FlowiseAI
ToolAgent Not Always Calling Webhook
I'm using Flowise to build a chatbot for a production use case where only an API should be called to generate the response. I've configured a ToolAgent node with a webhook tool (ai_agent_webhook) in my flow. However, the agent sometimes skips the tool call and returns an answer directly from the LLM, even though I have explicitly instructed it not to do so.
Expected Behavior: The ToolAgent should always call the webhook tool (ai_agent_webhook) for every question.
If the webhook response is invalid or empty, the LLM should return "I don't know."
The LLM should never generate an answer from its own internal knowledge.
Actual Behavior: In some cases, the LLM bypasses the ToolAgent and responds using its own model knowledge.
The webhook is not called at all for certain user questions, even when they are valid.
This breaks the intended functionality, which requires 100% reliance on API responses.
System prompt:
Is there any other approach available? I am embedding the chatbot into my site and overriding some global variables — such as the user's account number — which should be passed to the webhook.
I think its about how you create the tool name and description, and how you prompt it: https://platform.openai.com/docs/guides/function-calling?api-mode=chat#best-practices-for-defining-functions
@HenryHengZJ can i force tool agents to be called: I see a similar ask. https://github.com/FlowiseAI/Flowise/issues/3464