gpt-luna-5.6 is different in opencode go req fix gpt luna unuse

[wylasdasd]

部分支持。 大部分 Go 模型能用，但 gpt-5.6-luna 这类走 /responses 的模型，主聊天链路还不支持。

已支持
主客户端 GLMClient 只实现了两种协议：

协议 实际请求地址 适用模型
OpenAI
…/chat/completions
GLM、Kimi、DeepSeek、MiMo、Hy3 等
Anthropic
…/v1/messages
MiniMax、Qwen、Claude

core.ts
Ln 70–71

• Supports both OpenAI-compatible (/chat/completions) and
• Anthropic-compatible (/v1/messages) protocols. OpenCode Go 上这两类模型在 Copilot Chat 里都能正常用。

还不支持（Luna 的问题）
gpt-5.6-luna、grok-4.6、muse-spark-1.2-contributor 在 OpenCode Go 文档 里走的是 /v1/responses，不是 /chat/completions。

插件会把 Luna 当成普通 OpenAI 模型，路由到 opencode-go

[wylasdasd]

old

fixed

GPT-5.6 Luna, Grok-4.6, and Muse Spark 1.2 Contributor use /v1/responses in the OpenCode Go documentation, not /chat/completions.

[wylasdasd]

second commit fix bug

fixed

fix(vision): use getApiKey after unified auth removed getApiKeyForEndpoint
Auth was consolidated to a single Go/Zen API key, but automatic vision
still called authManager.getApiKeyForEndpoint(), causing image requests
to fail with "getApiKeyForEndpoint is not a function".(it is I delete go and zen selectitembox caused )