openclaw - 💡(How to fix) Fix Feature request: pre-delivery message hook (messageFilter / denyMessages) [1 participants]

TL;DR

Implement a pre-delivery hook to filter out unwanted messages before they reach users, either through a proposed denyMessages configuration or a preDeliveryHook that leverages an external validation service.

Guidance

• Consider implementing a preDeliveryHook as described in Option B, which would allow for flexible message filtering using an external command or service.
• If the denyMessages approach is preferred, it could be implemented similarly to denyCommands, with a configuration option to specify denied message patterns.
• To verify the effectiveness of the hook, monitor the messages that are being sent and check for any unwanted messages that are being blocked.
• It may be necessary to modify the existing validate_outbound() function to work with the pre-delivery hook, potentially by exposing it as a REST endpoint that can be called by the hook.

Example

{
  "hooks": {
    "internal": {
      "entries": {
        "message-filter": {
          "enabled": true,
          "type": "preDelivery",
          "command": "curl -s -X POST http://localhost:18800/api/validate_outbound -d '{\"message\": \"$MESSAGE\"}'"
        }
      }
    }
  }
}

Notes

The proposed solutions assume that the validate_outbound() function is effective in detecting unwanted messages. If this function is not reliable, additional work may be needed to improve its accuracy.

Recommendation

Apply workaround by implementing a preDeliveryHook as described, as it provides the most flexibility and allows for easy integration with the existing validate_outbound() function. This approach also does not require modifying the OpenClaw configuration or codebase directly.