feat: add MiniMax provider support (#2989)

Co-authored-by: octo-patch <octo-patch@users.noreply.github.com>

docs/docs/config-reference/llm/index.mdx

@@ -67,6 +67,11 @@ import { ConfigClassTable } from '@site/src/components/mdx/ConfigClassTable';
     "description": "LlamaServerParameters(name: str, provider: str = 'llama.cpp.server', verbose: Optional[bool] = False, concurrency: Optional[int] = 20, backend: Optional[str] = None, prompt_template: Optional[str] = None, context_length: Optional[int] = None, reasoning_model: Optional[bool] = None, path: Optional[str] = None, model_hf_repo: Optional[str] = None, model_hf_file: Optional[str] = None, device: Optional[str] = None, server_bin_path: Optional[str] = None, server_host: str = '127.0.0.1', server_port: int = 0, temperature: float = 0.8, seed: int = 42, debug: bool = False, model_url: Optional[str] = None, model_draft: Optional[str] = None, threads: Optional[int] = None, n_gpu_layers: Optional[int] = None, batch_size: Optional[int] = None, ubatch_size: Optional[int] = None, ctx_size: Optional[int] = None, grp_attn_n: Optional[int] = None, grp_attn_w: Optional[int] = None, n_predict: Optional[int] = None, slot_save_path: Optional[str] = None, n_slots: Optional[int] = None, cont_batching: bool = False, embedding: bool = False, reranking: bool = False, metrics: bool = False, slots: bool = False, draft: Optional[int] = None, draft_max: Optional[int] = None, draft_min: Optional[int] = None, api_key: Optional[str] = None, lora_files: List[str] = <factory>, no_context_shift: bool = False, no_webui: Optional[bool] = None, startup_timeout: Optional[int] = None)",
     "link": "./llama_cpp_adapter_llamaserverparameters_421f40"
   },
+  {
+    "name": "MiniMaxDeployModelParameters",
+    "description": "MiniMax proxy LLM configuration.",
+    "link": "./minimax_minimaxdeploymodelparameters_a1b2c4"
+  },
   {
     "name": "MoonshotDeployModelParameters",
     "description": "Moonshot proxy LLM configuration.",

docs/docs/config-reference/llm/minimax_minimaxdeploymodelparameters_a1b2c4.mdx

@@ -0,0 +1,97 @@
+---
+title: "MiniMax Proxy LLM Configuration"
+description: "MiniMax proxy LLM configuration."
+---
+
+import { ConfigDetail } from "@site/src/components/mdx/ConfigDetail";
+
+<ConfigDetail config={{
+  "name": "MiniMaxDeployModelParameters",
+  "description": "MiniMax proxy LLM configuration.",
+  "documentationUrl": "https://platform.minimax.io/docs/api-reference/text-openai-api",
+  "parameters": [
+    {
+      "name": "name",
+      "type": "string",
+      "required": true,
+      "description": "The name of the model."
+    },
+    {
+      "name": "backend",
+      "type": "string",
+      "required": false,
+      "description": "The real model name to pass to the provider, default is None. If backend is None, use name as the real model name."
+    },
+    {
+      "name": "provider",
+      "type": "string",
+      "required": false,
+      "description": "The provider of the model. If model is deployed in local, this is the inference type. If model is deployed in third-party service, this is platform name('proxy/<platform>')",
+      "defaultValue": "proxy/minimax"
+    },
+    {
+      "name": "verbose",
+      "type": "boolean",
+      "required": false,
+      "description": "Show verbose output.",
+      "defaultValue": "False"
+    },
+    {
+      "name": "concurrency",
+      "type": "integer",
+      "required": false,
+      "description": "Model concurrency limit",
+      "defaultValue": "100"
+    },
+    {
+      "name": "prompt_template",
+      "type": "string",
+      "required": false,
+      "description": "Prompt template. If None, the prompt template is automatically determined from model. Just for local deployment."
+    },
+    {
+      "name": "context_length",
+      "type": "integer",
+      "required": false,
+      "description": "The context length of the MiniMax API. If None, it is determined by the model."
+    },
+    {
+      "name": "reasoning_model",
+      "type": "boolean",
+      "required": false,
+      "description": "Whether the model is a reasoning model. If None, it is automatically determined from model."
+    },
+    {
+      "name": "api_base",
+      "type": "string",
+      "required": false,
+      "description": "The base url of the MiniMax API.",
+      "defaultValue": "${env:MINIMAX_API_BASE:-https://api.minimax.io/v1}"
+    },
+    {
+      "name": "api_key",
+      "type": "string",
+      "required": false,
+      "description": "The API key of the MiniMax API.",
+      "defaultValue": "${env:MINIMAX_API_KEY}"
+    },
+    {
+      "name": "api_type",
+      "type": "string",
+      "required": false,
+      "description": "The type of the OpenAI API, if you use Azure, it can be: azure"
+    },
+    {
+      "name": "api_version",
+      "type": "string",
+      "required": false,
+      "description": "The version of the OpenAI API."
+    },
+    {
+      "name": "http_proxy",
+      "type": "string",
+      "required": false,
+      "description": "The http or https proxy to use openai"
+    }
+  ]
+}} />

packages/dbgpt-core/src/dbgpt/model/proxy/__init__.py

@@ -11,6 +11,7 @@ if TYPE_CHECKING:
     from dbgpt.model.proxy.llms.gemini import GeminiLLMClient
     from dbgpt.model.proxy.llms.gitee import GiteeLLMClient
     from dbgpt.model.proxy.llms.infiniai import InfiniAILLMClient
+    from dbgpt.model.proxy.llms.minimax import MiniMaxLLMClient
     from dbgpt.model.proxy.llms.moonshot import MoonshotLLMClient
     from dbgpt.model.proxy.llms.ollama import OllamaLLMClient
     from dbgpt.model.proxy.llms.siliconflow import SiliconFlowLLMClient
@@ -39,6 +40,7 @@ def __lazy_import(name):
         "DeepseekLLMClient": "dbgpt.model.proxy.llms.deepseek",
         "GiteeLLMClient": "dbgpt.model.proxy.llms.gitee",
         "InfiniAILLMClient": "dbgpt.model.proxy.llms.infiniai",
+        "MiniMaxLLMClient": "dbgpt.model.proxy.llms.minimax",
     }
 
     if name in module_path:
@@ -69,4 +71,5 @@ __all__ = [
     "DeepseekLLMClient",
     "GiteeLLMClient",
     "InfiniAILLMClient",
+    "MiniMaxLLMClient",
 ]

packages/dbgpt-core/src/dbgpt/model/proxy/llms/minimax.py

@@ -0,0 +1,183 @@
+import os
+from dataclasses import dataclass, field
+from typing import TYPE_CHECKING, Any, Dict, Optional, Type, Union, cast
+
+from dbgpt.core import ModelMetadata, ModelRequest
+from dbgpt.core.awel.flow import (
+    TAGS_ORDER_HIGH,
+    ResourceCategory,
+    auto_register_resource,
+)
+from dbgpt.model.proxy.llms.proxy_model import ProxyModel, parse_model_request
+from dbgpt.util.i18n_utils import _
+
+from ..base import (
+    AsyncGenerateStreamFunction,
+    GenerateStreamFunction,
+    register_proxy_model_adapter,
+)
+from .chatgpt import OpenAICompatibleDeployModelParameters, OpenAILLMClient
+
+if TYPE_CHECKING:
+    from httpx._types import ProxiesTypes
+    from openai import AsyncAzureOpenAI, AsyncOpenAI
+
+    ClientType = Union[AsyncAzureOpenAI, AsyncOpenAI]
+
+_DEFAULT_MODEL = "MiniMax-M2.5"
+
+
+@auto_register_resource(
+    label=_("MiniMax Proxy LLM"),
+    category=ResourceCategory.LLM_CLIENT,
+    tags={"order": TAGS_ORDER_HIGH},
+    description=_("MiniMax proxy LLM configuration."),
+    documentation_url="https://platform.minimax.io/docs/api-reference/text-openai-api",
+    show_in_ui=False,
+)
+@dataclass
+class MiniMaxDeployModelParameters(OpenAICompatibleDeployModelParameters):
+    """Deploy model parameters for MiniMax."""
+
+    provider: str = "proxy/minimax"
+
+    api_base: Optional[str] = field(
+        default="${env:MINIMAX_API_BASE:-https://api.minimax.io/v1}",
+        metadata={
+            "help": _("The base url of the MiniMax API."),
+        },
+    )
+
+    api_key: Optional[str] = field(
+        default="${env:MINIMAX_API_KEY}",
+        metadata={
+            "help": _("The API key of the MiniMax API."),
+            "tags": "privacy",
+        },
+    )
+
+
+async def minimax_generate_stream(
+    model: ProxyModel, tokenizer, params, device, context_len=2048
+):
+    client: MiniMaxLLMClient = cast(MiniMaxLLMClient, model.proxy_llm_client)
+    request = parse_model_request(params, client.default_model, stream=True)
+    async for r in client.generate_stream(request):
+        yield r
+
+
+class MiniMaxLLMClient(OpenAILLMClient):
+    """MiniMax LLM Client.
+
+    MiniMax's API is compatible with OpenAI's API, so we inherit from OpenAILLMClient.
+
+    API Reference: https://platform.minimax.io/docs/api-reference/text-openai-api
+    """
+
+    def __init__(
+        self,
+        api_key: Optional[str] = None,
+        api_base: Optional[str] = None,
+        api_type: Optional[str] = None,
+        api_version: Optional[str] = None,
+        model: Optional[str] = _DEFAULT_MODEL,
+        proxies: Optional["ProxiesTypes"] = None,
+        timeout: Optional[int] = 240,
+        model_alias: Optional[str] = _DEFAULT_MODEL,
+        context_length: Optional[int] = None,
+        openai_client: Optional["ClientType"] = None,
+        openai_kwargs: Optional[Dict[str, Any]] = None,
+        **kwargs,
+    ):
+        api_base = (
+            api_base or os.getenv("MINIMAX_API_BASE") or "https://api.minimax.io/v1"
+        )
+        api_key = api_key or os.getenv("MINIMAX_API_KEY")
+        model = model or _DEFAULT_MODEL
+        if not context_length:
+            context_length = 204800
+
+        if not api_key:
+            raise ValueError(
+                "MiniMax API key is required, please set 'MINIMAX_API_KEY' in "
+                "environment variable or pass it to the client."
+            )
+        super().__init__(
+            api_key=api_key,
+            api_base=api_base,
+            api_type=api_type,
+            api_version=api_version,
+            model=model,
+            proxies=proxies,
+            timeout=timeout,
+            model_alias=model_alias,
+            context_length=context_length,
+            openai_client=openai_client,
+            openai_kwargs=openai_kwargs,
+            **kwargs,
+        )
+
+    def check_sdk_version(self, version: str) -> None:
+        if not version >= "1.0":
+            raise ValueError(
+                "MiniMax API requires openai>=1.0, please upgrade it by "
+                "`pip install --upgrade 'openai>=1.0'`"
+            )
+
+    @property
+    def default_model(self) -> str:
+        model = self._model
+        if not model:
+            model = _DEFAULT_MODEL
+        return model
+
+    def _build_request(
+        self, request: ModelRequest, stream: Optional[bool] = False
+    ) -> Dict[str, Any]:
+        payload = super()._build_request(request, stream)
+        # MiniMax requires temperature in (0.0, 1.0]; 0 is not allowed.
+        temperature = payload.get("temperature")
+        if temperature is not None:
+            if temperature <= 0:
+                payload["temperature"] = 1.0
+            elif temperature > 1.0:
+                payload["temperature"] = 1.0
+        return payload
+
+    @classmethod
+    def param_class(cls) -> Type[MiniMaxDeployModelParameters]:
+        """Get the deploy model parameters class."""
+        return MiniMaxDeployModelParameters
+
+    @classmethod
+    def generate_stream_function(
+        cls,
+    ) -> Optional[Union[GenerateStreamFunction, AsyncGenerateStreamFunction]]:
+        """Get the generate stream function."""
+        return minimax_generate_stream
+
+
+register_proxy_model_adapter(
+    MiniMaxLLMClient,
+    supported_models=[
+        ModelMetadata(
+            model="MiniMax-M2.5",
+            context_length=204800,
+            max_output_length=192000,
+            description=("MiniMax-M2.5 by MiniMax. Peak Performance. Ultimate Value."),
+            link="https://platform.minimax.io/docs/api-reference/text-openai-api",
+            function_calling=True,
+        ),
+        ModelMetadata(
+            model="MiniMax-M2.5-highspeed",
+            context_length=204800,
+            max_output_length=192000,
+            description=(
+                "MiniMax-M2.5-highspeed by MiniMax. Same performance, faster and "
+                "more agile."
+            ),
+            link="https://platform.minimax.io/docs/api-reference/text-openai-api",
+            function_calling=True,
+        ),
+    ],
+)