ai

ai nodes call large language models (LLMs) for moderation, summarization, or personalization tasks so you can use modern AI models inside your marketing campaign.

Summary

• Placement: Mid-flow.
• Context: Client apps and backend apps.
• Visual: No.
• Level: Intermediate (requires prompt design).
• Purpose: Let flows interpret user inputs or generate dynamic copy without leaving the campaign graph.

• Supports ResultFly-managed providers (OpenAI, Anthropic, internal models) with per-node API keys.
• Lets authors set prompts, temperature, max tokens, and output targets in the $state tree.
• Emits structured errors so flows can branch on success, failure, or rate-limit scenarios.

Example input payload:

{
  "prompt": "Summarize the user's feedback in one sentence",
  "provider": "openai:gpt-4o-mini",
  "statePath": "analysis.summary",
  "variables": {
    "feedback": "{{ state.session.latest_feedback }}"
  },
  "settings": {
    "temperature": 0.3,
    "maxTokens": 150
  }
}

Success output:

{
  "status": "ok",
  "tokensUsed": 128,
  "result": {
    "text": "Users love the onboarding but want faster reward payouts."
  },
  "stateWrites": {
    "analysis.summary": "Users love the onboarding but want faster reward payouts."
  }
}

Error output:

{
  "status": "error",
  "error": {
    "code": "rate_limit",
    "message": "Provider rejected the request due to quota exhaustion."
  },
  "stateWrites": {}
}