feat(provider): add custom body overrides

[ayu-exorcist]

Related issue

Related to #1674. This PR delivers the provider-scoped part of that request; model-scoped overrides remain a follow-up design decision.

Problem / 背景

English

Provider APIs and compatible gateways expose request fields that Kimi Code does not generate itself. For example, OpenAI Priority processing uses service_tier: "priority". Without a supported configuration surface, users need a reverse proxy or a local source patch to use those capabilities.

中文

供应商 API 与兼容网关提供了 Kimi Code 不会自行生成的请求字段。例如，OpenAI Priority 处理需要 service_tier: "priority"。缺少受支持的配置入口时，用户只能借助反向代理或修改本地源码来使用这些能力。

What changed / 改动内容

English

• Add the provider-level custom_body TOML field alongside custom_headers.
• Validate it as a recursive JSON object, preserve nested values through TOML and configuration PATCH round-trips, and keep user-native keys such as service_tier and cache_control unchanged.
• Apply the patch immediately before transport dispatch for Kimi, Anthropic, OpenAI Chat Completions, OpenAI Responses, Google GenAI, and Vertex AI. Object values merge recursively; arrays and scalar values replace generated values, including false, 0, empty strings, and SDK/RPC null.
• Treat a boolean custom_body.stream as a transport-control override. It selects matching request serialization and response decoding for the OpenAI-compatible and Anthropic providers; for Google GenAI and Vertex AI, it selects generateContent or generateContentStream and is not forwarded to the SDK request object.
• Reject __proto__, prototype, and constructor at every nesting level, and use safe own-property writes while merging to prevent prototype pollution.
• Add the klient provider contract and conformance coverage so customBody round-trips through supported client transports.
• Update English and Chinese configuration documentation.

中文

• 新增 Provider 级 custom_body TOML 字段，并与 custom_headers 配套使用。
• 将其校验为递归 JSON 对象，在 TOML 与配置 PATCH 往返中保留嵌套值，并保持 service_tier、cache_control 等供应商原生键不变。
• 在 Kimi、Anthropic、OpenAI Chat Completions、OpenAI Responses、Google GenAI 和 Vertex AI 的传输请求发出前应用补丁。对象递归合并；数组和标量值覆盖生成值，包含 false、0、空字符串以及 SDK/RPC 中的 null。
• 将布尔值 custom_body.stream 视为传输控制覆盖。它会让 OpenAI 兼容供应商与 Anthropic 选择匹配的请求序列化和响应解码；对 Google GenAI 与 Vertex AI，则选择 generateContent 或 generateContentStream，且不会透传到 SDK 请求对象。
• 在所有嵌套层级拒绝 __proto__、prototype 和 constructor，并在合并时安全写入自有属性，防止原型污染。
• 新增 klient provider contract 与 conformance 覆盖，使 customBody 可以在受支持的客户端传输中往返。
• 更新中英文配置文档。

Scope / 范围

English

This implementation intentionally uses provider scope: fields apply to every request sent through the configured provider. It does not add model-level custom_body configuration. That work needs an explicit precedence model between provider defaults, model overrides, and generated request fields, and remains tracked by #1674.

中文

本实现有意采用 Provider 级范围：字段会作用于该 Provider 发出的所有请求。它不新增模型级 custom_body 配置。模型级能力需要先明确 Provider 默认值、模型覆盖和生成请求字段之间的优先级，因此仍由 #1674 跟踪。

Validation / 验证

• pnpm exec vitest run …: 120 focused tests passed after rebasing onto the latest upstream/main.
• @moonshot-ai/kosong, @moonshot-ai/agent-core-v2, and @moonshot-ai/klient typecheck passed.
• @moonshot-ai/agent-core-v2 lint:domain, documentation build, and git diff --check passed.
• The shared klient conformance addition passes locally for memory and HTTP. IPC conformance still needs a Unix-socket-enabled runner because the local Windows environment rejects socket binding.

Checklist

• I have read the CONTRIBUTING document.
• I have linked a related issue.
• I have added tests that prove the feature works.
• I added a minor changeset for @moonshot-ai/kimi-code.
• I updated the English and Chinese user-facing configuration documentation.

[changeset-bot]

🦋 Changeset detected

Latest commit: 393972d

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@moonshot-ai/kimi-code	Minor

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 356aadf4b5

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".