feat: add MiniMax provider support

README.md

@@ -470,6 +470,10 @@ api_providers:
     api_key: your_deepseek_key
     base_url: https://api.deepseek.com/v1
     enable_proxy: false
+  minimax:
+    api_key: your_minimax_key
+    base_url: https://api.minimax.io/v1
+    enable_proxy: false
   dashscope:
     api_key: your_dashscope_key
     base_url: https://dashscope.aliyuncs.com/api/v1
@@ -509,6 +513,7 @@ generation:
 |    **OpenAI**    | `api_providers.openai.api_key` / `base_url`                    | GPT 文本、视觉模型和 OpenAI 图像模型 |
 |    **Gemini**    | `api_providers.gemini.api_key` / `base_url`                    | Gemini 文本、视觉模型                |
 |   **DeepSeek**   | `api_providers.deepseek.api_key` / `base_url`                  | DeepSeek 文本模型                    |
+| **MiniMax** | `api_providers.minimax.api_key` / `base_url`                  | MiniMax 文本模型                     |
 |  **DashScope**  | `api_providers.dashscope.api_key` / `base_url`                 | 通义千问、通义万相、Wan 图像/视频等  |
 | **火山方舟 ARK** | `api_providers.ark.api_key` / `base_url`                       | Seedream 图像、Seedance 视频等       |
 |    **Kling**    | `api_providers.kling.access_key` / `secret_key` / `base_url` | 可灵视频生成                         |
@@ -519,7 +524,7 @@ generation:
 
 |        类型        | 模型                                                                                                                                                                        |
 | :----------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-|   **LLM**   | qwen3.6-max-preview, qwen3-max, deepseek-chat, deepseek-reasoner, deepseek-v4-flash, deepseek-v4-pro, gpt-4o, gpt-5, gpt-5.4, gemini-2.5-flash, gemini-2.0-flash, kimi-k2.6 |
+|   **LLM**   | qwen3.6-max-preview, qwen3-max, deepseek-chat, deepseek-reasoner, deepseek-v4-flash, deepseek-v4-pro, MiniMax-M3, MiniMax-M2.7, gpt-4o, gpt-5, gpt-5.4, gemini-2.5-flash, gemini-2.0-flash, kimi-k2.6 |
 |   **VLM**   | qwen3.6-plus, qwen3.6-flash, kimi-k2.6, gpt-5.4, gemini-2.5-flash-image, gemini-2.0-flash                                                                                   |
 |  **文生图**  | wan2.7-image, wan2.7-image-pro, wan2.6-t2i, doubao-seedream-5.0/4.5/4.0, gpt-image-2                                                                                        |
 |  **图生图**  | wan2.7-image, wan2.7-image-pro, doubao-seedream-5.0/4.5/4.0, gpt-image-2                                                                                                    |

README_EN.md

@@ -456,6 +456,10 @@ api_providers:
     api_key: your_deepseek_key
     base_url: https://api.deepseek.com/v1
     enable_proxy: false
+  minimax:
+    api_key: your_minimax_key
+    base_url: https://api.minimax.io/v1
+    enable_proxy: false
   dashscope:
     api_key: your_dashscope_key
     base_url: https://dashscope.aliyuncs.com/api/v1
@@ -488,6 +492,7 @@ models:
 | **OpenAI** | `api_providers.openai.api_key` / `base_url` | GPT text/vision models and OpenAI image models |
 | **Gemini** | `api_providers.gemini.api_key` / `base_url` | Gemini text and vision models |
 | **DeepSeek** | `api_providers.deepseek.api_key` / `base_url` | DeepSeek text models |
+| **MiniMax** | `api_providers.minimax.api_key` / `base_url` | MiniMax text models |
 | **DashScope** | `api_providers.dashscope.api_key` / `base_url` | Qwen, Wan image/video models, and related Alibaba Cloud services |
 | **Volcengine ARK** | `api_providers.ark.api_key` / `base_url` | Seedream image models and Seedance video models |
 | **Kling** | `api_providers.kling.access_key` / `secret_key` / `base_url` | Kling video generation |
@@ -498,7 +503,7 @@ You only need to fill in the provider keys required by the models you choose. Fo
 
 | Type | Models |
 |:---:|:---|
-| **LLM** | qwen3.6-max-preview, qwen3-max, deepseek-chat, deepseek-reasoner, deepseek-v4-flash, deepseek-v4-pro, gpt-4o, gpt-5, gpt-5.4, gemini-2.5-flash, gemini-2.0-flash, kimi-k2.6 |
+| **LLM** | qwen3.6-max-preview, qwen3-max, deepseek-chat, deepseek-reasoner, deepseek-v4-flash, deepseek-v4-pro, MiniMax-M3, MiniMax-M2.7, gpt-4o, gpt-5, gpt-5.4, gemini-2.5-flash, gemini-2.0-flash, kimi-k2.6 |
 | **VLM** | qwen3.6-plus, qwen3.6-flash, kimi-k2.6, gpt-5.4, gemini-2.5-flash-image, gemini-2.0-flash |
 | **Text-to-Image** | wan2.7-image, wan2.7-image-pro, wan2.6-t2i, doubao-seedream-5.0/4.5/4.0, gpt-image-2 |
 | **Image-to-Image** | wan2.7-image, wan2.7-image-pro, doubao-seedream-5.0/4.5/4.0, gpt-image-2 |

video-claw/SKILL.md

@@ -162,6 +162,7 @@ sed -n '1,120p' video-claw/backend/config.yaml
 | 平台 | config.yaml 字段 | 获取链接 |
 |------|--------------|----------|
 | DeepSeek | `api_providers.deepseek.api_key` | https://platform.deepseek.com/api_keys |
+| MiniMax | `api_providers.minimax.api_key` | https://platform.minimax.io |
 | 阿里云 DashScope | `api_providers.dashscope.api_key` 或 `models.llm.api_key` | https://bailian.console.aliyun.com/cn-beijing/?tab=home#/home |
 | 字节火山方舟 | `api_providers.ark.api_key` | https://www.volcengine.com/product/ark |
 | 快手可灵 Kling | `api_providers.kling.access_key` / `api_providers.kling.secret_key` | https://klingai.com/cn/dev |

video-claw/video-claw/backend/config.py

@@ -40,6 +40,11 @@
             "base_url": "https://api.deepseek.com/v1",
             "enable_proxy": False,
         },
+        "minimax": {
+            "api_key": "",
+            "base_url": "https://api.minimax.io/v1",
+            "enable_proxy": False,
+        },
         "dashscope": {
             "api_key": "",
             "base_url": "https://dashscope.aliyuncs.com/api/v1",
@@ -220,6 +225,9 @@ class Config:
     DEEPSEEK_API_KEY = _get(CONFIG, "api_providers.deepseek.api_key")
     DEEPSEEK_BASE_URL = _get(CONFIG, "api_providers.deepseek.base_url")
     DEEPSEEK_ENABLE_PROXY = _get(CONFIG, "api_providers.deepseek.enable_proxy")
+    MINIMAX_API_KEY = _get(CONFIG, "api_providers.minimax.api_key")
+    MINIMAX_BASE_URL = _get(CONFIG, "api_providers.minimax.base_url")
+    MINIMAX_ENABLE_PROXY = _get(CONFIG, "api_providers.minimax.enable_proxy")
     DASHSCOPE_API_KEY = _get(CONFIG, "api_providers.dashscope.api_key")
     DASHSCOPE_BASE_URL = _get(CONFIG, "api_providers.dashscope.base_url")
     DASHSCOPE_ENABLE_PROXY = _get(CONFIG, "api_providers.dashscope.enable_proxy")
@@ -294,6 +302,9 @@ def update_config(cls, values: Dict[str, Any]) -> Dict[str, Any]:
         cls.DEEPSEEK_API_KEY = _get(clean, "api_providers.deepseek.api_key")
         cls.DEEPSEEK_BASE_URL = _get(clean, "api_providers.deepseek.base_url")
         cls.DEEPSEEK_ENABLE_PROXY = _get(clean, "api_providers.deepseek.enable_proxy")
+        cls.MINIMAX_API_KEY = _get(clean, "api_providers.minimax.api_key")
+        cls.MINIMAX_BASE_URL = _get(clean, "api_providers.minimax.base_url")
+        cls.MINIMAX_ENABLE_PROXY = _get(clean, "api_providers.minimax.enable_proxy")
         cls.DASHSCOPE_API_KEY = _get(clean, "api_providers.dashscope.api_key")
         cls.DASHSCOPE_BASE_URL = _get(clean, "api_providers.dashscope.base_url")
         cls.DASHSCOPE_ENABLE_PROXY = _get(clean, "api_providers.dashscope.enable_proxy")

video-claw/video-claw/backend/config.yaml.example

@@ -20,6 +20,10 @@ api_providers:
     api_key: ''
     base_url: https://api.deepseek.com/v1
     enable_proxy: false
+  minimax:
+    api_key: ''
+    base_url: https://api.minimax.io/v1
+    enable_proxy: false
   dashscope:
     api_key: ''
     base_url: https://dashscope.aliyuncs.com/api/v1

video-claw/video-claw/backend/models/config_model.py

@@ -39,6 +39,20 @@
                                 'concurrency': 10,
                                 'price_per_1k_input_token': 0.002,
                                 'price_per_1k_output_token': 0.008},
+            'MiniMax-M3': {'name': 'MiniMax M3',
+                           'provider': 'minimax',
+                           'family': 'minimax',
+                           'type': ['llm'],
+                           'concurrency': 10,
+                           'price_per_1k_input_token': 0.0003,
+                           'price_per_1k_output_token': 0.0012},
+            'MiniMax-M2.7': {'name': 'MiniMax M2.7',
+                             'provider': 'minimax',
+                             'family': 'minimax',
+                             'type': ['llm'],
+                             'concurrency': 10,
+                             'price_per_1k_input_token': 0.0003,
+                             'price_per_1k_output_token': 0.0012},
             'gpt-4o': {'name': 'GPT-4o',
                        'provider': 'openai',
                        'type': ['llm'],

video-claw/video-claw/backend/models/llm_client.py

@@ -12,12 +12,14 @@
     from models.llm_gpt import GPT
     from models.llm_gemini import Gemini
     from models.llm_deepseek import DeepSeek
+    from models.llm_minimax import MiniMax
     from models.llm_dashscope import QwenLLM
     from models.vlm_dashscope import QwenVLClient
 except ImportError:
     from llm_gpt import GPT
     from llm_gemini import Gemini
     from llm_deepseek import DeepSeek
+    from llm_minimax import MiniMax
     from llm_dashscope import QwenLLM
     from vlm_dashscope import QwenVLClient
 
@@ -26,18 +28,21 @@
 logger = logging.getLogger(__name__)
 
 class LLM:
-    def __init__(self, gemini_base_url="", gemini_api_key="", gpt_base_url="", gpt_api_key="", deepseek_base_url="", deepseek_api_key="", dashscope_api_key=""):
+    def __init__(self, gemini_base_url="", gemini_api_key="", gpt_base_url="", gpt_api_key="", deepseek_base_url="", deepseek_api_key="", minimax_base_url="", minimax_api_key="", dashscope_api_key=""):
         self._gemini_base_url = gemini_base_url or Config.GOOGLE_GEMINI_BASE_URL
         self._gemini_api_key = gemini_api_key or Config.GEMINI_API_KEY
         self._gpt_base_url = gpt_base_url or Config.OPENAI_BASE_URL
         self._gpt_api_key = gpt_api_key or Config.OPENAI_API_KEY
         self._deepseek_base_url = deepseek_base_url or Config.DEEPSEEK_BASE_URL
         self._deepseek_api_key = deepseek_api_key or Config.DEEPSEEK_API_KEY
+        self._minimax_base_url = minimax_base_url or Config.MINIMAX_BASE_URL
+        self._minimax_api_key = minimax_api_key or Config.MINIMAX_API_KEY
         self._dashscope_api_key = dashscope_api_key or Config.DASHSCOPE_API_KEY
 
         self._gemini_client = None
         self._gpt_client = None
         self._deepseek_client = None
+        self._minimax_client = None
         self._dashscope_client = None
         self._dashscope_vl_client = None
 
@@ -69,6 +74,15 @@ def deepseek_client(self):
             )
         return self._deepseek_client
 
+    @property
+    def minimax_client(self):
+        if self._minimax_client is None:
+            self._minimax_client = MiniMax(
+                base_url=self._minimax_base_url,
+                api_key=self._minimax_api_key,
+            )
+        return self._minimax_client
+
     @property
     def dashscope_client(self):
         if self._dashscope_client is None:
@@ -128,6 +142,9 @@ def query(self, prompt, image_urls=[], model="qwen3.6-max-preview", safe_content
         elif "gpt" in model_lower:
             # OpenAI series models
             result = self.gpt_client.query(prompt, image_urls=image_urls, model=model, web_search=web_search)
+        elif model_lower.startswith("minimax"):
+            # MiniMax series models (OpenAI-compatible interface)
+            result = self.minimax_client.query(prompt, image_urls=image_urls, model=model, web_search=web_search)
         elif "kimi" in model_lower or "qwen3.6-plus" in model_lower or "qwen3.6-flash" in model_lower or "vl" in model_lower:
             # DashScope VLM models (using MultiModalConversation API)
             result = self.dashscope_vl_client.chat(text=prompt, images=image_urls, model=model, stream=False)

video-claw/video-claw/backend/models/llm_minimax.py

@@ -0,0 +1,134 @@
+import os
+import sys
+
+models_dir = os.path.dirname(os.path.abspath(__file__))
+backend_dir = os.path.dirname(models_dir)
+if backend_dir not in sys.path:
+    sys.path.insert(0, backend_dir)
+
+import time
+import logging
+from openai import OpenAI
+from config import Config
+
+logger = logging.getLogger(__name__)
+
+DEFAULT_MINIMAX_BASE_URL = "https://api.minimax.io/v1"
+DEFAULT_MINIMAX_TIMEOUT = 300
+DEFAULT_MINIMAX_MAX_ATTEMPTS = 5
+DEFAULT_MINIMAX_MAX_TOKENS = 8000
+
+
+class MiniMax:
+    """
+    MiniMax LLM 客户端（OpenAI 兼容接口）
+
+    MiniMax-M3: 最新旗舰模型，1M 上下文，最大输出 128K，支持图片输入
+    MiniMax-M2.7: 上一代模型，作为兼容选项保留
+
+    注意：
+    - Base URL 使用海外版 https://api.minimax.io/v1
+    - temperature 取值范围为 (0.0, 1.0]，默认 1.0，不能传 0
+    """
+    def __init__(self, base_url="", api_key="", timeout=None, max_attempts=None):
+        import httpx
+        self.base_url = base_url or Config.MINIMAX_BASE_URL or DEFAULT_MINIMAX_BASE_URL
+        self.api_key = api_key or Config.MINIMAX_API_KEY
+        self.timeout = self._as_int(timeout, DEFAULT_MINIMAX_TIMEOUT)
+        self.max_attempts = self._as_int(max_attempts, DEFAULT_MINIMAX_MAX_ATTEMPTS)
+
+        if not self.api_key:
+            logger.warning("MINIMAX_API_KEY is not set")
+
+        kwargs = {"api_key": self.api_key, "base_url": self.base_url, "timeout": self.timeout}
+        proxy = Config.provider_proxy("minimax")
+        if proxy:
+            kwargs["http_client"] = httpx.Client(proxy=proxy, timeout=self.timeout)
+        self.client = OpenAI(**kwargs)
+        self.max_tokens = DEFAULT_MINIMAX_MAX_TOKENS
+
+    @staticmethod
+    def _as_int(value, default: int) -> int:
+        try:
+            return int(value)
+        except (TypeError, ValueError):
+            return default
+
+    def query(self, prompt, image_urls=[], model="MiniMax-M3", web_search=False):
+        """
+        Query MiniMax model via the OpenAI-compatible Chat Completions API.
+
+        :param prompt: Text prompt
+        :param image_urls: Optional image URLs for multimodal input
+        :param model: Model name (e.g., MiniMax-M3, MiniMax-M2.7)
+        :param web_search: Unused for MiniMax, kept for interface consistency
+        """
+        if not model:
+            model = "MiniMax-M3"
+
+        messages = [{"role": "system", "content": "You are a helpful assistant."}]
+        content = [{"type": "text", "text": prompt}]
+        if image_urls:
+            content.extend([{"type": "image_url", "image_url": {"url": url}} for url in image_urls])
+        messages.append({"role": "user", "content": content})
+
+        attempts = 0
+        while attempts < self.max_attempts:
+            try:
+                response = self.client.chat.completions.create(
+                    model=model,
+                    messages=messages,
+                    max_tokens=self.max_tokens,
+                    temperature=1.0,  # MiniMax requires temperature in (0.0, 1.0]
+                )
+
+                if response.choices and response.choices[0].message.content:
+                    return response.choices[0].message.content
+                logger.warning("MiniMax returned an empty response; retrying")
+                time.sleep(2)
+            except Exception as e:
+                logger.warning(
+                    "MiniMax request failed; retrying. model=%s attempt=%s/%s timeout=%ss error=%s",
+                    model,
+                    attempts + 1,
+                    self.max_attempts,
+                    self.timeout,
+                    e,
+                )
+                time.sleep(5)
+            attempts += 1
+
+        raise Exception("Max attempts reached, failed to get a response from MiniMax.")
+
+
+if __name__ == "__main__":
+    import sys
+    sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+    from config import Config
+
+    # 支持的模型列表
+    MODELS = ["MiniMax-M3", "MiniMax-M2.7"]
+
+    print("=== MiniMax 文本生成可用性测试 ===")
+    api_key = Config.MINIMAX_API_KEY
+    base_url = Config.MINIMAX_BASE_URL
+    if not api_key:
+        print("✗ MINIMAX_API_KEY 未设置，跳过")
+        sys.exit(1)
+    print(f"  API Key: {api_key[:6]}***{api_key[-4:]}")
+    if base_url:
+        print(f"  Base URL: {base_url}")
+
+    client = MiniMax(api_key=api_key, base_url=base_url)
+    prompt = "用一句话介绍你自己。"
+    print(f"  Prompt: {prompt}")
+
+    for model in MODELS:
+        print(f"\n--- 测试模型: {model} ---")
+        t0 = time.time()
+        try:
+            resp = client.query(prompt, model=model)
+            elapsed = time.time() - t0
+            print(f"✓ 响应 ({elapsed:.1f}s): {resp.strip()[:200]}")
+        except Exception as e:
+            print(f"✗ 失败: {e}")

video-claw/video-claw/backend/tests/test_llm_minimax.py

@@ -0,0 +1,59 @@
+"""Unit tests for the MiniMax LLM provider integration."""
+
+import os
+import sys
+
+BACKEND_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+if BACKEND_DIR not in sys.path:
+    sys.path.insert(0, BACKEND_DIR)
+
+
+def test_minimax_provider_in_config_defaults():
+    """MiniMax should be registered as an api_provider with the overseas base URL."""
+    from config import Config
+
+    config = Config.as_dict()
+    providers = config["api_providers"]
+    assert "minimax" in providers
+    assert providers["minimax"]["base_url"] == "https://api.minimax.io/v1"
+    assert "api_key" in providers["minimax"]
+    assert providers["minimax"]["enable_proxy"] is False
+
+
+def test_minimax_models_registered():
+    """MiniMax-M3 (new default) and MiniMax-M2.7 (compat) should be llm models."""
+    from models.config_model import get_model_config, get_models_by_type
+
+    llm_ids = {model["id"] for model in get_models_by_type("llm")}
+    assert "MiniMax-M3" in llm_ids
+    assert "MiniMax-M2.7" in llm_ids
+
+    m3 = get_model_config("MiniMax-M3")
+    assert m3["provider"] == "minimax"
+    assert "llm" in m3["type"]
+
+
+def test_minimax_client_default_base_url():
+    """Client should fall back to the overseas base URL and default model M3."""
+    from models.llm_minimax import DEFAULT_MINIMAX_BASE_URL, MiniMax
+
+    client = MiniMax(api_key="test-key", base_url="")
+    assert client.base_url == DEFAULT_MINIMAX_BASE_URL
+    assert client.base_url.endswith("/v1")
+
+
+def test_minimax_client_custom_base_url():
+    from models.llm_minimax import MiniMax
+
+    client = MiniMax(api_key="test-key", base_url="https://example.com/v1")
+    assert client.base_url == "https://example.com/v1"
+
+
+def test_minimax_client_uses_openai_sdk():
+    """The provider should be implemented on top of the OpenAI-compatible SDK."""
+    from models.llm_minimax import MiniMax
+
+    client = MiniMax(api_key="test-key")
+    # OpenAI client exposes chat.completions.create
+    assert hasattr(client.client, "chat")
+    assert hasattr(client.client.chat, "completions")