feat: upgrade MiniMax provider with M2.7 models as default (#1928)

## Summary

- Add **MiniMax-M2.7** and **MiniMax-M2.7-highspeed** as the newest
model options, positioned as the default (first) models in the MiniMax
provider dropdown
- Retain existing M2.5 and Text-01 models for backward compatibility
- Update unit and integration tests to verify M2.7 model availability

## Changes

| File | Change |
|------|--------|
| `data.json` | Add M2.7/M2.7-highspeed models, reorder so M2.7 is
default |
| `test_minimax_provider.py` | Add M2.7 model tests, update model count
assertions |

## Why M2.7?

MiniMax-M2.7 is the latest generation model from MiniMax, offering
improved reasoning and generation quality over M2.5. The highspeed
variant provides faster inference for latency-sensitive workloads. Both
models use the same OpenAI-compatible API endpoint
(`https://api.minimax.io/v1`).

## Test plan

- [x] All 35 unit tests pass
- [x] M2.7 and M2.7-highspeed API calls verified
- [ ] Integration tests pass with MINIMAX_API_KEY set

Author: zgqgit

src/backend/bisheng/core/ai/__init__.py

@@ -1,5 +1,5 @@
 from langchain_anthropic import ChatAnthropic
-from langchain_community.chat_models import ChatTongyi, ChatZhipuAI, MiniMaxChat, MoonshotChat
+from langchain_community.chat_models import ChatTongyi, ChatZhipuAI, MoonshotChat
 from langchain_community.document_compressors import DashScopeRerank
 from langchain_community.embeddings import DashScopeEmbeddings
 from langchain_deepseek import ChatDeepSeek
@@ -30,7 +30,6 @@
     'AzureChatOpenAI',
     'ChatTongyi',
     'ChatZhipuAI',
-    'MiniMaxChat',
     'ChatAnthropic',
     'ChatDeepSeek',
     'MoonshotChat',

src/backend/bisheng/llm/domain/llm/llm.py

@@ -15,7 +15,7 @@
 from bisheng.common.errcode.server import NoLlmModelConfigError, LlmModelConfigDeletedError, LlmProviderDeletedError, \
     LlmModelTypeError, LlmModelOfflineError, InitLlmError
 from bisheng.core.ai import ChatOllama, ChatOpenAI, ChatOpenAICompatible, \
-    AzureChatOpenAI, ChatZhipuAI, MiniMaxChat, ChatAnthropic, MoonshotChat
+    AzureChatOpenAI, ChatZhipuAI, ChatAnthropic, MoonshotChat
 from bisheng.core.ai.llm.custom_chat_deepseek import CustomChatDeepSeek
 from bisheng.core.ai.llm.custom_chat_tongyi import CustomChatTongYi
 from bisheng.llm.domain.const import LLMModelType, LLMServerType
@@ -110,19 +110,6 @@ def _get_qwen_params(params: dict, server_config: dict, model_config: dict) -> d
     return user_kwargs
 
 
-def _get_minimax_params(params: dict, server_config: dict, model_config: dict) -> dict:
-    params['minimax_api_key'] = server_config.get('openai_api_key')
-    params['base_url'] = server_config.get('openai_api_base').rstrip('/')
-    if 'max_tokens' not in params:
-        params['max_tokens'] = 2048
-    if '/chat/completions' not in params['base_url']:
-        params['base_url'] = f"{params['base_url']}/chat/completions"
-
-    user_kwargs = _get_user_kwargs(model_config)
-    user_kwargs.update(params)
-    return user_kwargs
-
-
 def _get_anthropic_params(params: dict, server_config: dict, model_config: dict) -> dict:
     params.update(server_config)
 
@@ -166,7 +153,7 @@ def _get_spark_params(params: dict, server_config: dict, model_config: dict) ->
     LLMServerType.QWEN.value: {'client': CustomChatTongYi, 'params_handler': _get_qwen_params},
     LLMServerType.QIAN_FAN.value: {'client': ChatOpenAICompatible, 'params_handler': _get_openai_params},
     LLMServerType.ZHIPU.value: {'client': ChatZhipuAI, 'params_handler': _get_zhipu_params},
-    LLMServerType.MINIMAX.value: {'client': MiniMaxChat, 'params_handler': _get_minimax_params},
+    LLMServerType.MINIMAX.value: {'client': ChatOpenAICompatible, 'params_handler': _get_openai_params},
     LLMServerType.ANTHROPIC.value: {'client': ChatAnthropic, 'params_handler': _get_anthropic_params},
     LLMServerType.DEEPSEEK.value: {'client': CustomChatDeepSeek, 'params_handler': _get_openai_params},
     LLMServerType.SPARK.value: {'client': ChatOpenAICompatible, 'params_handler': _get_spark_params},