feat: add MiniMax as native AI provider

.env.example

@@ -4,7 +4,7 @@
 # API Keys - At least one is required
 #
 # IMPORTANT: Choose ONE approach:
-# - Native APIs (Gemini/OpenAI/XAI) for direct access
+# - Native APIs (Gemini/OpenAI/XAI/MiniMax) for direct access
 # - DIAL for unified enterprise access
 # - OpenRouter for unified cloud access
 # Having multiple unified providers creates ambiguity about which serves each model.
@@ -29,6 +29,9 @@ AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
 # Get your X.AI API key from: https://console.x.ai/
 XAI_API_KEY=your_xai_api_key_here
 
+# Get your MiniMax API key from: https://platform.minimaxi.com/
+MINIMAX_API_KEY=your_minimax_api_key_here
+
 # Get your DIAL API key and configure host URL
 # DIAL provides unified access to multiple AI models through a single API
 DIAL_API_KEY=your_dial_api_key_here
@@ -105,6 +108,20 @@ DEFAULT_THINKING_MODE_THINKDEEP=high
 #   - grok3           (shorthand for grok-3)
 #   - grokfast        (shorthand for grok-3-fast)
 #
+#
+# Supported MiniMax models:
+#   - MiniMax-M2.5           (196K context, frontier reasoning)
+#   - MiniMax-M2.5-highspeed (204K context, fast reasoning)
+#   - MiniMax-M2.1           (204K context, advanced reasoning)
+#   - MiniMax-M2.1-highspeed (204K context, fast reasoning)
+#   - MiniMax-M2             (204K context, capable reasoning)
+#   - m2.5                   (shorthand for MiniMax-M2.5)
+#   - m2.5-fast              (shorthand for MiniMax-M2.5-highspeed)
+#   - m2.1                   (shorthand for MiniMax-M2.1)
+#   - m2.1-fast              (shorthand for MiniMax-M2.1-highspeed)
+#   - m2                     (shorthand for MiniMax-M2)
+#   - minimax                (shorthand for MiniMax-M2.5)
+#
 # Supported DIAL models (when available in your DIAL deployment):
 #   - o3-2025-04-16   (200K context, latest O3 release)
 #   - o4-mini-2025-04-16 (200K context, latest O4 mini)
@@ -133,6 +150,8 @@ DEFAULT_THINKING_MODE_THINKDEEP=high
 #   OPENAI_ALLOWED_MODELS=o4-mini                # Single model standardization
 #   GOOGLE_ALLOWED_MODELS=flash,pro              # Allow both Gemini models
 #   XAI_ALLOWED_MODELS=grok,grok-3-fast          # Allow both GROK variants
+#   MINIMAX_ALLOWED_MODELS=MiniMax-M2.5           # Only allow M2.5 model
+#   MINIMAX_ALLOWED_MODELS=m2.5,m2.5-fast         # Allow M2.5 and M2.5 Highspeed
 #   DIAL_ALLOWED_MODELS=o3,o4-mini                       # Only allow O3/O4 models via DIAL
 #   DIAL_ALLOWED_MODELS=opus-4.1,sonnet-4.1                  # Only Claude 4.1 models (without thinking)
 #   DIAL_ALLOWED_MODELS=opus-4.1-thinking,sonnet-4.1-thinking # Only Claude 4.1 with thinking mode
@@ -142,6 +161,7 @@ DEFAULT_THINKING_MODE_THINKDEEP=high
 # OPENAI_ALLOWED_MODELS=
 # GOOGLE_ALLOWED_MODELS=
 # XAI_ALLOWED_MODELS=
+# MINIMAX_ALLOWED_MODELS=
 # DIAL_ALLOWED_MODELS=
 
 # Optional: Custom model configuration file path

conf/minimax_models.json

@@ -0,0 +1,121 @@
+{
+  "_README": {
+    "description": "Model metadata for MiniMax API access.",
+    "documentation": "https://github.com/BeehiveInnovations/pal-mcp-server/blob/main/docs/custom_models.md",
+    "usage": "Models listed here are exposed directly through the MiniMax provider. Aliases are case-insensitive.",
+    "field_notes": "Matches providers/shared/model_capabilities.py.",
+    "field_descriptions": {
+      "model_name": "The model identifier (e.g., 'MiniMax-M2.5', 'MiniMax-M2')",
+      "aliases": "Array of short names users can type instead of the full model name",
+      "context_window": "Total number of tokens the model can process (input + output combined)",
+      "max_output_tokens": "Maximum number of tokens the model can generate in a single response",
+      "supports_extended_thinking": "Whether the model supports extended reasoning tokens (MiniMax uses reasoning_split instead)",
+      "supports_json_mode": "Whether the model can guarantee valid JSON output",
+      "supports_function_calling": "Whether the model supports function/tool calling",
+      "supports_images": "Whether the model can process images/visual input",
+      "supports_temperature": "Whether the model accepts temperature parameter in API calls",
+      "description": "Human-readable description of the model",
+      "intelligence_score": "1-20 human rating used as the primary signal for auto-mode model ordering"
+    }
+  },
+  "models": [
+    {
+      "model_name": "MiniMax-M2.5",
+      "friendly_name": "MiniMax (M2.5)",
+      "aliases": [
+        "m2.5",
+        "minimax",
+        "minimax-m2.5"
+      ],
+      "intelligence_score": 16,
+      "description": "MiniMax M2.5 (196K context) - Frontier reasoning model with tool calling",
+      "context_window": 196608,
+      "max_output_tokens": 196608,
+      "supports_extended_thinking": false,
+      "supports_system_prompts": true,
+      "supports_streaming": true,
+      "supports_function_calling": true,
+      "supports_json_mode": true,
+      "supports_images": false,
+      "supports_temperature": true
+    },
+    {
+      "model_name": "MiniMax-M2.5-highspeed",
+      "friendly_name": "MiniMax (M2.5 Highspeed)",
+      "aliases": [
+        "m2.5-fast",
+        "m2.5-highspeed",
+        "minimax-m2.5-fast"
+      ],
+      "intelligence_score": 15,
+      "description": "MiniMax M2.5 Highspeed (204K context) - Fast reasoning model",
+      "context_window": 204800,
+      "max_output_tokens": 100000,
+      "supports_extended_thinking": false,
+      "supports_system_prompts": true,
+      "supports_streaming": true,
+      "supports_function_calling": true,
+      "supports_json_mode": true,
+      "supports_images": false,
+      "supports_temperature": true
+    },
+    {
+      "model_name": "MiniMax-M2.1",
+      "friendly_name": "MiniMax (M2.1)",
+      "aliases": [
+        "m2.1",
+        "minimax-m2.1"
+      ],
+      "intelligence_score": 15,
+      "description": "MiniMax M2.1 (204K context) - Advanced reasoning model with tool calling",
+      "context_window": 204800,
+      "max_output_tokens": 100000,
+      "supports_extended_thinking": false,
+      "supports_system_prompts": true,
+      "supports_streaming": true,
+      "supports_function_calling": true,
+      "supports_json_mode": true,
+      "supports_images": false,
+      "supports_temperature": true
+    },
+    {
+      "model_name": "MiniMax-M2.1-highspeed",
+      "friendly_name": "MiniMax (M2.1 Highspeed)",
+      "aliases": [
+        "m2.1-fast",
+        "m2.1-highspeed",
+        "minimax-m2.1-fast"
+      ],
+      "intelligence_score": 14,
+      "description": "MiniMax M2.1 Highspeed (204K context) - Fast reasoning model",
+      "context_window": 204800,
+      "max_output_tokens": 100000,
+      "supports_extended_thinking": false,
+      "supports_system_prompts": true,
+      "supports_streaming": true,
+      "supports_function_calling": true,
+      "supports_json_mode": true,
+      "supports_images": false,
+      "supports_temperature": true
+    },
+    {
+      "model_name": "MiniMax-M2",
+      "friendly_name": "MiniMax (M2)",
+      "aliases": [
+        "m2",
+        "minimax-m2"
+      ],
+      "intelligence_score": 13,
+      "description": "MiniMax M2 (204K context) - Capable reasoning model with tool calling",
+      "context_window": 204800,
+      "max_output_tokens": 100000,
+      "supports_extended_thinking": false,
+      "supports_system_prompts": true,
+      "supports_streaming": true,
+      "supports_function_calling": true,
+      "supports_json_mode": true,
+      "supports_images": false,
+      "supports_temperature": true
+    }
+  ]
+}

providers/__init__.py

@@ -3,6 +3,7 @@
 from .azure_openai import AzureOpenAIProvider
 from .base import ModelProvider
 from .gemini import GeminiModelProvider
+from .minimax import MiniMaxModelProvider
 from .openai import OpenAIModelProvider
 from .openai_compatible import OpenAICompatibleProvider
 from .openrouter import OpenRouterProvider
@@ -16,6 +17,7 @@
     "ModelProviderRegistry",
     "AzureOpenAIProvider",
     "GeminiModelProvider",
+    "MiniMaxModelProvider",
     "OpenAIModelProvider",
     "OpenAICompatibleProvider",
     "OpenRouterProvider",

providers/minimax.py

@@ -0,0 +1,86 @@
+"""MiniMax model provider implementation."""
+
+import logging
+from typing import TYPE_CHECKING, ClassVar, Optional
+
+if TYPE_CHECKING:
+    from tools.models import ToolModelCategory
+
+from .openai_compatible import OpenAICompatibleProvider
+from .registries.minimax import MiniMaxModelRegistry
+from .registry_provider_mixin import RegistryBackedProviderMixin
+from .shared import ModelCapabilities, ProviderType
+
+logger = logging.getLogger(__name__)
+
+
+class MiniMaxModelProvider(RegistryBackedProviderMixin, OpenAICompatibleProvider):
+    """Integration for MiniMax models exposed over an OpenAI-style API.
+
+    Publishes capability metadata for the officially supported deployments and
+    maps tool-category preferences to the appropriate MiniMax model.
+    """
+
+    FRIENDLY_NAME = "MiniMax"
+
+    REGISTRY_CLASS = MiniMaxModelRegistry
+    MODEL_CAPABILITIES: ClassVar[dict[str, ModelCapabilities]] = {}
+
+    # Canonical model identifiers used for category routing.
+    PRIMARY_MODEL = "MiniMax-M2.5"
+    FAST_MODEL = "MiniMax-M2.5-highspeed"
+
+    def __init__(self, api_key: str, **kwargs):
+        """Initialize MiniMax provider with API key."""
+        # Set MiniMax base URL
+        kwargs.setdefault("base_url", "https://api.minimax.io/v1")
+        self._ensure_registry()
+        super().__init__(api_key, **kwargs)
+        self._invalidate_capability_cache()
+
+    def get_provider_type(self) -> ProviderType:
+        """Get the provider type."""
+        return ProviderType.MINIMAX
+
+    def get_preferred_model(self, category: "ToolModelCategory", allowed_models: list[str]) -> Optional[str]:
+        """Get MiniMax's preferred model for a given category from allowed models.
+
+        Args:
+            category: The tool category requiring a model
+            allowed_models: Pre-filtered list of models allowed by restrictions
+
+        Returns:
+            Preferred model name or None
+        """
+        from tools.models import ToolModelCategory
+
+        if not allowed_models:
+            return None
+
+        if category == ToolModelCategory.EXTENDED_REASONING:
+            # Prefer M2.5 for advanced reasoning tasks
+            if self.PRIMARY_MODEL in allowed_models:
+                return self.PRIMARY_MODEL
+            if self.FAST_MODEL in allowed_models:
+                return self.FAST_MODEL
+            return allowed_models[0]
+
+        elif category == ToolModelCategory.FAST_RESPONSE:
+            # Prefer M2.5-highspeed for speed
+            if self.FAST_MODEL in allowed_models:
+                return self.FAST_MODEL
+            if self.PRIMARY_MODEL in allowed_models:
+                return self.PRIMARY_MODEL
+            return allowed_models[0]
+
+        else:  # BALANCED or default
+            # Prefer M2.5 for balanced use
+            if self.PRIMARY_MODEL in allowed_models:
+                return self.PRIMARY_MODEL
+            if self.FAST_MODEL in allowed_models:
+                return self.FAST_MODEL
+            return allowed_models[0]
+
+
+# Load registry data at import time
+MiniMaxModelProvider._ensure_registry()

providers/registries/__init__.py

@@ -4,6 +4,7 @@
 from .custom import CustomEndpointModelRegistry
 from .dial import DialModelRegistry
 from .gemini import GeminiModelRegistry
+from .minimax import MiniMaxModelRegistry
 from .openai import OpenAIModelRegistry
 from .openrouter import OpenRouterModelRegistry
 from .xai import XAIModelRegistry
@@ -13,6 +14,7 @@
     "CustomEndpointModelRegistry",
     "DialModelRegistry",
     "GeminiModelRegistry",
+    "MiniMaxModelRegistry",
     "OpenAIModelRegistry",
     "OpenRouterModelRegistry",
     "XAIModelRegistry",

providers/registries/minimax.py

@@ -0,0 +1,19 @@
+"""Registry loader for MiniMax model capabilities."""
+
+from __future__ import annotations
+
+from ..shared import ProviderType
+from .base import CapabilityModelRegistry
+
+
+class MiniMaxModelRegistry(CapabilityModelRegistry):
+    """Capability registry backed by ``conf/minimax_models.json``."""
+
+    def __init__(self, config_path: str | None = None) -> None:
+        super().__init__(
+            env_var_name="MINIMAX_MODELS_CONFIG_PATH",
+            default_filename="minimax_models.json",
+            provider=ProviderType.MINIMAX,
+            friendly_prefix="MiniMax ({model})",
+            config_path=config_path,
+        )

providers/registry.py

@@ -40,6 +40,7 @@ class ModelProviderRegistry:
         ProviderType.OPENAI,  # Direct OpenAI access
         ProviderType.AZURE,  # Azure-hosted OpenAI deployments
         ProviderType.XAI,  # Direct X.AI GROK access
+        ProviderType.MINIMAX,  # Direct MiniMax access
         ProviderType.DIAL,  # DIAL unified API access
         ProviderType.CUSTOM,  # Local/self-hosted models
         ProviderType.OPENROUTER,  # Catch-all for cloud models
@@ -336,6 +337,7 @@ def _get_api_key_for_provider(cls, provider_type: ProviderType) -> Optional[str]
             ProviderType.OPENAI: "OPENAI_API_KEY",
             ProviderType.AZURE: "AZURE_OPENAI_API_KEY",
             ProviderType.XAI: "XAI_API_KEY",
+            ProviderType.MINIMAX: "MINIMAX_API_KEY",
             ProviderType.OPENROUTER: "OPENROUTER_API_KEY",
             ProviderType.CUSTOM: "CUSTOM_API_KEY",  # Can be empty for providers that don't need auth
             ProviderType.DIAL: "DIAL_API_KEY",

providers/shared/provider_type.py

@@ -12,6 +12,7 @@ class ProviderType(Enum):
     OPENAI = "openai"
     AZURE = "azure"
     XAI = "xai"
+    MINIMAX = "minimax"
     OPENROUTER = "openrouter"
     CUSTOM = "custom"
     DIAL = "dial"