程序员阿江(Relakkes)
814c02a786
Connecting Feishu meant creating a bot by hand on the open platform and pasting an App ID and App Secret back. Feishu also exposes an RFC 8628 device-authorization flow, so the desktop can now render a QR code, and confirming it in the app creates the bot and stores its credentials directly. `adapters/feishu/registration.ts` implements that protocol rather than importing `registerApp` from `@larksuiteoapi/node-sdk@1.73`: the repository pins 1.60 for the chat client, and the SDK runs the whole poll inside one un-cancellable promise where the desktop needs the stateless begin/poll pair the DingTalk registration already uses. The scan is create-only, so it can never rewrite the configuration of a bot the user already runs, and it pre-fills exactly the scopes, events and callbacks this adapter calls. International tenants finish on Lark's domain, which is now persisted and honoured by the client. WeCom, QQ and Slack join the same session model. WeCom and QQ bind by scanning; Slack has no scan flow, so it uses an app manifest that pre-fills the scopes and Socket Mode. All three run over long connections, so no public callback URL is needed, and all three accept private chats only — pairing authorizes one person, and answering in a group would extend that authorization to everyone else in the room. They are built on a new `adapters/common/chat-runtime.ts` instead of a fourth copy of the loop the five existing adapters each carry. A platform supplies a `ChatPort` — how to say something, how to open a streaming reply, optionally how to send an image — and the runtime owns pairing, command routing, session restore, permission bookkeeping and the translation of the server's stream. The existing five are deliberately left on their own copies; migrating them is a separate change with its own regression surface. Attachments are downloaded through a deferred loader that runs after the pairing gate and inside the per-chat queue. Resolving them eagerly would let an unpaired stranger make the adapter fetch bytes and write them under ~/.claude/im-downloads — on Slack with the bot token attached — and would let a slow attachment overtake a text message sent after it. The sidecar launcher's per-adapter branches become one table. It is declared above the mode dispatch on purpose: `runAdapters` is hoisted and runs at module top level, so a table declared below it is still in its temporal dead zone when the adapters mode reads it — which type checks, lints and unit tests all miss, and only the compiled binary reveals. Verified with the checks `check:impact` selects: adapters, server, desktop, electron, policy, chat-contract, agent-flow, docs, native (sidecar compile, packaging and an adapters-mode smoke against the real binary) and coverage. The scan flows themselves are not verified against live platforms — that needs real WeCom, QQ and Slack accounts and would create real bots. Claude-Session: https://claude.ai/code/session_01CCGoP316AK7wdQG3Ms6Uwq
5.1 KiB
title, nav_title, description, order
| title | nav_title | description | order |
|---|---|---|---|
| QQ Integration | Scan a QR code in Settings to authorize a QQ bot, then drive the Desktop app from a private chat with streaming replies. | 7 |
QQ Integration
For individuals on QQ. Select Scan to Bind in Settings, scan with the QQ mobile app, and the AppID and AppSecret land in local configuration — no manual app creation on the QQ Open Platform. Messages travel over the official WebSocket gateway, so your machine needs no public address.
Limits: private (C2C) chats only, no groups or guilds; permission approval is a text reply.
Authorize the bot by scanning
- Open Settings → IM Adapters and switch to the QQ tab.
- Select Scan to Bind. A QR code appears.
- Scan it with the QQ mobile app and confirm the authorization.
- On success the AppID and AppSecret are written to
~/.claude/adapters.jsonand the adapter restarts.
The code rotates on its own when it expires, and the image on screen refreshes with it. Once bound, the button becomes Scan Again and an Unbind bot account button appears next to it.
This step only gives the Desktop app the bot's credentials. It does not authorize anyone — that is what pairing below decides.
Pairing
Back at the top of the page, under Pairing, select Generate Code to get a six-character code. This takes effect immediately; no Save is needed.
Send that code to the newly authorized bot in a private QQ chat. Once pairing is confirmed, anything you type goes to Claude Code.
A code is valid for 60 minutes, works once, and is invalidated the moment a new one is generated. Five failed attempts within five minutes trigger rate limiting.
Allowed Users can stay empty, in which case only paired accounts can use the bot. To allow known accounts directly, enter their QQ openid values separated by commas. A QQ openid is a per-bot identifier, not a QQ number.
Supported commands
/helpor帮助— list the available commands/statusor状态— current project, model, and run state/projectsor项目列表— list recent projects and switch/newor新会话— start a fresh session, optionally with an index, name, or absolute path/clearor清空— clear the context, keep the project binding/stopor停止— stop the current turn- Permission approval: reply
1to allow once,2to allow for the session,3to deny;/allow <id>,/always <id>, and/deny <id>work too
How messages look
Replies use QQ stream messages (stream_messages): one message refreshes as the answer is generated, then settles. That API is private-chat only; if a turn has no usable reply anchor, the adapter falls back to ordinary chunked messages instead of dropping content.
A typing indicator is sent while Claude is thinking or running tools. Inbound images and files are downloaded into ~/.claude/im-downloads/qq/ and forwarded as attachments; voice notes use QQ's own transcript rather than downloading the audio. Local images referenced in the Agent's output (restricted to the current session's working directory) are uploaded and sent as separate image messages.
Agent capability and boundaries
QQ is not a separate question-and-answer model. Messages enter the same Claude Code Agent session as the current project, so they continue the multi-turn context and can use the files, terminal, Git, Skills, and MCP tools that session already has.
That means a paired account holds the full Agent capability of the current project. Permission confirmation is a gate on individual operations, not an OS sandbox. Do not hand the bot to people you do not trust, and do not install unreviewed Skills, plugins, or MCP servers from chat.
The adapter accepts only paired or allow-listed private chats, and project listing and name matching stay inside Allowed project directories.
Running it locally
Release builds start the adapter as a sidecar automatically. Manual startup is only needed when running from source:
cd adapters
bun install
bun run qq
Optional environment overrides:
export QQ_APP_ID="xxx"
export QQ_APP_SECRET="xxx"
export ADAPTER_SERVER_URL="ws://127.0.0.1:3456"
Troubleshooting
The QR code never appears. Scanning goes through the QQ Open Platform; confirm your machine can reach it. The request times out after 20 seconds, so try again.
Scanned successfully but the bot stays silent. Confirm the Desktop app is still running and that the QQ tab shows an AppID. Credentials are written the moment the scan succeeds, and the adapter restarts on its own.
Messages report "unauthorized". Check that a pairing code was generated, that it is still inside its 60-minute window, and that you sent the current one.
Mentioning the bot in a group does nothing. That is intended: only private chats are handled. Pairing authorizes one person, and answering in a group would extend that authorization to everyone in it.
Source entry points
adapters/qq/index.ts (runtime), adapters/qq/qr-auth.ts (scan authorization), adapters/qq/extract-payload.ts (inbound parsing), plus the shared session loop in adapters/common/chat-runtime.ts.