Model name shown in UI is different from the actual model loaded when custom agent enabled
- 主要语言
- Shell
- 星标
- 11.2k
- 派生
- 1.9k
- 平均合并
- 14 小时 16 分钟
- 30 天内合并 PR
- 6
描述
### Describe the bug
当我使用一个自定义agent来启动copilot,并且该自定义agent默认模型和cli全局的默认模型不同时,cli启动后ui右下角显示的模型会误显示为全局的默认模型,而此时实际加载的却是agent的默认模型,这会误导用户。
When I start copilot cli with a custom agent provided (e.g., --agent=orchestrator) whose default
model is different from the global default model of copilot cli, the model name shown in right-bottom corner of cli's ui will be different from the actual model loaded, which is misleading.
### Affected version
1.0.32
### Steps to reproduce the behavior
1. Add a custom agent to folder `~/.copilot/agents`. Take the oh-my-opencode-slim orchestrator (w/ opus 4.6) as an example:
```yaml
---
name: orchestrator
description: |
Multi-agent orchestrator that optimizes for quality, speed, cost, and reliability by delegating to specialist agents. Use when tasks benefit from parallel execution or specialized expertise. This agent coordinates among explorer (codebase search), oracle (architecture/code review), librarian (docs research), designer (UI/UX), and fixer (bounded implementation).
model: claude-opus-4.6
---
You are an AI coding orchestrator that optimizes for quality, speed, cost, and reliability by delegating to specialists when it provides net efficiency gains.
Blah blah blah
```
2. start copilot cli (with its global default model different from claude opus 4.6) with that custom agent
```bash
copilot --agent=orchestrator
```
3. The model actually loaded in copilot cli is claude opus 4.6, while the model name shown at right-bottom of ui is not.
### Expected behavior
_No response_
### Additional context
_No response_
贡献指南
调研方向
Start by reproducing the mismatch with the custom agent definition in ~/.copilot/agents and the entry point `copilot --agent=orchestrator`, using an agent model that differs from the global default. Trace where the CLI loads the agent model and where the bottom-right UI label gets its value; done means the displayed model matches the model actually loaded.
由索引模型根据 Issue 内容生成。
评估
- 技术栈
- shell
- 领域
- cli
- Issue 类型
- 缺陷
- 难度
- 3/5
- 预计耗时
- 1-2 天
- 活跃度
- 冷清
- 描述清晰度
- 基本清楚
- 新手友好度
- 57/100