Dingding Tool

The plain Dingding node sends a fixed message every time your workflow runs, full stop. Sometimes that's exactly wrong - you want the agent to decide whether a situation actually warrants notifying a DingTalk (钉钉) group, and to write the message itself. Dingding Tool is that version: the same webhook mechanism, wrapped as a function the LLM can call on its own judgment instead of one that fires unconditionally.

How it works

Add it to your agent's tool list, and the model gains a callable "post to DingTalk" function. Unlike the plain send node, there's no content field here - the LLM supplies the message text itself at call time, based on whatever it decided needs to go out. Everything else about the mechanism is identical to the direct-send version: this posts through a DingTalk custom robot webhook, the same feature DingTalk group chats use for any external integration.

The inputs and outputs that matter

• msgtype (text / markdown, default markdown) - set ahead of time, same as the plain send node; the LLM's message gets formatted according to whichever you pick.
• key_word - required if your DingTalk webhook has keyword-based security enabled (one of DingTalk's own anti-abuse options for custom robots) - a message missing the configured keyword gets silently rejected by DingTalk's servers before it even reaches the chat.
• url (optional) - the webhook URL for the DingTalk group you want the agent able to post into.
• is_enable - toggle the tool off without removing it from the graph.
• One output: tool - wire into your agent's tool-list input.

How to install it

• ComfyUI Manager - search "comfyui_LLM_party", install, restart.
• Manual - cd ComfyUI/custom_nodes && git clone https://github.com/heshengtao/comfyui_LLM_party, then restart.

Then pip install -r requirements.txt from inside the comfyui_LLM_party folder, in your ComfyUI Python environment (python_embeded\python.exe -m pip install -r requirements.txt on portable Windows). Setup lives outside ComfyUI: add a custom robot to your DingTalk group and copy its webhook URL - a DingTalk admin task, not a Python dependency.

Common issues & troubleshooting

The agent posts too eagerly, or never posts at all. Because this is a judgment call left to the LLM, it's sensitive to how you describe the situation in your persona/system prompt - if you want it to only notify on genuinely important events, say so explicitly, since the model has no other signal for "important" beyond the conversation.

Message rejected with no visible error. Same as the direct-send node: check key_word against your webhook's security settings first if messages aren't landing in the chat - DingTalk enforces this server-side and won't tell ComfyUI why it dropped the message.

Treat the webhook URL like a credential. Anyone with your DingTalk webhook url can post into your group as your bot. Don't leave it filled in on a workflow JSON you're sharing - scrub it the same way you'd scrub an API key.

The pack won't load at all. Separate from this node, comfyui_LLM_party has a real, current install issue where nodes fail to import after a Manager install - usually a CUDA/PyTorch mismatch with no clear compatibility list. Check that first if Dingding_tool isn't in your node browser.