github / github/copilot-cli

Model routing, per-agent model selection, and global hooks support

未關閉
#1,354 3 則留言 10 個 reaction 已指派 0 人 在 GitHub 檢視
area:agents area:models
主要語言
Shell
星號
11.2k
分支
1.9k
平均合併
14 小時 16 分鐘
30 天內合併 PR
6

描述

### Describe the feature or problem you'd like to solve

Copilot CLI lacks configurable model routing, per-subagent model selection (agent.md model field is ignored in CLI, see #980), and global hooks (~/.copilot/hooks/) support.

### Proposed solution

1. Configurable Model Routing
Add a model routing configuration (e.g., in ~/.copilot/config or a dedicated ~/.copilot/model-routing.json) that allows users to define rules for automatic model selection:
```json
{
"routing": {
"default": "claude-sonnet-4.5",
"rules": [
{
"when": { "agent": "explore" },
"model": "gpt-5-mini"
},
{
"when": { "agent": "task" },
"model": "gpt-5-mini"
},
{
"when": { "agent": "code-review" },
"model": "claude-opus-4.5"
},
{
"when": { "context": "planning" },
"model": "claude-opus-4.5"
}
]
}
}
```
This is similar to how tools like [oh-my-opencode](https://github.com/nicepkg/oh-my-opencode) and [oh-my-claude-code](https://github.com/nicepkg/oh-my-claude-code) provide model routing for other CLI agents. Users could optimize for cost, speed, and capability per task.

2. Respect model field in subagents
Honor the model frontmatter in custom agent definitions when running as subagents in the CLI, just as VS Code already does:
```markdown
---
name: quick-test
description: Run tests quickly with a fast model
model: gpt-5-mini
tools: ['execute']
---
```

Run the test suite and report results concisely.
When the main agent delegates to quick-test, the CLI should switch to gpt-5-mini for that subagent's execution context, then return to the parent model afterward. This enables cost-effective multi-agent orchestration.

3. Global Hooks
Support loading hooks from a global directory (~/.copilot/hooks/) in addition to the per-repository .github/hooks/. Global hooks should execute alongside (not replace) repository-level hooks, with a defined execution order (global first, then repo-local).
This enables:

Organization-wide security scanning before tool execution
Consistent audit logging across all projects
Shared pre/post-commit validation rules
Custom notification workflows

### Example prompts or workflows

_No response_

### Additional context

_No response_

貢獻指南

開啟貢獻指南

評估

這個 Issue 還沒有評估資料。

把新 issue 寄到你的電子郵件信箱

精選適合新手參與的 GitHub issue 摘要。