feat: add MiniMax Token Plan provider with hardcoded model list

[Blueteemo]

背景

本次 PR 解决 issue #7585 中反馈的 MiniMax Token Plan 用户无法正常使用 AstrBot 的问题。

调研发现：

• MiniMax Token Plan API 不提供 /models 接口，无法动态获取模型列表
• 现有的 anthropic_chat_completion provider 依赖 /models 接口，导致 Token Plan 用户无法获取模型列表
• 这是 Token Plan API 本身的限制，与 AstrBot 实现无关

解决方案

新增专门的 minimax_token_plan provider：

• 继承 ProviderAnthropic，复用所有聊天/补全/工具调用逻辑
• api_base 写死为 https://api.minimaxi.com/anthropic
• 通过 custom_headers 注入 Authorization: Bearer <token> header（Token Plan 要求此认证方式）
• get_models() 返回硬编码的模型列表，避免调用 /models 接口
• 模型列表：MiniMax-M2.7、MiniMax-M2.5、MiniMax-M2.1、MiniMax-M2
• 极速版（highspeed）模型需高级套餐，已排除
• 初始化时校验用户配置的模型名称，不在列表内则抛出明确的 ValueError

模型列表来源：MiniMax 官方 Anthropic 兼容 API 文档

后续展望

如果 MiniMax 官方后续支持了 /models 接口，只需将 get_models() 改为动态调用即可，无需修改其他逻辑。

---

Summary

Add a new provider minimax_token_plan for MiniMax Token Plan users, dedicated to MiniMax's Token Plan subscription API.

Problem: When using MiniMax Token Plan API key with the existing anthropic_chat_completion provider, users cannot get model list because the Token Plan API does not expose a /models endpoint. See issue #7585.

Solution:

• Create a new provider minimax_token_plan that inherits from ProviderAnthropic
• api_base hardcoded to https://api.minimaxi.com/anthropic
• Use custom_headers to inject Authorization: Bearer <token> header (required by Token Plan)
• get_models() returns hardcoded model list: MiniMax-M2.7, M2.5, M2.1, M2 (from official docs)
• Highspeed models excluded (require premium tier)
• Model name validation at initialization to catch unsupported models early

Future: If MiniMax adds a /models endpoint in the future, only get_models() needs to be updated.

Closes #7585

---

Summary by Sourcery

Add a dedicated MiniMax Token Plan provider that reuses the Anthropic adapter while handling Token Plan–specific configuration and model discovery.

New Features:

• Introduce a new minimax_token_plan provider adapter for MiniMax Token Plan subscriptions with a fixed API base and auth header requirements.

Enhancements:

• Provide a hardcoded MiniMax Token Plan model list in the provider to support model selection without relying on a /models endpoint.

[sourcery-ai]

Hey - I've found 2 issues, and left some high level feedback:

• Since api_base is intentionally fixed in __init__, consider omitting it from default_config_tmpl or clearly treating it as read-only to avoid confusing users who might think they can override it.
• Instead of mutating the incoming provider_config dict in-place when setting api_base and auth_header, consider copying it first (e.g., provider_config = {**provider_config, "api_base": ..., "auth_header": True}) to avoid unexpected side effects on shared configuration objects.

Prompt for AI Agents

Please address the comments from this code review:

## Overall Comments
- Since `api_base` is intentionally fixed in `__init__`, consider omitting it from `default_config_tmpl` or clearly treating it as read-only to avoid confusing users who might think they can override it.
- Instead of mutating the incoming `provider_config` dict in-place when setting `api_base` and `auth_header`, consider copying it first (e.g., `provider_config = {**provider_config, "api_base": ..., "auth_header": True}`) to avoid unexpected side effects on shared configuration objects.

## Individual Comments

### Comment 1
<location path="astrbot/core/provider/sources/minimax_token_plan_source.py" line_range="13-21" />
<code_context>
+@register_provider_adapter(
+    "minimax_token_plan",
+    "MiniMax Token Plan 提供商适配器",
+    default_config_tmpl={
+        "key": "",
+        "api_base": "https://api.minimaxi.com/anthropic",
+    },
+    provider_display_name="MiniMax Token Plan",
</code_context>
<issue_to_address>
**suggestion:** Configuration template exposes `api_base` even though it is always overridden.

Because `__init__` always overwrites `provider_config["api_base"]` with the MiniMax endpoint, exposing `api_base` here implies it’s user-configurable when it isn’t. Please either remove it from `default_config_tmpl` or mark it clearly as internal-only to avoid a misleading no-op config option.

```suggestion
@register_provider_adapter(
    "minimax_token_plan",
    "MiniMax Token Plan 提供商适配器",
    default_config_tmpl={
        "key": "",
    },
    provider_display_name="MiniMax Token Plan",
)
```
</issue_to_address>

### Comment 2
<location path="astrbot/core/provider/sources/minimax_token_plan_source.py" line_range="39-44" />
<code_context>
+            provider_settings,
+        )
+
+        self.set_model(provider_config.get("model", "MiniMax-M2.7"))
+
+    async def get_models(self) -> list[str]:
</code_context>
<issue_to_address>
**suggestion (bug_risk):** Model selection is not validated against the known Token Plan model list.

Given `MINIMAX_TOKEN_PLAN_MODELS` is the source of truth for supported models, consider validating `provider_config["model"]` against this list before calling `set_model`. Either normalize to a supported value or raise a clear error for unsupported models to avoid confusing downstream failures and to keep behavior consistent with `get_models()`.

```suggestion
        super().__init__(
            provider_config,
            provider_settings,
        )

        # Validate configured model against the supported Token Plan model list
        configured_model = provider_config.get("model", "MiniMax-M2.7")
        if configured_model not in MINIMAX_TOKEN_PLAN_MODELS:
            raise ValueError(
                f"Unsupported MiniMax Token Plan model: {configured_model!r}. "
                f"Supported models: {', '.join(MINIMAX_TOKEN_PLAN_MODELS)}"
            )

        self.set_model(configured_model)
```
</issue_to_address>

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}

[gemini-code-assist]

Code Review

This pull request introduces the MiniMax Token Plan provider adapter, extending the Anthropic provider with a fixed API base and a hardcoded model list. A review comment points out that the auth_header configuration is likely ineffective and suggests using custom_headers to correctly provide the Authorization token.

[y524331729]

我是购买了 MiniMax Plus-极速版 的用户，当前硬编码的限制导致我无法配置 MiniMax-M2.7-highspeed 并抛出 ValueError。建议将白名单校验改为 Warning 警告，或者把 highspeed 模型加回列表，不要做强行阻断。

[Blueteemo]

我是购买了 MiniMax Plus-极速版 的用户，当前硬编码的限制导致我无法配置 MiniMax-M2.7-highspeed 并抛出 ValueError。建议将白名单校验改为 Warning 警告，或者把 highspeed 模型加回列表，不要做强行阻断。

感谢反馈！新 PR #7692 已修复此问题：

• 将硬编码模型列表扩展为完整列表（含所有 highspeed 变体）
• ValueError 改为 logger.warning，不再强行阻断

在新 PR 合并之前，可参考 issue #7585 的临时方案先做配置。