注意
Copilot SDK 当前处于 技术预览版. 功能和可用性可能会发生更改。
两种交互模式允许用户在代理正常工作时发送消息:引导 可以中途重定向代理,排队 则将消息缓冲,以便在当前轮次完成后按顺序(先入先出,FIFO)处理。
当会话正在主动处理一个轮次时,传入消息可以通过mode字段以两种方式之一传递MessageOptions。
| 模式 | 行为 | 用例 |
|---|
`"immediate"` (转向) | 注入 **当前** LLM 轮次中 | 实际上,不要创建该文件,而是使用其他方法。 |
|
"enqueue" (排队) | 在当前轮次完成后排队并处理 | “在此之后,还要修复测试” |
有关转向和排队流的序列图,请参阅 github/copilot-sdk 存储库。
转向(即时模式)
转向模块会发送一条消息,该消息直接注入到代理的当前回合中。 代理实时看到消息,并相应地调整其响应 — 这对于修正方向而无需中止过程非常有用。
import { CopilotClient } from "@github/copilot-sdk";
const client = new CopilotClient();
await client.start();
const session = await client.createSession({
model: "gpt-4.1",
onPermissionRequest: async () => ({ kind: "approved" }),
});
// Start a long-running task
const msgId = await session.send({
prompt: "Refactor the authentication module to use sessions",
});
// While the agent is working, steer it
await session.send({
prompt: "Actually, use JWT tokens instead of sessions",
mode: "immediate",
});
有关 Python、Go 和 .NET 中的示例,请参阅 github/copilot-sdk 存储库。
内部转向工作原理
- 该消息将添加到运行时的
ImmediatePromptProcessor队列中。 - 在当前轮次中的下一个 LLM 请求之前,处理器会将消息注入到会话中。
- 代理将引导消息视为新用户消息并调整其响应。
- 如果在处理转向消息之前轮次完成,该轮次会自动移动到常规队列,用于下一轮。
注意
指导信息是在当前轮次内尽力而为的。 如果代理已参与了工具调用,在该调用完成后,指令调整将生效,但仍在同一操作序列内。
排队(入队模式)
在当前轮次完成后,排队缓冲区消息将按顺序处理。 每个排队的消息启动其自己的完整处理过程。 这是默认模式 - 如果省略 mode,则 SDK 使用 "enqueue"。
import { CopilotClient } from "@github/copilot-sdk";
const client = new CopilotClient();
await client.start();
const session = await client.createSession({
model: "gpt-4.1",
onPermissionRequest: async () => ({ kind: "approved" }),
});
// Send an initial task
await session.send({ prompt: "Set up the project structure" });
// Queue follow-up tasks while the agent is busy
await session.send({
prompt: "Add unit tests for the auth module",
mode: "enqueue",
});
await session.send({
prompt: "Update the README with setup instructions",
mode: "enqueue",
});
// Messages are processed in FIFO order after each turn completes
有关 Python、Go 和 .NET 中的示例,请参阅 github/copilot-sdk 存储库。
队列在内部的工作原理
- 消息作为
QueuedItem添加到会话的itemQueue中。 - 当当前轮次完成并且会话变为空闲时,
processQueuedItems()将运行。 - 项目按 FIFO 顺序出队,每个消息都会触发一个完整的代理操作周期。
- 如果一个转向指令在转弯结束时仍处于待处理状态,则将其移动到队列的前端。
- 处理将持续进行,直至队列为空,当队列为空时,会话将触发一个空闲事件。
组合转向和排队
可以在单个会话中同时使用这两种模式。 引导会影响当前轮次,而排队消息等待自己的轮次:
const session = await client.createSession({
model: "gpt-4.1",
onPermissionRequest: async () => ({ kind: "approved" }),
});
// Start a task
await session.send({ prompt: "Refactor the database layer" });
// Steer the current work
await session.send({
prompt: "Make sure to keep backwards compatibility with the v1 API",
mode: "immediate",
});
// Queue a follow-up for after this turn
await session.send({
prompt: "Now add migration scripts for the schema changes",
mode: "enqueue",
});
有关 Python 中的示例,请参阅 github/copilot-sdk 存储库。
在转向和排队之间进行选择
| 情景 | 图案 | 为什么 |
|---|---|---|
| 代理正在沿着错误的路径前进 | 转向 | 重定向当前轮次而不丢失进度 |
| 你想到代理应该执行的事情 | 队列 | 不会中断当前工作;下一步运行 |
| 代理即将出错 | 转向 | 在错误发生之前进行干预 |
| 您想要链式执行多个任务 | 队列 | FIFO 排序可确保可预测的执行 |
| 你想要将上下文添加到当前任务 | 转向 | 代理将其合并到其当前推理中 |
| 你想要对不相关的请求进行批处理 | 队列 | 每个实例都获得其完整的轮次,并具有清晰的上下文 |
使用排队和转向构建用户界面
下面是用于生成支持这两种模式的交互式 UI 的模式:
import { CopilotClient, CopilotSession } from "@github/copilot-sdk";
interface PendingMessage {
prompt: string;
mode: "immediate" | "enqueue";
sentAt: Date;
}
class InteractiveChat {
private session: CopilotSession;
private isProcessing = false;
private pendingMessages: PendingMessage[] = [];
constructor(session: CopilotSession) {
this.session = session;
session.on((event) => {
if (event.type === "session.idle") {
this.isProcessing = false;
this.onIdle();
}
if (event.type === "assistant.message") {
this.renderMessage(event);
}
});
}
async sendMessage(prompt: string): Promise<void> {
if (!this.isProcessing) {
this.isProcessing = true;
await this.session.send({ prompt });
return;
}
// Session is busy — let the user choose how to deliver
// Your UI would present this choice (e.g., buttons, keyboard shortcuts)
}
async steer(prompt: string): Promise<void> {
this.pendingMessages.push({
prompt,
mode: "immediate",
sentAt: new Date(),
});
await this.session.send({ prompt, mode: "immediate" });
}
async enqueue(prompt: string): Promise<void> {
this.pendingMessages.push({
prompt,
mode: "enqueue",
sentAt: new Date(),
});
await this.session.send({ prompt, mode: "enqueue" });
}
private onIdle(): void {
this.pendingMessages = [];
// Update UI to show session is ready for new input
}
private renderMessage(event: unknown): void {
// Render assistant message in your UI
}
}
API 参考
可以使用 Copilot SDK 会话 API 来引导和队列会话。
消息选项
| 语言 | 领域 | 类型 | 默认 | 说明 |
|---|---|---|---|---|
| Node.js | mode | "enqueue" | "immediate" | "enqueue" | 消息传送模式 |
| Python | mode | Literal["enqueue", "immediate"] | "enqueue" | 消息传送模式 |
| Go | Mode | string | "enqueue" | 消息传送模式 |
| .NET | Mode | string? | "enqueue" | 消息传送模式 |
传递模式
| 模式 | Effect | 在活动回合期间 | 空闲期间 |
|---|---|---|---|
"enqueue" | 排队等待下一轮 | 在 FIFO 队列中等待 | 立即启动新轮次 |
"immediate" | 注入到当前回合 | 在下一次 LLM 调用之前注入 | 立即启动新轮次 |
注意
当会话处于空闲状态(未处理)时,这两种模式的行为方式相同 — 消息会立即启动新的轮次。
最佳做法
-
**默认为排队** - 对大多数消息使用 `"enqueue"` (或省略 `mode`)。 这是可预测的,不会有中断正在进行的工作的风险。 -
**保留更正指导** - 当代理正在积极执行错误操作时使用 `"immediate"` ,并且你需要在操作进一步之前对其进行重定向。 -
**保持引导消息简洁** — 代理需要快速了解方向调整。 长而复杂的转向消息可能会混淆当前上下文。 -
**不要过度转向** - 多次快速转向可能会降低驾驶表现。 如果需要显著更改方向,请考虑中止轮次并重新开始。 -
**在 UI 中显示队列状态** - 显示排队消息数,以便用户知道挂起的内容。 监听空闲事件以清除显示。 -
**处理引导到队列的回退** — 如果引导消息在操作完成后到达,它将自动移动到队列。 设计 UI 以反映此转换。