chenhg5/cc-connect · error

wps-agentspace: not connected

Error message

wps-agentspace: not connected

What it means

sendText returns the sentinel error "wps-agentspace: not connected" when p.conn is nil, i.e. the platform has no active WebSocket session. This happens whenever a reply/send is attempted outside the window between a successful init and the deferred conn cleanup in connect (wpsagentspace.go:379-386). It is a state error telling the caller the message could not be delivered and should be retried after reconnection.

Source

Thrown at platform/wps-agentspace/wpsagentspace.go:614

	// Dispatch to handler
	if p.handler != nil {
		msg := &core.Message{
			SessionKey: sessionKey,
			Platform:   "wps-agentspace",
			Content:    content,
			ReplyCtx:   rc,
			UserID:     chatID,
		}
		go p.handler(p, msg)
	}

	return nil
}

// sendText sends a text message to a chat.
func (p *Platform) sendText(chatID, content string, rc *replyContext) error {
	if p.conn == nil {
		return fmt.Errorf("wps-agentspace: not connected")
	}

	msg := messageData{
		Role:       "assistant",
		Type:       "answer",
		Content:    content,
		SessionID:  rc.SessionID,
		ChatID:     chatID,
		MessageID:  rc.MessageID,
		Timestamp:  time.Now().UnixMilli(),
		DeviceUUID: p.deviceUuid,
		DeviceName: p.deviceName,
	}

	return p.writeJSON("message", msg)
}

// sendTyping sends a typing indicator.

View on GitHub (pinned to 4000b2338a)

Solutions

1. Ensure the write path goes through p.sendCh / writeLoop so sends are queued and flushed on reconnect instead of requiring a live conn, or retry after connectLoop re-establishes the session
2. Check connectLoop logs for why the connection is down (dial/init/read errors upstream) and fix that root cause
3. If this happens at shutdown, treat it as expected: messages cannot be delivered after Stop()
4. Verify network connectivity to agentspace.wps.cn and valid credentials so a connection can be maintained

Example fix

// before
func (p *Platform) sendText(chatID, content string, rc *replyContext) error {
	if p.conn == nil {
		return fmt.Errorf("wps-agentspace: not connected")
	}
	...
}
// after — snapshot conn under lock to avoid racing the cleanup in connect
func (p *Platform) sendText(chatID, content string, rc *replyContext) error {
	p.mu.Lock()
	conn := p.conn
	p.mu.Unlock()
	if conn == nil {
		return fmt.Errorf("wps-agentspace: not connected")
	}
	...
}

// caller-side pre-check
p.mu.Lock()
connected := p.conn != nil
p.mu.Unlock()
if !connected { /* queue message or wait for reconnect before replying */ }

if err := platform.Send(msg); err != nil {
	if strings.Contains(err.Error(), "not connected") {
		// transient: re-enqueue and deliver after reconnect
	}
}

When it happens

Trigger: Engine replies to a user message while connectLoop is between attempts (socket dropped, backoff pending); Stop() was called and closed the connection; sendText races with the deferred cleanup that nils p.conn; the platform never connected because earlier dial/init errors kept failing.

Common situations: Agent finishes a long-running task just as the AgentSpace WebSocket drops, so the reply finds no socket; cc-connect was stopped/restarted while a pipeline was mid-reply; server outage means no connection is ever established, so every send fails with this error.

Related errors

• qq: ws connect failed (%s): %w
• dial: %s
• wps-agentspace: dial: %w
• wps-agentspace: init: %w
• yuanbao: not connected

AI-assisted analysis of chenhg5/cc-connect@4000b2338a (2026-09-06). Data as JSON: /api/errors/60a6045e92b5b1ad. Report an issue: GitHub.