@best-linux-code/opencode-copilot-device-auth

GitHub Repository: https://github.com/best-linux-code/opencode-copilot-device-auth Package on npm: https://www.npmjs.com/package/@best-linux-code/opencode-copilot-device-auth

This plugin is an enhanced OpenCode authentication provider for GitHub Copilot. It builds upon the official opencode-copilot-auth plugin, introducing major improvements in model metadata handling, authentication flow reliability, and support for advanced model features like Claude thinking budgets and Gemini reasoning efforts.

Features

• API-Driven Model Metadata: Fetches real-time model context limits directly from the Copilot API. For example, it correctly identifies the 200K context limit for Claude Opus 4.6 instead of falling back to a static 128K limit.
• Dynamic Claude Thinking Budgets: Automatically generates low, medium, and high thinking_budget variants based on API-reported model capabilities.
• Dynamic Gemini Reasoning Efforts: Automatically generates low, medium, and high reasoning_effort variants based on API-reported model capabilities.
• Unified Headers: Implements consistent vscode-chat headers across all model families, including Claude, Gemini, GPT, and Grok.
• Enhanced Token Exchange: Features a two-level fallback mechanism (v2/token to copilot_internal/user) to ensure reliable authentication.
• Robust Device Flow: Implements the OAuth device code flow with proper handling for slow_down responses.
• Automatic Cleanup: Removes stale provider configurations from opencode.json upon logout.
• Debug Logging: Detailed response status and error body logging available at /tmp/copilot-device-auth.log.
• Zero Cost: Provides access to all available Copilot models at no additional cost.
• Future-Proof: Dynamically registers new models like claude-opus-4.6-1m if they are reported by the API.

Model Limits Comparison

This plugin patches live per-model limits from Copilot, ensuring OpenCode uses accurate context window sizes.

| Model | This Plugin (Live) | static models.dev | | --- | --- | --- | | claude-opus-4.6 | 200,000 | 128,000 | | claude-sonnet-4.6 | 200,000 | 128,000 | | claude-haiku-4.5 | 144,000 | 128,000 |

Installation

Install the package via npm:

npm install @best-linux-code/opencode-copilot-device-auth

Add the plugin to your opencode.json configuration:

{
  "plugin": [
    "@best-linux-code/opencode-copilot-device-auth"
  ]
}

Known Limitations

• Input Tokens Undefined: Some models (e.g., gpt-5, claude-opus-41) may display input=undefined because the Copilot API does not currently return max_prompt_tokens for these specific identifiers.
• Grok Initiator Error: The grok-code-fast-1 model may occasionally return an "invalid initiator" error. This is a known issue with the upstream Copilot API and is not a bug within this plugin.
• Duplicate Loader Calls: The internal CopilotAuthPlugin may still run alongside this plugin. This results in duplicate loader calls but is harmless and idempotent.

Publishing

Publishing is handled via GitHub Actions. You can manually trigger the workflow to release a new version (patch, minor, or major).

@best-linux-code/opencode-copilot-device-auth (中文)

GitHub 仓库: https://github.com/best-linux-code/opencode-copilot-device-auth npm 软件包: https://www.npmjs.com/package/@best-linux-code/opencode-copilot-device-auth

本插件是针对 GitHub Copilot 的增强版 OpenCode 身份验证提供程序。它基于官方的 opencode-copilot-auth 插件构建，在模型元数据处理、身份验证流程可靠性以及对 Claude 思考预算 (thinking budget) 和 Gemini 推理强度 (reasoning effort) 等高级功能的支持方面进行了重大改进。

功能特性

• API 驱动的模型元数据：直接从 Copilot API 获取实时的模型上下文限制。例如，它能正确识别 Claude Opus 4.6 的 200K 上下文限制，而不是回退到静态的 128K 限制。
• 动态 Claude 思考预算：根据 API 报告的模型能力，自动生成低、中、高三种思考预算 (thinking_budget) 变体。
• 动态 Gemini 推理强度：根据 API 报告的模型能力，自动生成低、中、高三种推理强度 (reasoning_effort) 变体。
• 统一请求头：在包括 Claude、Gemini、GPT 和 Grok 在内的所有模型系列中实现一致的 vscode-chat 请求头。
• 增强型令牌交换：具备两级回退机制（从 v2/token 回退到 copilot_internal/user），以确保身份验证的可靠性。
• 稳健的设备流：实现 OAuth 设备代码流，并能妥善处理 slow_down 响应。
• 自动清理：注销时自动从 opencode.json 中移除过时的提供程序配置。
• 调试日志：在 /tmp/copilot-device-auth.log 中提供详细的响应状态和错误主体日志。
• 零成本：免费访问所有可用的 Copilot 模型。
• 面向未来：如果 API 报告了如 claude-opus-4.6-1m 等新模型，插件将动态注册这些模型。

模型限制对比

本插件会根据 Copilot 的实时数据修补每个模型的限制，确保 OpenCode 使用准确的上下文窗口大小。

| 模型 | 本插件 (实时) | 静态 models.dev | | --- | --- | --- | | claude-opus-4.6 | 200,000 | 128,000 | | claude-sonnet-4.6 | 200,000 | 128,000 | | claude-haiku-4.5 | 144,000 | 128,000 |

安装

通过 npm 安装软件包：

npm install @best-linux-code/opencode-copilot-device-auth

将插件添加到您的 opencode.json 配置中：

{
  "plugin": [
    "@best-linux-code/opencode-copilot-device-auth"
  ]
}

已知限制

• 输入令牌未定义：某些模型（例如 gpt-5、claude-opus-41）可能会显示 input=undefined，这是因为 Copilot API 目前未针对这些特定标识符返回 max_prompt_tokens。
• Grok 启动器错误：grok-code-fast-1 模型可能会偶尔返回 "invalid initiator" 错误。这是上游 Copilot API 的已知问题，并非本插件的错误。
• 重复加载调用：内置的 CopilotAuthPlugin 可能会与本插件同时运行。这会导致重复的加载调用，但由于操作是幂等的，因此并无大碍。

发布

发布工作由 GitHub Actions 处理。您可以手动触发工作流以发布新版本（修订号 patch、次版本号 minor 或主版本号 major）。