log.warning(f'Unsupported provider {name}, fallback to gemini')
             name = 'gemini'
         self.aiAgent = self.create_engine(name, params)
 
     #创建一个AI封装实例
     def create_engine(self, name, params):
         return simple_ai_provider.SimpleAiProvider(name, params.get('api_key', ''), 
-            model=params.get('model', ''), api_host=params.get('api_host', ''))
+            model=params.get('model', ''), apiHost=params.get('api_host', ''))
 
     #给一段文字做摘要，记住不要太长
     #返回 {'error': '', 'summary': ''}

application/lib/simple_ai_provider.py

@@ -3,140 +3,161 @@
 #简单封装几个流行的AI服务接口，对外提供一个统一的接口
 #Author: cdhigh <https://github.com/cdhigh>
 # 使用示例：
-#provider = SimpleAiProvider("openai", api_key="xxxxxx")
+#provider = SimpleAiProvider("openai", apiKey="xxxxxx")
 #response = provider.chat("你好，请讲一个笑话")
 #response = provider.chat([{"role": "system", "content": "你是一个专业的物理学家。"},{"role": "user", "content": "黑洞是怎么形成的？"}])
 #import requests
+from urllib.parse import urljoin
 from urlopener import UrlOpener
 
 #支持的AI服务商列表，models里面的第一项请设置为默认要使用的model
 #rpm(requests per minute)是针对免费用户的，如果是付费用户，一般会高很多，可以自己修改
 #大语言模型发展迅速，估计没多久这些数据会全部过时
 #{'name': '', 'rpm': , 'context': },
-_PROV_AI_LIST = {
-    'Gemini': [
+AI_LIST = {
+    'google': {'host': 'https://generativelanguage.googleapis.com', 'models': [
         {'name': 'gemini-1.5-flash', 'rpm': 15, 'context': 128000}, #其实支持100万
         {'name': 'gemini-1.5-flash-8b', 'rpm': 15, 'context': 128000}, 
-        {'name': 'gemini-1.5-pro', 'rpm': 2, 'context': 128000},],
-    'Openai': [
+        {'name': 'gemini-1.5-pro', 'rpm': 2, 'context': 128000},],},
+    'openai': {'host': 'https://api.openai.com', 'models': [
         {'name': 'gpt-4o-mini', 'rpm': 3, 'context': 128000},
         {'name': 'gpt-4o', 'rpm': 3, 'context': 128000},
         {'name': 'gpt-4-turbo', 'rpm': 3, 'context': 128000},
         {'name': 'gpt-3.5-turbo', 'rpm': 3, 'context': 16000},
-        {'name': 'gpt-3.5-turbo-instruct', 'rpm': 3, 'context': 4000},],
-    'Anthropic': [
+        {'name': 'gpt-3.5-turbo-instruct', 'rpm': 3, 'context': 4000},],},
+    'anthropic': {'host': 'https://api.anthropic.com', 'models': [
         {'name': 'claude-2', 'rpm': 5, 'context': 100000},
         {'name': 'claude-3', 'rpm': 5, 'context': 200000},
-        {'name': 'claude-2.1', 'rpm': 5, 'context': 100000},],
-    'Grok': [
-        {'name': 'grok-beta', 'rpm': 60, 'context': 128000},],
-    'Mistral': [
+        {'name': 'claude-2.1', 'rpm': 5, 'context': 100000},],},
+    'xai': {'host': 'https://api.x.ai', 'models': [
+        {'name': 'grok-beta', 'rpm': 60, 'context': 128000},
+        {'name': 'grok-2', 'rpm': 60, 'context': 128000},],},
+    'mistral': {'host': 'https://api.mistral.ai', 'models': [
         {'name': 'open-mistral-7b', 'rpm': 60, 'context': 32000},
         {'name': 'mistral-small-latest', 'rpm': 60, 'context': 32000},
         {'name': 'open-mixtral-8x7b', 'rpm': 60, 'context': 32000},
         {'name': 'open-mixtral-8x22b', 'rpm': 60, 'context': 64000},
         {'name': 'mistral-medium-latest', 'rpm': 60, 'context': 32000},
         {'name': 'mistral-large-latest', 'rpm': 60, 'context': 128000},
-        {'name': 'pixtral-12b-2409', 'rpm': 60, 'context': 128000},],
-    'Groq': [
+        {'name': 'pixtral-12b-2409', 'rpm': 60, 'context': 128000},],},
+    'groq': {'host': 'https://api.groq.com', 'models': [
         {'name': 'gemma2-9b-it', 'rpm': 30, 'context': 8000},
         {'name': 'gemma-7b-it', 'rpm': 30, 'context': 8000},
         {'name': 'llama-guard-3-8b', 'rpm': 30, 'context': 8000},
         {'name': 'llama3-70b-8192', 'rpm': 30, 'context': 8000},
         {'name': 'llama3-8b-8192', 'rpm': 30, 'context': 8000},
-        {'name': 'mixtral-8x7b-32768', 'rpm': 30, 'context': 32000},],
-    'Alibaba': [
+        {'name': 'mixtral-8x7b-32768', 'rpm': 30, 'context': 32000},],},
+    'perplexity': {'host': 'https://api.perplexity.ai', 'models': [
+        {'name': 'llama-3.1-sonar-small-128k-online', 'rpm': 60, 'context': 128000},
+        {'name': 'llama-3.1-sonar-large-128k-online', 'rpm': 60, 'context': 128000},
+        {'name': 'llama-3.1-sonar-huge-128k-online', 'rpm': 60, 'context': 128000},],},
+    'alibaba': {'host': 'https://dashscope.aliyuncs.com', 'models': [
         {'name': 'qwen-turbo', 'rpm': 60, 'context': 128000}, #其实支持100万
         {'name': 'qwen-plus', 'rpm': 60, 'context': 128000},
         {'name': 'qwen-long', 'rpm': 60, 'context': 128000},
-        {'name': 'qwen-max', 'rpm': 60, 'context': 32000},],
+        {'name': 'qwen-max', 'rpm': 60, 'context': 32000},],},
 }
 
 class SimpleAiProvider:
     #name: AI提供商的名字
-    def __init__(self, name, api_key, model=None, api_host=None):
-        if name not in _PROV_AI_LIST:
+    #apiHost: 支持自搭建的API转发服务器，传入以分号分割的地址列表字符串，则逐个使用
+    #singleTurn: 一些API转发服务不支持多轮对话模式，设置此标识，当前仅支持 openai
+    def __init__(self, name, apiKey, model=None, apiHost=None, singleTurn=False):
+        name = name.lower()
+        if name not in AI_LIST:
             raise ValueError(f"Unsupported provider: {name}")
         self.name = name
-        self.api_key = api_key
+        self.apiKey = apiKey
+        self.singleTurn = singleTurn
+        self.apiHosts = (apiHost or AI_LIST[name]['host']).split(';')
+        self.hostIdx = 0
+        self.host = '' #如果提供多个api host，这个变量保存当前使用的host
+        self._models = AI_LIST[name]['models']
 
-        index = 0
-        for idx, item in enumerate(_PROV_AI_LIST[name]):
-            if model == item['name']:
-                index = idx
-                break
+        #如果传入的model不在列表中，默认使用第一个
+        item = next((m for m in self._models if m['name'] == model), self._models[0])
+        self.model = item['name']
+        self.rpm = item['rpm']
+        self.context_size = item['context']
+        if self.rpm <= 0:
+            self.rpm = 2
+        if self.context_size < 1000:
+            self.context_size = 1000
 
-        item = _PROV_AI_LIST[name][index]
-        self._model = item['name']
-        self._rpm = item['rpm']
-        self._context = item['context']
-        if self._rpm <= 0:
-            self._rpm = 2
-        if self._context < 4000:
-            self._context = 4000
-        self.api_host = api_host
         self.opener = UrlOpener()
 
     @property
-    def model(self):
-        return self._model
-    @property
-    def rpm(self):
-        return self._rpm
-    @property
     def request_interval(self):
-        return (60 / self._rpm) if (self._rpm > 0) else 30
-    @property
-    def context_size(self):
-        return self._context
+        return (60 / self.rpm) if (self.rpm >= 1) else 20
 
     def __repr__(self):
-        return f'{self.name}({self._model})'
+        return f'{self.name}/{self.model}'
 
     #返回支持的AI供应商列表，返回一个python字典
     def ai_list(self):
-        return _PROV_AI_LIST
+        return AI_LIST
+
+    #返回需要使用的api host，自动轮换使用多host
+    @property
+    def apiHost(self):
+        self.host = self.apiHosts[self.hostIdx]
+        self.hostIdx += 1
+        if self.hostIdx >= len(self.apiHosts):
+            self.hostIdx = 0
+        return self.host
 
     #外部调用此函数即可调用简单聊天功能
     #message: 如果是文本，则使用各项默认参数
     #传入 list/dict 可以定制 role 等参数
+    #返回 respTxt，如果要获取当前使用的主机，可以使用 host 属性
     def chat(self, message):
         name = self.name
-        if name == "Openai":
+        if name == "openai":
             return self._openai_chat(message)
-        elif name == "Anthropic":
+        elif name == "anthropic":
             return self._anthropic_chat(message)
-        elif name == "Gemini":
-            return self._gemini_chat(message)
-        elif name == "Grok":
-            return self._grok_chat(message)
-        elif name == "Mistral":
+        elif name == "google":
+            return self._google_chat(message)
+        elif name == "xai":
+            return self._xai_chat(message)
+        elif name == "mistral":
             return self._mistral_chat(message)
-        elif name == 'Groq':
+        elif name == 'groq':
             return self._groq_chat(message)
-        elif name == "Alibaba":
+        elif name == 'perplexity':
+            ret = self._perplexity_chat(message)
+        elif name == "alibaba":
             return self._alibaba_chat(message)
         else:
             raise ValueError(f"Unsupported provider: {name}")
 
     #openai的chat接口
-    def _openai_chat(self, message, defaultUrl='https://api.openai.com/v1/chat/completions'):
-        url = self.api_host if self.api_host else defaultUrl
-        headers = {'Authorization': f"Bearer {self.api_key}",
-            'Content-Type': 'application/json'}
-        payload = {
-            "model": self._model,
-            "messages": [{"role": "user", "content": message}] if isinstance(message, str) else message
-        }
-        response = self.opener.post(url, headers=headers, json=payload)
-        response.raise_for_status()
-        return response.json()["choices"][0]["message"]["content"]
+    def _openai_chat(self, message, path='v1/chat/completions'):
+        url = urljoin(self.apiHost, path)
+        headers = {'Authorization': f'Bearer {self.apiKey}', 'Content-Type': 'application/json'}
+        if isinstance(message, str):
+            msg = [{"role": "user", "content": message}]
+        elif self.singleTurn and (len(message) > 1): #将多轮对话手动拼接为单一轮对话
+            msgArr = ['Previous conversions:\n']
+            roleMap = {'system': 'background', 'assistant': 'Your responsed'}
+            msgArr.extend([f'{roleMap.get(e["role"], "I asked")}:\n{e["content"]}\n' for e in message[:-1]])
+            msgArr.append(f'\nPlease continue this conversation based on the previous information:\n')
+            msgArr.append("I ask:")
+            msgArr.append(message[-1]['content'])
+            msgArr.append("You Response:\n")
+            msg = [{"role": "user", "content": '\n'.join(msgArr)}]
+        else:
+            msg = message
+        payload = {"model": self.model, "messages": msg}
+        resp = self.opener.post(url, headers=headers, json=payload)
+        resp.raise_for_status()
+        return resp.json()["choices"][0]["message"]["content"]
 
     #anthropic的chat接口
-    def _anthropic_chat(self, message):
-        url = self.api_host if self.api_host else 'https://api.anthropic.com/v1/complete'
+    def _anthropic_chat(self, message, path='v1/complete'):
+        url = urljoin(self.apiHost, path)
         headers = {'Accept': 'application/json', 'Anthropic-Version': '2023-06-01',
-            'Content-Type': 'application/json', 'x-api-key': self.api_key}
+            'Content-Type': 'application/json', 'x-api-key': self.apiKey}
 
         if isinstance(message, list): #将openai的payload格式转换为anthropic的格式
             msg = []
@@ -152,17 +173,13 @@ def _anthropic_chat(self, message):
             prompt = f"\n\nHuman: {message}\n\nAssistant:"
             payload = {"prompt": prompt, "model": self.model, "max_tokens_to_sample": 256}
 
-        response = self.opener.post(url, headers=headers, json=payload)
-        response.raise_for_status()
-        return response.json()["completion"]
+        resp = self.opener.post(url, headers=headers, json=payload)
+        resp.raise_for_status()
+        return resp.json()["completion"]
 
-    #gemini的chat接口
-    def _gemini_chat(self, message):
-        if self.api_host:
-            url = f'{self.api_host}?key={self.api_key}'
-        else:
-            url = f'https://generativelanguage.googleapis.com/v1beta/models/{self._model}:generateContent?key={self.api_key}'
-
+    #google的chat接口
+    def _google_chat(self, message):
+        url = urljoin(self.apiHost, f'v1beta/models/{self.model}:generateContent?key={self.apiKey}')
         if isinstance(message, list): #将openai的payload格式转换为gemini的格式
             msg = []
             for item in message:
@@ -174,45 +191,32 @@ def _gemini_chat(self, message):
             payload = message
         else:
             payload = {'contents': [{'role': 'user', 'parts': [{'text': message}]}]}
-        response = self.opener.post(url, json=payload)
-        response.raise_for_status()
-        contents = response.json()["candidates"][0]["content"]
+        resp = self.opener.post(url, json=payload)
+        resp.raise_for_status()
+        contents = resp.json()["candidates"][0]["content"]
         return contents['parts'][0]['text']
 
-    #grok的chat接口
-    def _grok_chat(self, message):
-        #直接使用openai兼容接口
-        return self._openai_chat(message, defaultUrl='https://api.x.ai/v1/chat/completions')
+    #xai的chat接口
+    def _xai_chat(self, message):
+        return self._openai_chat(message, path='v1/chat/completions')
 
     #mistral的chat接口
     def _mistral_chat(self, message):
-        #直接使用openai兼容接口
-        return self._openai_chat(message, defaultUrl='https://api.mistral.ai/v1/chat/completions')
-
+        return self._openai_chat(message, path='v1/chat/completions')
+
     #groq的chat接口
     def _groq_chat(self, message):
-        #直接使用openai兼容接口
-        return self._openai_chat(message, defaultUrl='https://api.groq.com/openai/v1/chat/completions')
+        return self._openai_chat(message, path='openai/v1/chat/completions')
+
+    #perplexity的chat接口
+    def _perplexity_chat(self, message):
+        return self._openai_chat(message, path='chat/completions')
 
     #通义千问
     def _alibaba_chat(self, message):
-        #直接使用openai兼容接口
-        return self._openai_chat(message, defaultUrl='https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions')
+        return self._openai_chat(message, path='compatible-mode/v1/chat/completions')
 
-    def _baidu_chat(self, message):
-        url = self.api_host if self.api_host else 'https://aip.baidubce.com/rpc/2.0/ai_custom/v1/ernie-bot'
-        headers = {"Content-Type": "application/json"}
-        params = {"access_token": self.api_key}
-        payload = {
-            "messages": [{"role": "user", "content": message}] if isinstance(message, str) else message,
-            "model": self._model,
-            "max_tokens": 300
-        }
-        response = self.opener.post(url, headers=headers, params=params, json=payload)
-        response.raise_for_status()
-        return response.json()["result"]["content"]
-
 if __name__ == '__main__':
-    provider = SimpleAiProvider("gemini", api_key="xxx")
-    response = provider.chat("你好，请讲一个笑话")
-    print(response)
+    client = SimpleAiProvider("gemini", apiKey="xxx")
+    resp = client.chat("你好，请讲一个笑话")
+    print(resp)

application/static/base.css

@@ -134,6 +134,20 @@ button {
     padding: 1em .5em;
 }
 
+/* 语言选择框 */
+.home-menu .language-select {
+    border: none;
+    background: none;
+    outline: none;
+    padding: 5px;
+    font-size: 0.6em;
+    color: #333;
+    text-align: center;
+}
+.home-menu .language-select:focus {
+    outline: none;
+}
+
 .app-menu {
     background-color: #0078e7;
 }

application/static/base.js

@@ -94,6 +94,16 @@ function MakeAjaxRequest(url, method, data, callback) {
   });
 }
 
+///[start] base.html
+//选择其他语种进行页面显示
+function handleLanguageChange() {
+  const lang = $('#languageSelect').val();
+  MakeAjaxRequest(`/setlocale`, "POST", {'lang': lang}, function (resp) {
+    location.reload(); //刷新页面
+  });
+}
+///[end] base.html
+
 ///[start] my.html
 var show_menu_box = false;
 
@@ -1620,7 +1630,7 @@ function PopulateSummarizerFields(currEngineName) {
 function SummarizerEngineFieldChanged(model) {
   let hasSelected = false;
   var engineName = $('#summarizer_engine').val();
-  var models = g_ai_engines[engineName] || [];
+  var models = g_ai_engines[engineName].models || [];
 
   let modelSel = $('#summarizer_model');
   modelSel.empty();