EWW Bridge

Stream your active VS Code Copilot Chat session to Feishu / Lark as a live card — and send new tasks from Feishu back into the same chat session.

)

Chinese quick start: 中文说明

What it does

Copilot Chat ──realtime──▶ chatSessions / transcripts ──▶ EWW Bridge ──▶ Feishu card
      ▲                                                                     │
      └────────────── workbench.action.chat.open ◀──── Feishu message ◀─────┘

• Outbound: every assistant turn is mirrored to a Feishu card. The card is created instantly as a placeholder, then patched in place while the answer streams.
• Inbound: reply to the bot in Feishu and the text is appended to the current Copilot chat session and submitted.
• No public callback URL, no tunnel: Feishu long connection (WebSocket) is used.
• Does not modify Copilot or your model provider.

Features

Requirements

• VS Code 1.95+ with the built-in Copilot Chat
• A Feishu/Lark self-built app with the bot capability (see below)

Install

• Marketplace: search for EWW Bridge or install huangdan.eww-bridge
• VSIX: code --install-extension eww-bridge-x.y.z.vsix
• Source: npm install && npm run compile && npm run package

Feishu app setup (one time)

1. Feishu Open Platform → create a self-built app.
2. Add app capability → enable Bot.
3. Permissions → import:

{
  "scopes": {
    "tenant": [
      "im:message",
      "im:message.group_msg",
      "im:message.p2p_msg:readonly",
      "im:message:readonly",
      "im:message:send_as_bot",
      "im:message:update",
      "im:chat:readonly",
      "im:resource"
    ]
  }
}

4. Events → subscribe im.message.receive_v1, delivery method long connection.
5. Publish the app, then in VS Code run EWW Bridge: 设置飞书应用凭据 (App ID + App Secret).
6. Run EWW Bridge: 设置默认推送会话 and pick the chat/group to mirror into.

Commands

The output channel also logs card updates with (+XXms) API round-trip timings.

Chat commands (sent from Feishu)

Settings

See settings.example.jsonc.

Rendering mapping (VS Code → Feishu card)

Cards are capped at 24 KB of UTF-8 JSON and shrink automatically to stay under Feishu limits.

Security & privacy

• App secret is stored with VS Code SecretStorage, never in settings or logs.
• Inbound messages are rejected unless the sender's open_id is in security.allowedOpenIds.
• Everything runs locally; only card/text payloads are sent to Feishu.
• Recommended: set security.prefix so unrelated messages in the chat are ignored.

Troubleshooting

• Not connected: check the log (EWW Bridge: 显示日志), app published, long connection enabled, im.message.receive_v1 subscribed.
• No cards: ensure a default chat id is set and the bot is in that chat.
• Slow updates: model first-token latency dominates (can be tens of seconds); the card shows 已等待 Ns… heartbeats meanwhile. Feishu patch round-trip is ~0.6–1.2 s.
• Truncated look: oversized cards shrink automatically; the footer shows model and elapsed time.

Known limitations

• One VS Code window per Feishu app: multiple windows share the same app connection and Feishu load-balances events. Multi-project routing is planned.
• Copilot Chat runs one request per session, so tasks are queued, not parallel.
• Injection relies on built-in workbench commands; a hidden chat webview may delay submission by seconds.

Development

npm install
npm run compile
npm test
npm run package
# one-time: vsce login <publisher>
npm run publish -- --patch

Update publisher in package.json to your own Marketplace publisher id before publishing.

中文说明

把当前 VS Code 里 Copilot Chat 的会话实时镜像到飞书卡片，并支持从飞书直接给当前会话下发任务。

快速上手

1. 飞书开放平台创建「企业自建应用」→ 开启机器人 → 导入上文权限 → 事件订阅选「长连接」并订阅 im.message.receive_v1 → 发布应用
2. VS Code 执行 EWW Bridge: 设置飞书应用凭据（App ID/Secret），再执行 EWW Bridge: 设置默认推送会话
3. 在飞书里直接发消息即可；/status /stop /new /clear 为内置指令
4. 安全：务必配置 ewwBridge.security.allowedOpenIds，建议同时设置 security.prefix（如 #ai）

行为说明

• 每条提问先发占位卡，回答边生成边更新同一张卡
• 默认不转发思考过程；forward.thinking: true 可开启折叠面板
• 代码块统一变成「〔代码 N 行已省略〕」，空代码块（编辑占位）直接丢弃
• 卡片超过 24KB 会自动缩容，避免飞书 400/500
• 同一提问只投递一次，不会因自动重投产生重复消息

License

MIT © EWW Bridge contributors