feat: add MiniMax as alternative LLM provider

ai.py

@@ -16,10 +16,17 @@ def __init__(self, cfg: Config):
         self._chat_model: GPTModel = cfg.open_ai_chat_model
         self._embedding_model: EmbeddingModel = cfg.open_ai_embedding_model
         self._use_stream = cfg.use_stream
-        self._encoding = tiktoken.encoding_for_model(self._chat_model.name)
+        try:
+            self._encoding = tiktoken.encoding_for_model(self._chat_model.name)
+        except KeyError:
+            # Fallback for non-OpenAI models (e.g. MiniMax)
+            self._encoding = tiktoken.get_encoding('cl100k_base')
         self._language = cfg.language
         self._temperature = cfg.temperature
-        self.client = OpenAI(api_key=cfg.open_ai_key)
+        client_kwargs = dict(api_key=cfg.open_ai_key)
+        if cfg.open_ai_base_url:
+            client_kwargs['base_url'] = cfg.open_ai_base_url
+        self.client = OpenAI(**client_kwargs)
 
     def _chat_stream(self, messages: list[dict], use_stream: bool = None) -> str:
         use_stream = use_stream if use_stream is not None else self._use_stream
@@ -33,9 +40,10 @@ def _chat_stream(self, messages: list[dict], use_stream: bool = None) -> str:
         if use_stream:
             data = ""
             for chunk in response:
-                if chunk.choices[0].delta.get('content', None) is not None:
-                    data += chunk.choices[0].delta.content
-                    print(chunk.choices[0].delta.content, end='')
+                content = getattr(chunk.choices[0].delta, 'content', None)
+                if content is not None:
+                    data += content
+                    print(content, end='')
             print()
             return data.strip()
         else:

config.example.json

@@ -1,5 +1,6 @@
 {
   "open_ai_key": "",
+  "open_ai_base_url": "",
   "temperature": 0.1,
   "language": "English",
   "open_ai_chat_model": "gpt-3.5-turbo",

config.py

@@ -32,6 +32,10 @@ class GPTModel:
 
     GPTModel('gpt-3.5-turbo-16k', 16385, 0.0005, 0.0015),  # gpt-3.5-turbo-16k-0613
     GPTModel('gpt-3.5-turbo-16k-0613', 16385, 0.0005, 0.0015),
+
+    # MiniMax models (OpenAI-compatible API at https://api.minimax.io/v1)
+    GPTModel('MiniMax-M2.7', 1_000_000, 0.004, 0.016),
+    GPTModel('MiniMax-M2.7-highspeed', 1_000_000, 0.001, 0.004),
 ]
 
 
@@ -61,6 +65,7 @@ def __init__(self):
             self.language = self.config.get('language', 'Chinese')
             self.open_ai_key = self.config.get('open_ai_key')
             self.open_ai_proxy = self.config.get('open_ai_proxy')
+            self.open_ai_base_url = self.config.get('open_ai_base_url')
             gpt_model = self.config.get('open_ai_chat_model', 'gpt-3.5-turbo')
             self.open_ai_chat_model = self.get_gpt_model(gpt_model)
             embedding_model = self.config.get('open_ai_embedding_model', 'text-embedding-ada-002')
@@ -71,6 +76,9 @@ def __init__(self):
             if self.temperature < 0 or self.temperature > 1:
                 raise ValueError(
                     'temperature must be between 0 and 1, less is more conservative, more is more creative')
+            # MiniMax requires temperature > 0
+            if self.is_minimax_model() and self.temperature == 0:
+                self.temperature = 0.01
             self.use_stream = self.config.get('use_stream', False)
             self.use_postgres = self.config.get('use_postgres', False)
             if not self.use_postgres:
@@ -98,3 +106,6 @@ def get_embedding_model(self, model: str):
         if model not in name_list:
             raise ValueError('open_ai_embedding_model must be one of ' + ', '.join(name_list))
         return next(m for m in SUPPORTED_EMBEDDING_MODELS if m.name == model)
+
+    def is_minimax_model(self):
+        return self.open_ai_chat_model.name.startswith('MiniMax-')

readme.md

@@ -75,6 +75,25 @@ if you prefer, you can also run this project using docker:
 }
 ```
 
+## Using MiniMax as Chat Model
+
+ChatWeb supports [MiniMax](https://www.minimaxi.com/) models via their OpenAI-compatible API. MiniMax M2.7 offers a 1M token context window.
+
+- Edit `config.json` and set:
+```json
+{
+  "open_ai_key": "your-minimax-api-key",
+  "open_ai_base_url": "https://api.minimax.io/v1",
+  "open_ai_chat_model": "MiniMax-M2.7"
+}
+```
+
+Available MiniMax models:
+- `MiniMax-M2.7` — flagship model with 1M context window
+- `MiniMax-M2.7-highspeed` — faster variant with 1M context window
+
+> **Note:** MiniMax models are used for chat only. Embeddings still use OpenAI's embedding API, so a separate OpenAI API key may be needed for embedding if you want to use MiniMax for chat.
+
 ## Install PostgreSQL (Optional)
 
 - Edit `config.json` and set `use_postgres` to `true`.

readme.zh.md

@@ -73,6 +73,25 @@ python3 main.py
 }
 ```
 
+## 使用MiniMax作为Chat模型
+
+ChatWeb支持通过OpenAI兼容API使用[MiniMax](https://www.minimaxi.com/)模型。MiniMax M2.7拥有100万token上下文窗口。
+
+- 编辑`config.json`，设置：
+```json
+{
+  "open_ai_key": "你的MiniMax API密钥",
+  "open_ai_base_url": "https://api.minimax.io/v1",
+  "open_ai_chat_model": "MiniMax-M2.7"
+}
+```
+
+可用的MiniMax模型：
+- `MiniMax-M2.7` — 旗舰模型，100万token上下文窗口
+- `MiniMax-M2.7-highspeed` — 高速版本，100万token上下文窗口
+
+> **注意：** MiniMax模型仅用于聊天功能。Embedding仍使用OpenAI的Embedding API，如果您使用MiniMax进行聊天，可能需要单独的OpenAI API密钥用于Embedding。
+
 ## 安装postgresql(可选)
 
 - 编辑`config.json`, 设置`use_postgres`为`true`