feat(kiro): 支持思考模型 (Thinking Mode) 并通过多配额故障转移增强稳定性 #20
Conversation
…d improve tool calling support
Summary of ChangesHello @Ravens2121, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! 此拉取请求对Kiro（AWS CodeWhisperer/Amazon Q）模块进行了重大升级，旨在增强其智能性和可靠性。它引入了对思考/推理模型的原生支持，允许模型在生成响应时展示其内部思考过程。同时，通过实现多端点故障转移系统，显著提升了服务面对速率限制时的稳定性。此外，还优化了配置灵活性，并对模型注册表进行了规范化，以更好地支持不同的使用场景和模型变体。这些改进在保持完全向后兼容性的同时，为用户提供了更强大、更稳定的AI辅助开发体验。 Highlights
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
| // - Amazon Q endpoint (CLI origin) uses Amazon Q Developer quota | ||
| // - CodeWhisperer endpoint (AI_EDITOR origin) uses Kiro IDE quota | ||
| // Also supports multi-endpoint fallback similar to Antigravity implementation. | ||
| func (e *KiroExecutor) executeWithRetry(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options, accessToken, profileArn string, kiroPayload, body []byte, from, to sdktranslator.Format, reporter *usageReporter, currentOrigin, kiroModelID string, isAgentic, isChatOnly bool) (cliproxyexecutor.Response, error) { |
There was a problem hiding this comment.
executeWithRetry（以及 executeStreamWithRetry）的函数签名包含了 kiroPayload 和 currentOrigin 参数。然而，在函数开头的 for endpointIdx... 循环中，这些参数在被使用之前立即被覆盖了。
kiroPayload 在第 306 行被重新构建，currentOrigin 在第 302 行被重新赋值。这使得从调用方（Execute 和 ExecuteStream）传入的参数变得多余。
为了提高代码清晰度并简化代码，建议进行重构。你可以从 executeWithRetry 和 executeStreamWithRetry 的函数签名中移除 kiroPayload 和 currentOrigin 参数。这些变量可以在这些函数内部声明为局部变量，并且可以移除 Execute 和 ExecuteStream 中现在多余的、初始的 payload 创建逻辑。
| "claude-opus-4-5": "claude-opus-4.5", | ||
| "claude-opus-4.5": "claude-opus-4.5", | ||
| "claude-haiku-4-5": "claude-haiku-4.5", | ||
| "claude-haiku-4.5": "claude-haiku-4.5", | ||
| "claude-sonnet-4-5": "claude-sonnet-4.5", | ||
| "claude-sonnet-4-5-20250929": "claude-sonnet-4.5", | ||
| "claude-sonnet-4.5": "claude-sonnet-4.5", | ||
| "claude-sonnet-4": "claude-sonnet-4", | ||
| "claude-sonnet-4-20250514": "claude-sonnet-4", | ||
| "auto": "auto", |
There was a problem hiding this comment.
在 mapModelToKiro 函数的 modelMap 中，原生模型格式存在几个重复的键。例如，claude-opus-4.5 被列出了两次。
虽然这不会导致运行时错误（map 中一个键的最后一个条目会生效），但这使得代码更难阅读和维护。最好删除这些重复的条目以保持代码整洁。
| "claude-opus-4-5": "claude-opus-4.5", | |
| "claude-opus-4.5": "claude-opus-4.5", | |
| "claude-haiku-4-5": "claude-haiku-4.5", | |
| "claude-haiku-4.5": "claude-haiku-4.5", | |
| "claude-sonnet-4-5": "claude-sonnet-4.5", | |
| "claude-sonnet-4-5-20250929": "claude-sonnet-4.5", | |
| "claude-sonnet-4.5": "claude-sonnet-4.5", | |
| "claude-sonnet-4": "claude-sonnet-4", | |
| "claude-sonnet-4-20250514": "claude-sonnet-4", | |
| "auto": "auto", | |
| "claude-opus-4-5": "claude-opus-4.5", | |
| "claude-haiku-4-5": "claude-haiku-4.5", | |
| "claude-sonnet-4-5": "claude-sonnet-4.5", | |
| "claude-sonnet-4-5-20250929": "claude-sonnet-4.5", | |
| "claude-sonnet-4": "claude-sonnet-4", | |
| "claude-sonnet-4-20250514": "claude-sonnet-4", | |
| "auto": "auto", |
2 participants
PR Title / 拉取请求标题
feat(kiro): Add Thinking Mode support & enhance reliability with multi-quota failoverfeat(kiro): 支持思考模型 (Thinking Mode) 并通过多配额故障转移增强稳定性PR Description / 拉取请求描述
📝 Summary / 摘要
This PR introduces significant upgrades to the Kiro (AWS CodeWhisperer/Amazon Q) module. It adds native support for Thinking/Reasoning models (similar to OpenAI o1/Claude 3.7), implements a robust Multi-Endpoint Failover system to handle rate limits (429), and optimizes configuration flexibility.
本次 PR 对 Kiro (AWS CodeWhisperer/Amazon Q) 模块进行了重大升级。它增加了对 思考/推理模型 (Thinking/Reasoning models) 的原生支持（类似 OpenAI o1/Claude 3.7），实现了一套健壮的 多端点故障转移 (Multi-Endpoint Failover) 系统以应对速率限制 (429)，并优化了配置灵活性。
✨ Key Changes / 主要变更
1. 🧠 Thinking Mode Support / 思考模式支持
reasoning_effortparameter (low/medium/high) to Claude'sbudget_tokens(4k/16k/32k).reasoning_effort参数（low/medium/high）映射为 Claude 的budget_tokens（4k/16k/32k）。<thinking>...</thinking>tags, even across chunk boundaries.<thinking>...</thinking>标签内的内容，即使标签跨越了数据块边界。reasoning_contentfields (for non-stream) orthinking_deltaevents (for stream).reasoning_content字段（非流式）或thinking_delta事件（流式）。2. 🛡️ Robustness & Failover / 稳健性与故障转移
kiroEndpointConfigto distinguish between IDE (CodeWhisperer) and CLI (Amazon Q) quotas.kiroEndpointConfig结构，明确区分 IDE (CodeWhisperer) 和 CLI (Amazon Q) 的配额来源。429 Too Many Requests, the request seamlessly retries on the next available endpoint/quota.429 Too Many Requests，请求将无缝在下一个可用端点/配额上重试。OriginandX-Amz-Targetheaders for each endpoint to prevent403 Forbiddenerrors due to protocol mismatches.Origin和X-Amz-Target头信息，防止因协议不匹配导致的403 Forbidden错误。3. ⚙️ Configuration & Models / 配置与模型
KiroPreferredEndpoint(global) andPreferredEndpoint(per-key) settings to allow users to prioritize specific quotas (e.g., "ide" or "cli").KiroPreferredEndpoint（全局）和PreferredEndpoint（单 Key）设置，允许用户优先选择特定的配额（如 "ide" 或 "cli"）。-agenticvariants optimized for large code generation tasks.-agentic变体。4. 🔧 Fixes / 修复
ErrortoDebugto reduce log noise.Error降级为Debug，减少日志噪音。