FFlow
RUStart free
← Workflow examples

FAQ bot

Takes the routine delivery, payment and refund questions off a human: one system prompt, nothing else.

Graph preview

Open in Flow →

How it works

The same skeleton as the simple chat, but the agent already carries a system prompt: answer briefly, stay on topic and do not invent facts that were not in the conversation.

  1. Entry receives the user’s question.
  2. LLM sets the model (linked to the agent via link_llm).
  3. Agent answers under the rules from the system prompt.
  4. Answer returns the text to the user.

What to configure after import

  • Rewrite the agent’s system prompt for your product: the more concrete the subject and the tone, the less the bot makes things up.
  • Pick a model in the LLM node.
  • If answers must come from documents rather than from the prompt, start with Support bot (RAG + agent) instead.