Merge branch 'main' into chuanhuAgent

README.md

@@ -42,8 +42,22 @@
 
 ## 目录
 
-| [使用技巧](#使用技巧) | [安装方式](https://github.com/GaiZhenbiao/ChuanhuChatGPT/wiki/使用教程) | [常见问题](https://github.com/GaiZhenbiao/ChuanhuChatGPT/wiki/常见问题) | [给作者买可乐🥤](#捐款) |
-| ------------------ | -------------------------------------------------------------------- | -------------------------------------------------------------------- | -------------------- |
+| [支持模型](#支持模型) | [使用技巧](#使用技巧) | [安装方式](https://github.com/GaiZhenbiao/ChuanhuChatGPT/wiki/使用教程) | [常见问题](https://github.com/GaiZhenbiao/ChuanhuChatGPT/wiki/常见问题) | [给作者买可乐🥤](#捐款) |
+| ----- | ----- | ----- | ----- | ----- |
+
+
+## 支持模型
+**通过API调用的语言模型**：
+- [ChatGPT](https://chat.openai.com) ([GPT-4](https://openai.com/product/gpt-4))
+- [Inspur Yuan 1.0](https://air.inspur.com/home)
+- [MiniMax](https://api.minimax.chat/)
+- [XMChat](https://github.com/MILVLG/xmchat)
+
+**本地部署语言模型**：
+- [ChatGLM](https://github.com/THUDM/ChatGLM-6B)
+- [LLaMA](https://github.com/facebookresearch/llama)
+- [StableLM](https://github.com/Stability-AI/StableLM)
+- [MOSS](https://github.com/OpenLMLab/MOSS)
 
 ## 使用技巧
 
@@ -64,7 +78,7 @@ cd ChuanhuChatGPT
 pip install -r requirements.txt
 ```
 
-在项目文件夹中复制一份 `config_example.json`，并将其重命名为 `config.json`，在其中填入 `API-Key` 等设置。
+然后，在项目文件夹中复制一份 `config_example.json`，并将其重命名为 `config.json`，在其中填入 `API-Key` 等设置。
 
 ```shell
 python ChuanhuChatbot.py
@@ -78,7 +92,7 @@ python ChuanhuChatbot.py
 
 ## 疑难杂症解决
 
-在遇到各种问题查阅相关信息前，您可以先尝试手动拉取本项目的最新更改并更新 gradio，然后重试。步骤为：
+在遇到各种问题查阅相关信息前，您可以先尝试手动拉取本项目的最新更改并更新依赖库，然后重试。步骤为：
 
 1. 点击网页上的 `Download ZIP` 下载最新代码，或
    ```shell
@@ -88,10 +102,6 @@ python ChuanhuChatbot.py
    ```
    pip install -r requirements.txt
    ```
-3. 更新gradio
-   ```
-   pip install gradio --upgrade --force-reinstall
-   ```
 
 很多时候，这样就可以解决问题。
 

config_example.json

@@ -6,6 +6,9 @@
     "xmchat_api_key": "",
     "usage_limit": 120, // API Key的当月限额，单位：美元
     // 你的xmchat API Key，与OpenAI API Key不同
+    // MiniMax的APIKey（见账户管理页面 https://api.minimax.chat/basic-information）和Group ID，用于MiniMax对话模型
+    "minimax_api_key": "",
+    "minimax_group_id": "",
     "language": "auto",
     // 如果使用代理，请取消注释下面的两行，并替换代理URL
     // "https_proxy": "http://127.0.0.1:1079",

modules/config.py

@@ -77,6 +77,11 @@ my_api_key = os.environ.get("OPENAI_API_KEY", my_api_key)
 xmchat_api_key = config.get("xmchat_api_key", "")
 os.environ["XMCHAT_API_KEY"] = xmchat_api_key
 
+minimax_api_key = config.get("minimax_api_key", "")
+os.environ["MINIMAX_API_KEY"] = minimax_api_key
+minimax_group_id = config.get("minimax_group_id", "")
+os.environ["MINIMAX_GROUP_ID"] = minimax_group_id
+
 render_latex = config.get("render_latex", True)
 
 if render_latex:

modules/models/base_model.py

@@ -107,7 +107,8 @@ class ModelType(Enum):
     StableLM = 4
     MOSS = 5
     YuanAI = 6
-    ChuanhuAgent = 7
+    Minimax = 7
+    ChuanhuAgent = 8
 
     @classmethod
     def get_type(cls, model_name: str):
@@ -127,6 +128,8 @@ class ModelType(Enum):
             model_type = ModelType.MOSS
         elif "yuanai" in model_name_lower:
             model_type = ModelType.YuanAI
+        elif "minimax" in model_name_lower:
+            model_type = ModelType.Minimax
         elif "川虎助理" in model_name_lower:
             model_type = ModelType.ChuanhuAgent
         else:

modules/models/minimax.py

@@ -0,0 +1,161 @@
+import json
+import os
+
+import colorama
+import requests
+import logging
+
+from modules.models.base_model import BaseLLMModel
+from modules.presets import STANDARD_ERROR_MSG, GENERAL_ERROR_MSG, TIMEOUT_STREAMING, TIMEOUT_ALL, i18n
+
+group_id = os.environ.get("MINIMAX_GROUP_ID", "")
+
+
+class MiniMax_Client(BaseLLMModel):
+    """
+    MiniMax Client
+    接口文档见 https://api.minimax.chat/document/guides/chat
+    """
+
+    def __init__(self, model_name, api_key, user_name="", system_prompt=None):
+        super().__init__(model_name=model_name, user=user_name)
+        self.url = f'https://api.minimax.chat/v1/text/chatcompletion?GroupId={group_id}'
+        self.history = []
+        self.api_key = api_key
+        self.system_prompt = system_prompt
+        self.headers = {
+            "Authorization": f"Bearer {api_key}",
+            "Content-Type": "application/json"
+        }
+
+    def get_answer_at_once(self):
+        # minimax temperature is (0,1] and base model temperature is [0,2], and yuan 0.9 == base 1 so need to convert
+        temperature = self.temperature * 0.9 if self.temperature <= 1 else 0.9 + (self.temperature - 1) / 10
+
+        request_body = {
+            "model": self.model_name.replace('minimax-', ''),
+            "temperature": temperature,
+            "skip_info_mask": True,
+            'messages': [{"sender_type": "USER", "text": self.history[-1]['content']}]
+        }
+        if self.n_choices:
+            request_body['beam_width'] = self.n_choices
+        if self.system_prompt:
+            request_body['prompt'] = self.system_prompt
+        if self.max_generation_token:
+            request_body['tokens_to_generate'] = self.max_generation_token
+        if self.top_p:
+            request_body['top_p'] = self.top_p
+
+        response = requests.post(self.url, headers=self.headers, json=request_body)
+
+        res = response.json()
+        answer = res['reply']
+        total_token_count = res["usage"]["total_tokens"]
+        return answer, total_token_count
+
+    def get_answer_stream_iter(self):
+        response = self._get_response(stream=True)
+        if response is not None:
+            iter = self._decode_chat_response(response)
+            partial_text = ""
+            for i in iter:
+                partial_text += i
+                yield partial_text
+        else:
+            yield STANDARD_ERROR_MSG + GENERAL_ERROR_MSG
+
+    def _get_response(self, stream=False):
+        minimax_api_key = self.api_key
+        history = self.history
+        logging.debug(colorama.Fore.YELLOW +
+                      f"{history}" + colorama.Fore.RESET)
+        headers = {
+            "Content-Type": "application/json",
+            "Authorization": f"Bearer {minimax_api_key}",
+        }
+
+        temperature = self.temperature * 0.9 if self.temperature <= 1 else 0.9 + (self.temperature - 1) / 10
+
+        messages = []
+        for msg in self.history:
+            if msg['role'] == 'user':
+                messages.append({"sender_type": "USER", "text": msg['content']})
+            else:
+                messages.append({"sender_type": "BOT", "text": msg['content']})
+
+        request_body = {
+            "model": self.model_name.replace('minimax-', ''),
+            "temperature": temperature,
+            "skip_info_mask": True,
+            'messages': messages
+        }
+        if self.n_choices:
+            request_body['beam_width'] = self.n_choices
+        if self.system_prompt:
+            lines = self.system_prompt.splitlines()
+            if lines[0].find(":") != -1 and len(lines[0]) < 20:
+                request_body["role_meta"] = {
+                    "user_name": lines[0].split(":")[0],
+                    "bot_name": lines[0].split(":")[1]
+                }
+                lines.pop()
+            request_body["prompt"] = "\n".join(lines)
+        if self.max_generation_token:
+            request_body['tokens_to_generate'] = self.max_generation_token
+        else:
+            request_body['tokens_to_generate'] = 512
+        if self.top_p:
+            request_body['top_p'] = self.top_p
+
+        if stream:
+            timeout = TIMEOUT_STREAMING
+            request_body['stream'] = True
+            request_body['use_standard_sse'] = True
+        else:
+            timeout = TIMEOUT_ALL
+        try:
+            response = requests.post(
+                self.url,
+                headers=headers,
+                json=request_body,
+                stream=stream,
+                timeout=timeout,
+            )
+        except:
+            return None
+
+        return response
+
+    def _decode_chat_response(self, response):
+        error_msg = ""
+        for chunk in response.iter_lines():
+            if chunk:
+                chunk = chunk.decode()
+                chunk_length = len(chunk)
+                print(chunk)
+                try:
+                    chunk = json.loads(chunk[6:])
+                except json.JSONDecodeError:
+                    print(i18n("JSON解析错误,��到的内容: ") + f"{chunk}")
+                    error_msg += chunk
+                    continue
+                if chunk_length > 6 and "delta" in chunk["choices"][0]:
+                    if "finish_reason" in chunk["choices"][0] and chunk["choices"][0]["finish_reason"] == "stop":
+                        self.all_token_counts.append(chunk["usage"]["total_tokens"] - sum(self.all_token_counts))
+                        break
+                    try:
+                        yield chunk["choices"][0]["delta"]
+                    except Exception as e:
+                        logging.error(f"Error: {e}")
+                        continue
+        if error_msg:
+            try:
+                error_msg = json.loads(error_msg)
+                if 'base_resp' in error_msg:
+                    status_code = error_msg['base_resp']['status_code']
+                    status_msg = error_msg['base_resp']['status_msg']
+                    raise Exception(f"{status_code} - {status_msg}")
+            except json.JSONDecodeError:
+                pass
+            raise Exception(error_msg)

modules/models/models.py

@@ -603,6 +603,11 @@ def get_model(
         elif model_type == ModelType.YuanAI:
             from .inspurai import Yuan_Client
             model = Yuan_Client(model_name, api_key=access_key, user_name=user_name, system_prompt=system_prompt)
+        elif model_type == ModelType.Minimax:
+            from .minimax import MiniMax_Client
+            if os.environ.get("MINIMAX_API_KEY") != "":
+                access_key = os.environ.get("MINIMAX_API_KEY")
+            model = MiniMax_Client(model_name, api_key=access_key, user_name=user_name, system_prompt=system_prompt)
         elif model_type == ModelType.ChuanhuAgent:
             from .ChuanhuAgent import ChuanhuAgent_Client
             model = ChuanhuAgent_Client(model_name, access_key, user_name=user_name)

modules/presets.py

@@ -72,6 +72,8 @@ ONLINE_MODELS = [
     "yuanai-1.0-translate",
     "yuanai-1.0-dialog",
     "yuanai-1.0-rhythm_poems",
+    "minimax-abab4-chat",
+    "minimax-abab5-chat",
 ]
 
 LOCAL_MODELS = [

readme/README_en.md

@@ -44,6 +44,19 @@
   </p>
 </div>
 
+## Supported LLM Models
+**LLM models via API**:
+- [ChatGPT](https://chat.openai.com) ([GPT-4](https://openai.com/product/gpt-4))
+- [Inspur Yuan 1.0](https://air.inspur.com/home)
+- [MiniMax](https://api.minimax.chat/)
+- [XMChat](https://github.com/MILVLG/xmchat)
+
+**LLM models via local deployment**:
+- [ChatGLM](https://github.com/THUDM/ChatGLM-6B)
+- [LLaMA](https://github.com/facebookresearch/llama)
+- [StableLM](https://github.com/Stability-AI/StableLM)
+- [MOSS](https://github.com/OpenLMLab/MOSS)
+
 ## Usage Tips
 
 - To better control the ChatGPT, use System Prompt.
@@ -87,10 +100,6 @@ When you encounter problems, you should try manually pulling the latest changes
    ```
    pip install -r requirements.txt
    ```
-3. Update Gradio
-   ```
-   pip install gradio --upgrade --force-reinstall
-   ```
 
 Generally, you can solve most problems by following these steps.
 

readme/README_ja.md

@@ -44,6 +44,19 @@
   </p>
 </div>
 
+## サポートされている大規模言語モデル
+**APIを通じてアクセス可能な大規模言語モデル**：
+- [ChatGPT](https://chat.openai.com) ([GPT-4](https://openai.com/product/gpt-4))
+- [Inspur Yuan 1.0](https://air.inspur.com/home)
+- [MiniMax](https://api.minimax.chat/)
+- [XMChat](https://github.com/MILVLG/xmchat)
+
+**ローカルに展開された大規模言語モデル**：
+- [ChatGLM](https://github.com/THUDM/ChatGLM-6B)
+- [LLaMA](https://github.com/facebookresearch/llama)
+- [StableLM](https://github.com/Stability-AI/StableLM)
+- [MOSS](https://github.com/OpenLMLab/MOSS)
+
 ## 使う上でのTips
 
 - ChatGPTをより適切に制御するために、システムプロンプトを使用できます。
@@ -86,10 +99,6 @@ python ChuanhuChatbot.py
    ```
    pip install -r requirements.txt
    ```
-3. Gradioを更新
-   ```
-   pip install gradio --upgrade --force-reinstall
-   ```
 
 一般的に、以下の手順でほとんどの問題を解決することができます。
 

requirements.txt

@@ -8,7 +8,7 @@ tqdm
 colorama
 duckduckgo_search==2.9.5
 Pygments
-langchain==0.0.170
+langchain==0.0.142
 markdown
 PyPDF2
 pdfplumber